Principles of Power Measurement
Principles of
Power Measurement
A Primer on RF & Microwave Power Measurement
Principles of
Power Measurement
A Primer on RF & Microwave Power Measurement
As we constantly strive to improve our products, the information in this document gives only a general indication of the product capacity,
performance and suitability, none of which shall form part of any contract. We reserve the right to make design changes without notice.
This material has been complied from a number of sources and Wireless Telecom Group accepts no responsibility for errors or omissions.
Parent company Wireless Telecom Group © Wireless Telecom Group, 2011. All trademarks are acknowledged .
www.WirelessTelecomGroup.com
About Us
Wireless Telecom Group is a global designer and manufacturer of radio frequency (“RF”) and microwave-based products for wireless and advanced communications industries. We market our products and services worldwide under the Boonton Electronics (“Boonton”), Microlab/FXR (“Microlab”)
and Noisecom brands. Our Brands and products have maintained a reputation for their accuracy and
performance as they support our customers’ technological advancements within communications.
We offer our customers a complementary suite of high performance instruments and components
meeting a variety of standards including peak power meters, signal analyzers, noise sources, power
splitters, combiners, diplexers, noise modules and precision noise generators. We serve commercial
and government markets within the satellite, cable, radar, avionics, medical, and computing applications. We are headquartered in Parsippany, New Jersey, in the New York City metropolitan area and we
maintain a global network of Sales offices dedicated to providing excellent product support.
Wireless Telecom Group, Inc. continuously targets opportunities that allow us to capitalize on our
synergies and our talents. Our technological capabilities along with our customer service strategies
remain essential competencies for our success.
iii
Principles of Power Measurement
A Primer on RF & Microwave Power Measurement
Table of Contents
Section 1 RF and Microwave Power Measurement Fundamentals
1
Chapter 1: Power Measurement Basics
3
1.1 What is Power?
3
1.2 Why Measure Power?
5
1.3 Power Measurement History
7
Chapter 2: Power Measurement Technologies
11
2.1 Thermal RF Power Sensors
11
2.2 Detector (diode) RF Power Sensors
14
2.3 Receiver-based Amplitude Measurement
18
2.4 Monolithic Amplitude Measurement
19
2.5 Direct RF Sampling Amplitude Measurement
19
2.6 What is an RF Power Meter?
20
Chapter 3: CW, Average and Peak Power
23
3.1 CW Power Meter Limitations
23
3.2 The “Peak Power” Solution
24
3.3 It’s all about Bandwidth
25
3.4 Understanding the Importance of Dynamic Range
27
Section 2 Making Power Measurements
31
Chapter 4: Equipment Selection
33
4.1 Choosing the Right Power Meter
33
4.2 Choosing an RF Power Sensor
35
4.3 Selecting a Measurement Mode
37
4.4 Power Meters versus Spectrum Analyzers
40
4.5 Oscilloscopes and Detectors
48
Chapter 5: Calibration Issues
52
5.1 Factory Open-loop Calibration
52
5.2 Single, Double, and Multipoint Linearity Calibration
53
5.3 Field Linearity Calibration Methods
56
5.4 Frequency Response Correction
59
iv
Chapter 6: RF Power Analysis
62
6.1 Continuous Measurements
63
6.2 Triggered and Pulse Analysis
64
6.3 Statistical Power Analysis
68
Chapter 7: Power Measurement Applications
75
7.1 Low Duty-Cycle Pulse Measurements
75
7.2 Statistical Analysis of Modern Communication Signals
80
7.3 Using Power Meters for EMC Testing
87
Chapter 8: Performance Tips
91
8.1 Reducing Measurement Noise
91
8.2 Optimizing ATE Performance
94
8.3 Communication Amplifier Testing
104
Chapter 9: Measurement Accuracy
110
9.1 Introduction to Uncertainty
110
9.2 Power Measurement Uncertainty Contributions
112
9.3 Sample Uncertainty Calculations
116
Section 3 Power Measurement Reference 125
Chapter 10: Reference Tables
126
10.1 Amplitude Measurement Conversions
126
10.2 Return Loss / Reflection Coefficient / VSWR Conversions
127
10.3 Wireless and Radar/Microwave Bands
129
10.4 Sensor Cable Length Effects
129
Chapter 11: Boonton Solutions
131
11.1 4240 Series RF Power Meter
131
11.2 4530 Series RF Power Meter
132
11.3 4540 Series RF Power Meter
133
11.4 4500B RF Peak Power Analyzer
134
11.5 Boonton CW and Peak RF Power Sensors
135
11.6 Most Popular Peak Sensors
138
Table of Contents
v
Section 1
RF & Microwave Power
Measurement Fundamentals
RF Power Measurement is a broad topic that has been of importance to designers and
operators since the earliest days of wire line and wireless communication and information transmission. With today’s complex modulation schemes, increased popularity of wireless transmission and pulsed communication modes, the need to accurately
and efficiently measure RF power has become crucial to obtaining optimum performance from communication systems and components.
1
What is Power? A discussion of the physical definition of power, and the electrical concepts of volts, amps, and watts. This leads to how the measurement of AC and RF power is
complicated by complex impedance and phase shift.
Why do we want to measure Power? There are many reasons to measure RF power, spanning a wide range of industries and technologies. This subsection discusses the common
uses of power measurement instruments, and the rationale.
A brief history of RF power measurements. Power measurement has evolved considerably since the earliest days of wireless. Some of this history can be traced to well-known
radio pioneers, and much of the innovation took place among companies still involved in
the measurement industry. A considerable amount of history took place in the northern
New Jersey region that is still home to Boonton Electronics.
Power Measurement Technologies. A discussion of the key methods in use today for
measuring RF power, including Thermal, Diode, Receiver-based, Direct RF Sampling, and
Monolithic (IC) solutions.
CW versus Peak Power. Power measurement has come a long way since early methods,
which only produced meaningful measurements for unmodulated signals. This section
focuses on the limitations of various types of power meters when measuring modulated
signals, and how modern solutions have improved the situation.
Bandwidth and Dynamic Range Issues. Not every signal aligns neatly with the capabilities of power measurement instruments. By understanding the bandwidth and dynamic
range characteristics of your signal, it becomes easier to select the best measurement technology.
Section 1: RF & Microwave Power Measurement Fundamentals
2
Chapter 1: Power Measurement Basics
1.1 What is Power?
In physics terms, power is the transfer rate of energy per unit time. Just as energy has
many different forms (kinetic, potential, heat, electrical, chemical), so does power. One
mechanical definition of energy is force multiplied by distance – the force moving an object
multiplied by the distance it is moved.
Energy = Force x Distance
To get the power, or transfer rate of that energy, we divide that energy by the length of
time to perform the move. Since distance per unit time is velocity, mechanical power is
often computed as force times velocity.
Power = Force x Distance / Time
= Force x Velocity
In electrical terms, force equates to voltage, also known as Electromotive Force (EMF). This
describes how much “pressure” the electrons are under to move. The velocity is analogous
to electrical current, which is the charge (number of electrons) per unit time.
Power(electrical) = EMF x Current
EMF is typically measured in volts, and current is typically measured in amperes, or amps.
One ampere is one coulomb (a unit of charge, equal to 6.2 x 1018 electrons) per second.
Multiplying current and voltage together yields the power in watts.
Watts = Volts x Amps
In the case of steady voltage or steady current flow, computing the average power is simple
– just multiply average volts by average amps. However, if both values fluctuate, as will
be the case with alternating current, or AC, the average power can only be computed by
performing a mathematical average of the instantaneous power over one or more full signal periods.
Limiting our discussion only to sinusoidal, AC waveforms, we can see that the power will
fluctuate in synchronization with the voltage and current. For resistive loads, the current
and voltage will be in-phase. That is both will be positive at the same time, and both negative. Analyzing graphically, one can see that either case produces positive power, since
multiplying two negative numbers yields a positive result. See Fig. 1.1.1
3
If there is a phase shift between current and voltage, there will be times that the voltage
and current are of opposite polarities, resulting in a negative power flow. This has the effect of reducing the average power, even though the magnitude of the voltage and current
has not changed.
For this reason, it is not generally sufficient to simply measure the voltage or current to
characterize a signal’s power. A direct power measurement is best, in which the signal is
applied to a precision termination (load), which keeps the voltage and current very close to
in-phase. If this is done properly, a voltage measurement across the load can be performed
to yield a meaningful power value, or the dissipated power can be computed directly by
measuring the heating effect of the signal upon the load. This is discussed in great detail
in the next section.
50 ohm Resistive Load: 0 deg phase shift
Voltage
0.06
1.50
Current
1.25
Instantaneous
Power
0.04
1.00
Average
Power
0.75
Volts
0.25
0.00
0.00
-0.25
Amps, Watts
0.02
0.50
-0.02
-0.50
-0.75
-0.04
-1.00
-1.25
-1.50
-0.06
0
360
720
1080
1440
Phase (degrees)
50 ohm Complex Load: 45 deg phase shift
Voltage
1.50
0.06
Current
1.25
Instantaneous
Power
0.04
1.00
Average
Power
0.75
0.50
Volts
0.00
0.00
-0.25
Amps, Watts
0.02
0.25
-0.02
-0.50
-0.75
-0.04
-1.00
-1.25
-1.50
-0.06
0
360
720
1080
1440
Phase (degrees)
Figure 1.1.1 instantaneous and average power when voltage and current are in-phase with a resistive load (Top)
and when voltage and current are phase shifted due to complex load impedence (Bottom)
Chapter 1: Power Measurement Basics
4
1.2 Why Measure Power?
The first question is why measure power at all, rather than voltage? While it is true that
very accurate and traceable voltage measurements can be performed at DC, this becomes
more difficult with AC. At audio and low RF frequencies (below about 10 MHz), it can be
practical to individually measure the current and voltage of a signal. As frequency increases, this becomes more difficult, and a power measurement is a simpler and more accurate
method of measuring a signal’s amplitude.
As RF signals approach microwave frequencies, the propagation wavelength in conductors
becomes much smaller, and signal reflections, standing waves, and impedance mismatch
can all become very significant error sources. A properly designed power detector can
minimize these effects and allow accurate, repeatable amplitude measurements. For these
reasons, POWER has been adopted as the primary amplitude measurement quantity of any
RF or microwave signal.
There are many reasons it may be necessary to measure RF power. The most common
needs are for proof-of-design, regulatory, safety, system efficiency, and component protection purposes, but there are thousands of unique applications for which RF power measurement is required or helpful.
In the communication and wireless industries, there are usually a number of regulatory
specifications that must be met by any transmitting device, and maximum transmitted
power is almost always near the top of the list. The FCC and other regulatory agencies responsible for wireless transmissions place strict limits on how much power may be radiated
in specific bands to ensure that devices do not cause unacceptable interference to others.
Although the real need is usually to limit the actual radiated energy, the more common and
practical regulatory requirement is to specify the maximum power which may be delivered
to the transmitting antenna.
Transmission interference
5
In addition to the regulatory issues, transmitter power needs to be limited in many
communication systems to allow optimum use of wireless spectral and geographical space. If two transmitting devices are operating in the same frequency band and
physical proximity, receivers can have a more difficult time discriminating the signals if one signal is much too large relative to the other. Even in commercial broadcast, the transmitting power of each broadcast site is licensed and must be constantly
monitored to ensure that operators do not interfere with other stations occupying the
same or nearby frequencies in neighboring cities.
Controlling transmission power is particularly necessary in modern cellular networks, where
operators constantly strive to maximize system capacity and throughput. Many modern
wireless protocols use some form of multiplexing, in which multiple mobile transmitters
(for example, cellular handsets) must simultaneously transmit data to a common base station. In these situations, it is necessary to carefully monitor and control the transmitted
power of each device so that their signals arrive at the base station with approximately
equal amplitudes. If one device on a channel has too much power, it will “step on” the
transmission of other devices sharing that channel, and make it impossible for the base
station to decode those signals.
Another power control issue in cellular systems is due to the close proximity of base stations in congested areas. If a device is transmitting with too much power, it will not only
interfere with signals in its own cell, but can possibly interfere with the transmissions of
devices in neighboring cells. Mobile devices for these systems typically implement both
open-loop and closed-loop, real-time power control of their transmitters. Without accurate
power control of every single device within range of a base station, cellular network capacity can be severely degraded.
Too much power has other dangers as well. For higher power systems, too much RF power
can present biological hazards to personnel and animals. Safety limits are often placed on
transmitted power to protect users and bystanders from the dangers of high-power RF
radiation. A good example of the potential dangers of RF energy is a common microwave
oven, which can severely burn human flesh just as easily as it can heat a meal. Radio and
RADAR transmitters operate at still higher power levels, and present their own special hazards. It is hypothesized that even low-power RF transmitting devices such as cell phones
may have potential to cause lasting biological effects. In all of these cases, there will be
times when the actual power present must be monitored to ensure compliance with safety
standards or guidelines.
Chapter 1: Power Measurement Basics
6
Measuring power is important for circuit designers as well. Any electronic device can be
overloaded or damaged by too high a signal. Too much steady-state power can cause
heating effects and destroy passive and active components alike. Too much instantaneous
(“peak”) power can overstress semiconductor devices, or cause dielectric breakdown or
arcing in passive components, connectors, and cables.
But even at power levels well below the damage threshold of the circuit components, excessive power can cause overload of system, clipping, distortion, data loss or a number of
other adverse effects. Similarly, insufficient power can cause a signal to fall below the noise
floor of a transmission system, again resulting in signal degradation or loss.
1.3 Power Measurement History
Since the late 1800s, when Nikola Tesla first demonstrated wireless transmission, there has
been a need to measure the output of RF circuits. A major focus of Telsa’s work was wireless transmission of electrical power, so he was often working in the megawatt range, and
a relative indication of power was the discharge length of the “RF lightning” he produced.
For obvious reasons, there was little incentive to attempt any sort of “contact” measurement!
Around 1888, an Austrian physicist named Ernst Lecher developed his “wires” technique as
a method for measuring the frequency of an RF or microwave oscillator. The apparatus, often known as Lecher Wires, consisted of two parallel rods or wires, held a constant distance
apart, with a sliding short circuit between them. The wires formed an RF transmission line,
and by moving the shorting bar, Lecher could create standing waves in the line, resulting
in a series of the peaks and nulls. By measuring the physical distance between two peaks
or two nulls, the signal’s wavelength in the transmission line, and thus its frequency could
be calculated.
Initially, Lecher used a simple incandescent light bulb across the lines as power detector to
locate the peaks and nulls. The apparent brightness of the bulb at the peaks also gave him
a rough indication of the oscillator’s output amplitude. One of the problems with using
a bulb, however, was that the low (and variable) impedance of its filament changed the
line’s characteristics, and could affect the resonant frequency and output amplitude of the
oscillator.
This was addressed by substituting a high-impedance, gas-discharge glow tube for the
incandescent bulb. The glass tube was laid directly across the wires, and the field from a
medium-voltage RF signal was adequate to excite a glow discharge in the gas tube. This
didn’t change the tank impedance as much, while keeping it easy to visually determine the
peak and null locations as the tube was slid up and down the wires. Later, a neon bulb was
used, but the higher striking voltage of neon made the nulls difficult to locate precisely.
7
Historical photo of Tesla lightning
In 1933, H.V. Noble, a Westinghouse engineer, refined some of Tesla’s research, and was
able to transmit several hundred watts at 100 MHz a distance of ten meters or so. This
wireless RF power transmission was demonstrated at the Chicago World’s Fair at the Westinghouse exhibit. His frequencies were low enough that the transmitted and received
signal voltages could be directly measured by conventional electronic devices of the day
– vacuum tube and cat’s whisker detectors. At higher frequencies, however, these simple
methods did not work as well – the tubes and cat’s whiskers of the day simply lost rectification efficiency and repeatability.
Earnst Lecher’s apparatus for measuring RF amplitude
Chapter 1: Power Measurement Basics
8
The Varian brothers used another indicator technique in the late 1930s during their development of the Klystron. They drilled a small hole in the side of the resonant cavity and
put a fluorescent screen next to it. A glow would indicate that the device was oscillating,
and the brightness gave a very rough power indication as adjustments were made. In fact
some small transmitters manufactured into the 1960s had a small incandescent or neon
lamp in the final tank circuit for tuning. The tank was tuned for maximum lamp brightness.
These techniques all fall more under the category of RF indicators than actual measurement instruments.
The water-flow calorimeter, a common device for other uses, was adapted for higher power
RF measurements to measure the heating effect of RF energy, and found its way into use
anywhere you could install a “dummy load.” By monitoring flow of water and temperature
rise as it cooled the load, it was simple to measure long-term average power dissipated by
the load.
The thermocouple is one of the oldest ways of directly measuring low RF power levels. This
is done by measuring its heating effect upon a load, and is still in common use today for the
measurement of “true-RMS” power. Thermocouple RF ammeters have been in use since
before 1930 but were restricted to the lower frequencies. It was not until the 1970s that
thermocouples were developed that allowed their use as sensors in the VHF and Microwave
range.
In later years, thermocouples and semiconductor diodes improved both in sensitivity and
high-frequency ability. By the mid 1940s, the fragile, galena-based “cat-whisker” detectors were being replaced by stable, durable packaged diodes that could be calibrated
against known standards, and used for more general-purpose RF power measurement.
Diode-based power measurement was further improved in the 50s and 60s, and Boonton
Electronics made some notable contributions to the industry, initially in RF voltage measurement. The Model 91B was introduced in 1958 and could measure from below one millivolt to several volts. With a suitable termination, this yielded a calibrated dynamic range
of about -50 dBm to +22 dBm over a frequency range of 200 kHz to 500 MHz.
RF voltmeters and power meters continued to evolve throughout the 70s with the application of digital and microprocessor technology, but these were all “average-only” instruments and few had any ability to quantify peak measurements. When a pulsed signal had
to be characterized, the accepted technique was to use an oscilloscope and crystal detector
to view the waveform in a qualitative fashion, and perform an average power measurement on the composite signal using either a CW power meter or a higher power measurement such as a calorimeter.
9
Boonton 91C RF Voltmeter
The “slideback wattmeter” used a diode detector, and substituted a DC voltage for the RF
pulse while the pulse was off, giving a way to measure the pulse’s amplitude while compensating for duty cycle. However, a more common approach was to simply characterize a
diode detector to correct for its pulse response – a technique pioneered by Boonton Radio,
a local company that provided a great deal of technology to Boonton Electronics.
The modern realization of the peak power meter came into being in the early 1990s. Boonton Electronics, Hewlett Packard (later Agilent Technologies) and Wavetek all introduced
instruments that were specifically designed to measure pulsed or modulated signals, and
correct for non-linear response of the detector diodes in real time. These instruments have
evolved over time with the application of better detectors and high-speed digital signal
processing technology.
Chapter 1: Power Measurement Basics
10
Chapter 2: Key Power Measurement Technologies
There are a several different technologies available for the measurement of RF power.
These generally fall into four categories:
Thermal
Detector
Receiver
RF Sampling
The heating effect of RF power upon a sensing element is measured.
The RF signal is rectified or “detected” to yield a DC voltage proportional to the signal’s amplitude.
A “tuner” type circuit is used to receive the signal, then measure its amplitude component.
The RF signal is treated as a baseband AC signal, and is directly digitized.
Both thermal and detector type measurements are typically “direct sensing,” in which the
amplitude of the RF signal applied to a load element is measured by converting the RF to
an easily-measured DC quantity. This RF-to-DC conversion is typically performed close to
the signal source by connecting a small converter probe known as an RF power sensor to
the device under test.
Power Meter System
Amplifier
Cell Phone Radio
Base Station
Transmitter
RF Signal
RF Source
Thermistors
OR
Thermocouple
OR
Diode
Digitizer
Processing
Proportional DC
Signal
Terminating
Power Detector
Display
Power Meter Base Module
Direct Power Measurement Block Diagram
The receiver and RF sampling methods are usually indirect – the signal is brought into an
instrument via a cable connection, processed through a multi-stage circuit to yield amplitude information, then scaled to power.
Following is a discussion of each of these technologies.
2.1 Thermal RF Power Sensors
Thermal sensors use the incoming RF energy to produce a temperature rise in a terminating load. The temperature rise of the load is measured either directly or indirectly, and
the corresponding input power is computed. The simplest is the early “light bulb” power
detector used by Ernst Lecher in the late 1800s.
11
DC Bias Current
R
R
+
RT
RF Power
R
Thermistor
Thermistor Sensor Diagram
Thermistor (Bolometer) sensors use a thermal element, known as a thermistor, as both the
RF load and the temperature measurement device. The thermistor’s resistance changes
with temperature, making it simple to measure its temperature by detecting in-circuit resistance.
The most common implementation places the thermistor element in one corner of a wheatstone bridge, and uses a DC substitution technique, in which a controlled DC bias current
is applied to the bridge to heat the thermistor until its resistance equals that of the other
bridge resistors and the bridge is in balance. An auto-balancing circuit is used to amplify
the bridge output and drive the entire bridge with this bias signal, heating the thermistor until balance is achieved. The net effect is that the thermistor will be operated at a
constant temperature point where its resistance remains at the correct value to properly
terminate the incoming RF – typically 50 or 100 ohms.
Bolometer
Reference
Sensor
+
DC
Power
Ts
Tr
Thermal Leak
Heat
Power
Bolometer Diagram
Chapter 2: Key Power Measurement Technologies
12
The total power dissipated by the thermistor is the sum of the incoming RF power and the
power due to the DC bias. The power dissipated due to the RF heating can be computed
by subtracting the thermistor’s reference “DC-only”power (measured and stored when no
RF is applied) from its total (DC+RF) power. When the bridge is balanced, the thermistor’s
dissipation due to the DC bias is easily computed as one-quarter of the total bridge power
(bridge current multiplied by bridge voltage). The other three resistors in the bridge are
designed to have a negligible temperature coefficient of resistance.
In practical implementations, there are two identical thermistor bridges, but only one is
exposed to the RF. The second bridge is used to compensate for ambient temperature
changes.
An RF signal is applied to the terminated load of a thermocouple sensor and the rise
in temperature is measured. The rise in temperature is due to the Thermocouple
Principle. A thermocouple is formed by a metallurgical junction between two dissimilar metals which produces a small voltage in response to a temperature gradient across each metal segment – typically just a few tens of microvolts per degree C.
In a practical thermocouple power sensor, a number of thermocouples may be electrically
connected in series to form a thermopile. This increases the output voltage so it can be
more easily amplified and measured by the meter. The thermopile often forms the RF load
as well, and is connected in such a way that the RF energy flows through and heats only
one end (the “hot junction”) of each thermocouple. This is done by capacitively coupling
the RF while maintaining DC coupling for the output signal.
Iron Wire
+
Vh
-
Hot
Junction
Copper Wire
Iron Wire
Hot
Junction
Copper Wire
Diagram illustrating Thermocouple Principle
13
Cold
Junction
Weld Area
+
-
The voltage between wires is
produced by increasing the
temperature at the welded
junction
The output voltage of a thermocouple type power sensor is very linear with input power
and has a relatively long time constant due to heat flow delays. This means that they will
tend to produce a reading which is proportional to the average of the applied RF power.
Because of this, thermocouple sensors are commonly used for measuring the average power of a modulated signal. Their relatively low sensitivity, however, limits their usefulness
when the RF power level is less than several microwatts.
Thermopile
Output Voltage
Copper Wire
Iron Wire
Copper Wire
Iron Wire
Copper Wire
Iron Wire
Copper Wire
Iron Wire
Copper Wire
Iron Wire
Copper Wire
+
+
+
+
+
+
+
+
+
+
Diagram illustrating a Thermopile
2.2 Detector (Diode) RF Power Sensors
Diode sensors use high-frequency semiconductor diodes to detect the RF voltage developed across a terminating load resistor. The diodes directly perform an AC to DC conversion, and the DC voltage is measured by the power meter and scaled to produce a power
readout. In the strictest sense these are not power detectors, but rather voltage detectors,
so termination impedance variations can cause more error in the reading due to mismatch
than would be seen using thermal sensors. Early devices were simple crystal detectors using galena and a cat-whisker to form a crude diode junction.
In a diode type RF power sensor, one or more diodes perform a rectification (peak detection) function at high levels and act as a nonlinear resistor at lower levels, conducting
more current in the forward direction than reverse. This is shown in Figure 2.2.1. Usually a
smoothing capacitor is connected to the output of the diode to convert the pulsating DC to
a steady DC voltage. Often, two diodes are used so both the positive and negative carrier
cycles are detected; this makes the sensor relatively insensitive to even harmonic distortion. A diode detector’s DC output voltage is proportional to power at low signal levels and
Chapter 2: Key Power Measurement Technologies
14
proportional to the peak RF voltage at higher levels. To achieve high sensitivities, the load
resistance driven by the diode’s output is typically several megohms.
Below about -20 dBm (30mV peak carrier voltage), the RF input is not high enough to
cause the diodes to fully conduct in the forward direction. Instead, they behave as non-linear resistors as shown in Figure 2.2.2 below, and produce a DC output that is closely proportional to the square of the applied RF voltage. This is referred to as the “square-law” region
of the diode sensor. When operated in this region, the average DC output voltage will be
proportional to average RF power, even if modulation is present. This means a diode sensor
can be used to measure the average power of a modulated signal, provided the instantaneous (peak) power remains within the square-law region of the diodes at all times.
RO
RSource
VSource
RTerm
CL
RL
Vout
CL
RL
RO
Figure 2.2.1. A balanced, dual-diode sensor diagram
Above about 0 dBm (300mV peak input voltage), the diodes go into forward conduction
on each cycle of the carrier, and the peak RF voltage is held by the smoothing capacitors.
In this region, the sensor is behaving as a peak detector (also called an envelope detector), and the DC output voltage will be equal to the peak-to-peak RF input voltage minus
two diode drops. This is known as the “peak detecting” region of the diode sensor. When
operated in this region, the average DC output voltage will be proportional to the peak RF
voltage.
While the dynamic range of diode detectors is very large, operation in these two regions is
quite different and the sensor’s response is not linear across its entire dynamic range. The
square-law and peak-detecting regions, as well as the “transition region” between them
(typically from about –20 dBm to 0 dBm), must be linearized in the power meter. This linearization process does not present any difficulties for modern power meters.
15
Substituted DC
or low frequency
equivalent
-
Hot Junction
2 +
3
5
8
9
10
15
20 +
30
60
V1
+
Metal 1
-
Metal 2
Vh
-
V2
V0 = V1 + Vh V2
Seebeck Voltage
Junction
(heated)
Seebeck Voltage
DC
Power
Ts
-40
Thermal Leak
-20
300
200
Thermopile
Output Voltage
100
Copper Wire
Iron Wire
+
Copper Wire
Iron Wire
0
Copper Wire
20
40
Iron Wire
+
Heat
Power
PIN[watts]
0.01 nW
-20 dBm
400
+
Tr
0.1 nW
-70 dBm
500
261.8
271.7
283.0
303.0
330.6
339.6
344.8
364.1
390.6
526.0
714.3
Sensor
+
Noise Floor
250.0
4.00
248.5
8.05
247.9
12.1
242.7
20.6
231.9
34.5 Cold Junction
227.8
39.5
219.8
45.5
191.1
78.5
173.9
115
147.0
204
94.1
531
-3.82
-1
-7.36
-2
-10.6
-3
Iron Wire
Small voltage between wires
+
more voltage produced
as -16.5
-5
junction temperature increases.
Copper Wire
-24.2
-8
-26.5
-9
-29.0
-10
-41.2
-15
-51.2
-20
Bolometer
-57.0
-30
-70.0
-50
Reference
Vo(log)
600
R(kΩ)
Copper Wire
-100
+
Iron Wire
e (mV)
Copper Wire
Iron Wire
+
+
The Principles behind the thermocouple
1
I(nA)
+
e(mV)
+
Power Sensor
Linear Region
i (nA)
Display
+
Diode Detectors
Net RF Power
absorbed by sensor
Square Law Region of
Diode Sensor
Power
Meter
+
Thermistors
Thermocouples
Copper Wire
90
600
80
500
70
400
60
50
nA
mA
300
200
40
30
20
100
10
0
0
-100
-10
-50
-40
-30
-20
-10
-0
-10
-20
-30
-40
-50
-500
-400
-300
-200
-100
-0
100
200
300
400
500
mV
mV
Figure 2.2.2. (Top) I-V curve showing “non-linear resistor” characteristic and (Bottom) Diode I-V characteristic in
low level “square-law” region (RIGHT) and high-level “peak detecting region (LEFT)
Although very sensitive and easily linearized with digital techniques, diode sensors are
challenged by modulation when the signal’s peak amplitude exceeds the upper boundary
of the square-law region. In a case where high-level modulation is present, the RF amplitude enters the peak detecting region of the diode detector. In this situation, the detector’s
output voltage will rapidly slew towards the highest peaks, then slowly decay once the
signal drops. Since the input signal could be at any amplitude during the time the capacitor voltage is decaying, it is no longer possible to deduce the actual average power of a
modulated signal once the peak RF power gets into this peak-detecting region of the diode.
Chapter 2: Key Power Measurement Technologies
16
Graph illustrating Square-Law, Linear, and Compression Region of a Detector Circuit
One solution to this problem is to load the diode detector in such a way that the output
voltage decays more quickly, and follows the envelope fluctuations of the modulation. This
is normally done by reducing the load resistance and capacitance that follows the diodes
(RL and CL in Figure 2.2.1). If the sensor’s output accurately tracks the signal’s envelope
without significant time lag or filtering effect, then it is generally possible to properly linearize the output in real time and perform any necessary filtering on this linearized signal
(see Figure 2.2.3). This allows a sufficiently fast diode sensor to accurately measure both
the instantaneous and average power of modulated signals at any power level within the
sensor’s dynamic range. This type of sensor is commonly referred to as a Peak Power Sensor, and is discussed in greater detail in Section 4.2.
Wide Bandwidth Detector
Volts
Volts
Detector with Insufficient Bandwidth
Time
Figure 2.2.3. A Wide Bandwidth Detector Correctly Tracks a Pulse Envelope
17
Time
2.3 Receiver-Based Amplitude Measurement
In some situations, RF power is indirectly measured using a “receiver” process. The equipment may vary in type from a receiver to a spectrum analyzer, specialized test set or a VSA.
The measurement technique is similar for all of these and is essentially the same process
used in an ordinary AM radio. The input signal is coarsely tuned, and downconverted to an
intermediate frequency (IF) by combining the incoming RF with the output of a local oscillator (LO) using a mixer. Included in the mixer’s output are sum and difference products
of the original signal. The LO frequency is adjusted so that the difference product falls at
the desired intermediate frequency. This IF is then fed to one or more tuned stages, which
amplify the signal and limit its bandwidth so that only the desired input RF range is measured. The amplified and tuned IF is then either digitized directly or demodulated by some
sort of detector. (see Figure 2.3.1)
Some measurement instruments in this category, such as spectrum analyzers, can adjust or
sweep the tuning parameters of the receiving circuit, such as the tuned frequency and RF
(resolution) bandwidth. This offers considerable benefit and flexibility where information
on the signal’s spectral content is needed, but can be a hindrance when trying to perform
accurate power measurements.
The primary reason for this is that receiver-based measurements are not truly power measurements at all, but rather a measurement of the absolute amplitude of a signal’s voltage component over a specific frequency range. This narrowband, or tuned measurement
method is quite different from a wideband sensor-based power measurement, and the
reported result will often not agree with a true power measurement. The differences between power measurements performed by power meters and those performed by spectrum analyzers is discussed in detail in Section 4.4 of this guide.
Figure 2.3.1. Generic Receiver Block Diagram
Chapter 2: Key Power Measurement Technologies
18
2.4 Monolithic RF Amplitude Measurement
As discussed in Chapter 1 of this guide, part of the “wireless revolution” has been focused
on expanding wireless capacity. Part of this capacity increase comes from various types
of multiplexing schemes which allow multiple mobile devices to operate on the same upstream channel simultaneously. Many of these protocols depend upon the wireless devices
to monitor and control their transmitted power so no single device’s signal dominates the
composite signal seen by the base-station’s receiver. By balancing the received amplitudes
of all mobile transmitters, the base station can separate the individual signals.
This requirement has given rise to a family of integrated circuits designed to monitor
the amplitude of an RF signal in real time. There are several different types of IC’s that
have been introduced over the years including true RMS voltage detectors, demodulating
log amps, analog multipliers, and dedicated RSSI chips. Most share a common operating
characteristic that they have a “fast” RF input stage and output a DC voltage that is in
some way proportional to the amplitude of the input signal.
Typical RF Detection ICs
These integrated solutions are usually low-cost and often have non-linear amplitude and
frequency response. They are nearly always uncalibrated and generally tailored to a specific application. Also, most of them perform a voltage measurement function rather than
detecting true power, although a proper input circuit can terminate the signal so an equivalent power level can be computed. For these reasons, RF detection ICs are limited in their
ability to be used for general-purpose RF power measurement.
2.5 Direct RF Sampling Amplitude Measurement
In cases where the carrier frequency is low enough, it is possible to treat the signal as a
baseband AC voltage, and directly digitize it to yield amplitude information. A high speed
digitizer or digital storage oscilloscope (DSO) may be used for this purpose.
19
For accurate amplitude measurements, the sampling rate should be well above the Nyquist
rate – typically four times the carrier frequency for CW, and ten times the carrier frequency
for wideband modulated signals. For many modern communication and radar signals with
carrier frequencies approaching or into the GHz range, a fast enough sample rate to satisfy
this criteria becomes prohibitively expensive. (see Digitizer block diagram)
An alternative to the Nyquist sampling rate dilemma is to undersample the signal, while
maintaining a sufficiently high sampling bandwidth to track the carrier. This technique
requires a wide-bandwidth sample-and-hold, but does not need the A/D converter to run
as quickly. It is a viable alternative to Nyquist sampling if full reconstruction of the RF
carrier is not required for single-shot events. However, care must be taken to choose the
sample rate carefully relative to the RF carrier frequency in order to avoid aliasing artifacts.
Digitizer Block Diagram
2.6 What is an RF Power Meter?
An RF Power Meter is a precision instrument designed specifically for measurement of RF
power. It usually measures the actual power dissipated across a terminating load, and
therefore is a “single-port” device. Early RF power meters were often called RF microwattmeters, but that term is outdated and the term wattmeter usually refers to a different class
of devices, discussed below. (see RF Wattmeter)
In most cases, an RF power meter performs its task using one of the “direct” RF power
measurement techniques discussed above – either thermal or detector-based – with the
termination and power detector integrated into a single, wideband module. This module is
commonly referred to as a “power sensor” or “power head,” small enough so its input connector can be connected directly to an RF signal source without any cabling.
Chapter 2: Key Power Measurement Technologies
20
Boonton 4240 Power Meter with Sensors connected
RF power sensors are calibrated for amplitude and frequency linearity, and often contain
temperature stabilization as well. They are designed to operate at low power levels (generally less than 1 W), but are sometimes extended to medium power levels (as high as 50 W
or so) by integrating high-power input attenuators. If an input attenuator is present, the
detector and attenuator are generally calibrated as a unit to maximize accuracy. The power
sensor may or may not include active electronics following the detector.
The output of the power sensor can be single-ended or differential, and is a DC or baseband
representation of the input signal’s envelope. It is typically buffered or amplified, then
routed through a cable to the power meter base unit where it may be further conditioned
by precision analog stages, linearized and displayed. Older model power meters used an
analog meter to display the readings, but modern models will digitize, process, and analyze
the signal as needed prior to displaying the results. Because the power sensor and power
meter combination measures power directly, there is usually no input switching, RF amplification, or bandwidth limiting – all common sources of error. Therefore, they generally
provide the most accurate method for measuring the total power of an RF signal.
Another common device for measuring RF power is the RF wattmeter This is a two-port
device (input and output), in which the RF power passing through the meter is measured.
These devices are called “throughline wattmeters.” This is different from a power meter,
which has an input connector and terminates the signal. A wattmeter is useful for measuring the actual power delivered to a load or antenna, but since the load’s impedance
may vary the wattmeter does not give as good an indication of a transmitter’s capability
and is more commonly used for in-system power monitoring rather than precision power
measurement. Wattmeters usually measure power flowing in one direction (from input
to output) and may be used to measure forward or reflected power depending upon their
connection. Some have built-in controls to select which component is measured.
21
Wattmeters are more often used for higher power levels,
and in many cases can be totally passive – extracting the
power to drive the display or meter from the RF signal itself.
A power meter can always be used as a wattmeter by using a three- or four-port RF device such as a directional coupler. By choosing the coupler and attenuators appropriately,
it is possible to measure signals ranging from milliwatts to
megawatts. Using a two-channel power meter (with two
power sensors) permits simultaneous measurement of forward and reflected power. Most of these will perform ratiometric measurements between channels, which allows
return loss to be directly displayed. Some models will even
display the computed VSWR. This makes power meters very
useful for spotting problems with either the source (transmitter or power amplifier) or load (antenna) in a transmitting system.
RF Amplifier
RF Wattmeter
Power Sensor
DC Out
DUT
Power Meter
Base Unit
Power
Meter
“TERMINATED” POWER MEASUREMENT
RF Amplifier
Power Sensor
RF Out
DUT
RL
Watt
Meter
DC Out
Power Meter
Base Unit
“THROUGH” POWER MEASUREMENT
Connection Diagram of “Through” vs. “Terminated” Power Measurement
Chapter 2: Key Power Measurement Technologies
22
Chapter 3: CW, Average, and Peak Power
Classical RF power measurements are performed on steady-state CW signals. Any sort of
power detector can be linearized and corrected to yield a reasonably accurate and predictable reading with a CW signal. However, when modulation is present, additional challenges
arise. The average power of a modulated carrier with varying amplitude can be measured
accurately by a CW type power meter only if the meter is using a thermal sensor or a diode
sensor operating in its low-power, square-law region.
3.1 CW Power Meter Limitations
Traditional “CW” power meters are designed for the measurement of unmodulated or CW
signals, however they may be used with modulated signals under certain conditions, which
extends their usefulness to many applications. From a power meter’s view, any constantenvelope signal is CW, so FM or PM signals can always be measured accurately. However,
once any sort of amplitude modulation occurs, there are a number of issues that arise.
Power meters using thermal or square-law diode sensors can provide the true average
power of an envelope modulated signal, which is sufficient for most RF engineers. However,
both of these sensor types are burdened by a rather limited dynamic range, which makes it
difficult to measure complex signals. When a diode detector is used above its square-law
region with modulated signals, a CW power meter yields a non-linear and unpredictable
response.
One solution is to integrate several “true power” detectors (typically, square-law dual-diode detectors) into a single sensor package, and operate each at a different signal level.
This is done by using an integrated power splitter and scaled attenuators so that each
detector operates in its “sweet spot” over a portion of the sensor’s total dynamic range.
As long as these ranges overlap, the power meter is able to splice these detector outputs
together to yield an accurate, average power measurement over a relatively wide dynamic
range. Problems with this technique can include: mismatch issues, varying frequency response of each detector, measurement and detector range switching artifacts.
When the peak power of a pulse-modulated waveform is required, the pulse power is determined traditionally by adjusting the average power reading for the duty cycle of the modulating pulse. In addition to dynamic range limitations, this method becomes inaccurate if
the pulse shape is not ideal and useless for complex modulation. These issues are discussed
in greater detail in Chapter 7.
23
3.2 The “Peak Power” Solution
Although they can accurately measure CW power within the square-law region, CW diode
sensors cannot track rapid power changes (amplitude modulation), and will yield erroneous readings if power peaks occur that are above the square-law region. By optimizing
the sensor for response time (at the tradeoff of some low-level sensitivity), it is possible
for the diode detector to track amplitude changes due to modulation. Peak sensors use a
low-impedance load across the smoothing capacitors to discharge them very quickly when
the RF amplitude drops. This, in combination with a very small smoothing capacitance,
permits peak power sensors to achieve video bandwidths of several tens of megahertz and
risetimes in the ten nanosecond range.
Boonton 4542 RF Power Meter
It should be noted that the term video bandwidth is used to describe the frequency range
of the power envelope fluctuations or the AM component of the modulation only. If a
signal has other modulation components that do not affect the envelope (FM or phase
modulation), the frequency components of those modulating signals does not have any
direct effect on the video bandwidth unless it causes additional AM modulation as an intermodulation product. A pure FM or phase modulated signal contains very little AM, and
may be considered a CW signal for the purposes of power measurement. Power sensors are
sensitive to the amplitude of an RF signal and not to the frequency or phase.
Although the output of the sensor tracks the signal envelope, the transfer function is nonlinear – it is proportional to RF voltage at higher levels and proportional to the square of RF
voltage at lower levels. By sampling the sensor output and performing linearity correction
on each sample before any signal integration or averaging occurs, it is possible to calculate
average and peak power of a modulated signal even if the input signal does not stay within
the square-law region of the diode. Additionally, a large number of power samples can be
analyzed to yield statistics about the signal’s power distribution, and assembled into an
oscilloscope-like power-vs-time trace. (see Figure 3.2.1)
Chapter 3: CW, Average, and Peak Power
24
Figure 3.2.1. 4500B Pulse profile screen shot
Figure 3.2.2. 4540 Dual CCDF screen shot
High-speed digital signal acquisition and processing techniques have made it possible to
measure peak power as well as average power accurately with total dynamic range and
modulation bandwidth as the only limiting conditions. Knowledge of the modulation
method or modulating signal is not required. A peak power meter is accurate with CW
signals, pulsed signals such as RADAR, and modern digitally modulated communication
signals.
The term “peak” power indicates more than one might assume. Not only is the peak power
of a modulated signal available, but so is the instantaneous power at any instant in time
as well as the average power over any defined interval of time. A “peak” power meter
captures all of the amplitude-related information of a signal.
3.3 It’s All About Bandwidth
Although a peak power meter may seem like the ideal solution for measuring power of any
RF signal, there are still some caveats and tradeoffs. The most important issue is that the
sensor’s diode detector must be able to accurately track the envelope fluctuations of the
signal at all times. As was discussed in Chapter 2, the response speed of a diode sensor
may be adjusted by selection of circuit components that follow the detector diode.
If the detector is too slow, it will not accurately follow the envelope and there will be
points in time when the carrier power is unknown. This will manifest itself as measurement error, and can be in either direction depending upon the signal and detector characteristics. The rate at which a detector can follow the envelope can be described by rating the sensor’s risetime with a pulse signal or its small-signal bandwidth
with an AM signal.
25
Risetime and bandwidth are always inversely proportional, but their exact relationship is
affected to some extent by nonlinear parameters such as slew rate, as well as by high-order
filtering of the signal. For this reason, both parameters are often provided for peak power
sensors. However, a rule-of-thumb for the relationship of video bandwidth to risetime is:
Video Bandwidth = 0.35 / Risetime or Risetime = 0.35 / Video Bandwidth
The effects of limited video bandwidth are shown in Figure 3.3.1. On the left, the detector
is too slow to track the pulse envelope’s rise and fall and measurement error will result.
It should be noted that not only will the instantaneous power be wrong, but the average
power of the pulse will be incorrect. The sensor on the right has adequate video bandwidth
for the pulse to track the envelope accurately, so error is minimized.
Wide Bandwidth Detector
Volts
Volts
Detector with Insufficient Bandwidth
Time
Time
Figure 3.3.1. Effects of Sensor Video Bandwidth on detected pulse signal
Figure 3.3.1 clearly illustrates the effect of insufficient video bandwidth on power measurement accuracy, but the same errors can occur during the fast peaks and dips of modern
wide-bandwidth communication signals. When bandwidth limiting occurs on a digitallymodulated signal, the first thing that is generally affected will be the very short duration
peaks. These will be “rounded off” and the power peak will read low. The lower peak will
cause the measured peak-to-average power ratio (PAPR) to be reduced as well as changing
the statistical distribution of power levels. (CCDF, CDF or PDF, discussed in Sections 6.3 and
8.3 of this guide)(see Figure 3.2.2)
Although detector design in the sensor is usually the determining factor for video bandwidth, it is important to remember that other factors will have an effect as well. The entire
signal path that follows the detector is equally important – a chain is only as strong as its
Chapter 3: CW, Average, and Peak Power
26
weakest link. The sensor’s internal post-detection amplifier and cable driver, the sensor
cable itself, and the conditioning and conversion circuitry within the power meter can all
limit bandwidth. The bandwidths of all of the individual stages combine in inverse-squares
fashion to yield an overall “system” video bandwidth.
Power meter manufacturers typically describe the video bandwidth and risetime of a sensor as it will perform when mated with a particular power meter, and through a standardlength connecting cable. Section 10.4 of this guide includes a table showing the impact
of extended length sensor cables upon video bandwidth. USB power sensors sample or
digitize the signal within the sensor so the USB cable length and characteristics of the base
unit (if present) do not impact the video bandwidth. These sensors stop working when
the cable reaches a certain, maximum length and data can no longer be transmitted at the
required rate.
Modern peak power sensors are available with video bandwidths approaching 100 MHz,
and typical system response speeds are about 80 MHz or 4.5ns rise times with current, high
end peak power analyzers. While this bandwidth is sufficient for the majority of today’s
signals, there are still applications which will exceed these figures. Multi-carrier wireless,
high data rate satellite signals, and uncompressed video microwave links are some examples – bandwidths can be 200 MHz or more in some instances.
In these cases, a peak power meter is not usually the best option and the signal must
be analyzed with other techniques – either using a conventional, average-responding RF
power meter, or a swept frequency measurement (spectrum analyzer). Either of these
solutions will deliver the signal’s average power, but it will not be possible to capture the
instantaneous power information (time-domain and statistical) that a peak power meter
would yield. In most cases the RF power meter will offer better accuracy – the tradeoffs
between power meters and swept measurement techniques are discussed in more detail in
Section 4.4 of this guide.
3.4 The Importance of Dynamic Range
One of the most often asked questions for power measurement applications is “How low
and how high can I measure?” The answer can be as simple as making sure your expected
signal power will fall within the rated dynamic range of the sensor, but there may be more
to it than that.
For CW signals, the situation is simple. Too much power will overload the sensor – causing
either reading errors due to compression, and in severe cases, permanent sensor damage.
Too little power will cause errors due to noise and drift as the signal-to-noise ratio approaches unity. A CW diode sensor typically has 80 to 90 dB dynamic range, and can easily
accommodate a wide range of signals. (see Figure 3.4.1)
27
50
40
Thermal Sensor
“Extended” Area
30
Peak Pulse Power (dBm)
20
10
Peak Power Sensor
Dynamic Range
0
-10
-20
Avg Diode Sensor
Dynamic Range
-30
-40
-50
-60
-70
100
10
1
0.1
0.01
0.001
0.0001
0.00001
Duty Cycle (%)
Figure 3.4.1. Dynamic Range Chart showing range and various types of sensors
Once modulation is introduced, things become more complicated. Most sensors have an
average power rating for continuous signals, but many also have a peak rating for short
duration peak events – typically specifying a power level and time limit. In reality, there
is a “Safe Operating Area” curve for various duty cycles and pulse widths, but this is not
generally published.
There are few modulation-related limitations for thermal sensors – they typically can handle peaks that are well in excess of their average power rating and these peaks will average
linearly. However, very narrow duty cycle pulse signals can still exceed the sensor’s peak
power rating, so care must be taken. It is not uncommon for the average power level to be
30 dB or more below the peak power level for pulsed signals such as RADAR.
The major limitation of thermal sensors then becomes their low sensitivity. The dynamic
range is about 50 dB due to average power limits at the top, and noise floor at the bottom.
Most signals can be scaled to fit within these constraints. (see Figure 3.4.1)
When an average-responding diode sensor is used to measure modulated signals, additional concerns arise due to the inherent square-law limitations of the diode detector. The
square-law region of a diode detector has about 50 dB of usable dynamic range comparable range to a thermal sensor, but with much greater sensitivity. However, for accurate
Chapter 3: CW, Average, and Peak Power
28
measurements, the peak of the signal as well as its average must remain within the squarelaw region, so in practice a diode sensor will offer less useful dynamic range for signals with
a high peak-to-average power ratio (PAPR, also sometimes called Crest Factor).
As discussed earlier in this chapter, this limitation is sometimes addressed by integrating
multiple square-law detectors with varying attenuation into a single sensor package so
that each detector operates in its “sweet spot” over a portion of the sensor’s total dynamic range. This has the effect of “stacking” the dynamic range of each detector to yield
a wider-range composite sensor. Careful attention must be paid to matching and range
overlap, but this type of sensor can offer dynamic range approaching 80 dB for communication signals. For pulse signals, which often have a much higher PAPR, this dynamic range is
significantly reduced and the built-in detector overlap can actually leave non-linear “holes”
in the response curve. (see Figure 3.4.2)
Two-path Square Law Diode Sensor
50
40
30
20
UPPER PATH:
46db attenuation
Peak Power (dBm)
10
0
Neither path includes signals in this region
-10
-20
LOWER PATH:
6db attenuation
-30
-40
-50
-60
-70
100
10
1
0.1
0.01
0.001
0.0001
0.00001
Duty Cycle (%)
Figure 3.4.2. Dynamic range chart showing range and various types of sensors. Dynamic range of each path in a
two-path sensor is reduced for pulse signals, leaving gaps when the duty cycle is narrower than 20%. This is equivalent to a 7dB peak-to-average ratio, and is adequate for many communication signals, but few pulsed signals.
29
Peak sensors are not burdened by square-law limitations since their detectors track the
signal envelope fast enough to allow real-time linearization of the diode’s transfer function
followed by averaging. Although their peak power burnout rating is typically much higher
than the average (thermal) rating, peak sensor operation must be maintained within the
calibrated portion of the curve, usually limited by the average rating. And on the lower
end, peak sensors trade off sensitivity to yield fast response times, so net dynamic range
generally is 45 to 75 dB for peak power sensors.
Chapter 3: CW, Average, and Peak Power
30
Section 2
Making Power Measurements
Modern communication and radar signals have become complex and require advanced instruments, including RF Power Meters, to measure them.. This section will
discuss power measurement equipment and applications, and how to best match your
measurement needs to available techniques and instruments.
31
Equipment Selection. Choosing the Right Power Meter, Choosing an RF Power Sensor, Power Meters versus Spectrum Analyzers, Oscilloscopes and Detectors
Calibration Issues. Factory Open-loop Calibration, Field Linearity Calibration Methods, Single, Double, and Multipoint Linearity Calibration, Frequency Response Correction
RF Power Analysis. Continuous Measurements, Triggered and Pulse Analysis, Statistical
Power Analysis
Power Measurement Applications. Low Duty-Cycle Pulse Measurements, Measuring Modern Communication Signals, Using Power Meters for EMC Testing
Performance Tips. Reducing Measurement Noise, Optimizing ATE Performance, Amplifier
Testing
Measurement Accuracy. Introduction to Uncertainty, Power Measurement Uncertainty
Contributions, Sample Uncertainty Calculations
Section 2: Making Power Measurements
32
Chapter 4: Equipment Selection
RF power measurement can be very straightforward for simple signals, but things get complicated quickly as frequency, amplitude and modulation issues affect the accuracy of measurements. This chapter discusses the most common measurement options, and how to
align your application with available equipment.
Boonton 4500B RF Peak Power Analyzer
4.1 Choosing the Right Power Meter
Okay – so you need to measure power. The first question is what does your signal look like?
A little information on what you expect to measure is paramount to selecting the correct
equipment.
•
•
•
•
•
•
•
Frequency range – minimum and maximum carrier?
Video bandwidth – narrowband or wideband?
Rise time requirement – How fast does your pulse rise?
Dynamic range – what is the minimum and maximum expected power level?
Modulation – CW, pulse, analog or digital modulation?
Connection – connector type, coaxial or waveguide?
Impedance – 50 ohm, or something else?
Next, consider what signal measurements might be required.
• Power only, or is spectral information also needed?
• Average only (pulse power computed by duty-cycle method if needed)?
• Limited peak information (peak-to-average power ratio)?
• Time-domain measurements (pulse profiling)?
• Statistical power analysis?
33
For simple CW signals, there is a wide choice of solutions. A CW power meter is usually the
most economical choice, and can measure average power easily and accurately. If the signal is modulated, a CW power meter may still be a good choice, provided a suitable power
sensor is chosen. The first step in sensor selection is to find sensors which are compatible
with the primary characteristics of the signal to be measured – the expected minimum and
maximum power levels, carrier frequency, and source impedance. See the next section of
this guide on power sensor selection.
The chief limitation of a power meter is that it yields only amplitude information. In cases
where spectral information is also required, other solutions may be more appropriate. Vector signal analyzers, spectrum analyzers and measurement receivers are all instruments
which perform amplitude measurement while yielding information about the spectral distribution of the signal.
None of these are true power meters, since they are generally measuring narrowband voltage amplitude rather than broadband power amplitude (heating effect). However, in certain applications narrowband measurement may be preferable. Additionally, these types
of instruments often perform a swept measurement across a frequency band, and that
sweep may miss occasional signal events that generate power peaks at specific frequencies. This happens when the analyzer’s swept filter is not aligned with the center frequency
of the peak power event at the precise instant it occurs. The tradeoffs between power
meters and spectrum analyzers are discussed in detail later in this chapter.
The average power of a modulated signal can be measured by a CW or average-responding
power meter with suitable sensor, but if the user needs any sort of peak information or if
the signal has a particularly high peak-to-average power ratio, a peak power meter is often
a better choice. Peak power meters have various capabilities which must be aligned with
the signal and application to achieve accurate power measurements. Of chief importance
is video bandwidth, discussed in Chapter 3 of this guide. The sensor and power meter must
both have sufficient video bandwidth for the signal, or modulation-induced power errors
will occur.
Peak power meters nearly always measure the average power and peak power simultaneously, and usually provide the ratio between the two. Most have the ability to perform triggered waveform acquisition, and can do pulse profiling in some form. The most advanced
models offer detailed waveform analysis, sub-nanosecond time resolution, and statistical
power information.
Whether or not peak power information is necessary for modulated measurements is usually an application issue. For simple go/no-go tests in which a device is being compared to
a “known good” reading, an average power measurement is often sufficient. It will indiChapter 4: Equipment Selection
34
cate that the device being tested has an RF power at about the level expected, so is likely
functional. However, when trying to quantify performance parameters, it often becomes
necessary to measure peak power parameters, or perform signal or pulse profiling.
For pulse signals such as RADAR, average power meters have been the traditional choice.
If the pulse is very close to rectangular, its duty cycle is accurately known, and there is
minimal signal bleed and noise during the “pulse off” interval, a simple duty-cycle correction can be performed to yield the pulse power. In many cases, however, these constraints
cannot be guaranteed, and it is necessary to monitor the waveform’s shape (typically with
an oscilloscope and crystal detector) to assure that everything is as it should be.
In these cases, a peak power meter often is a more economical solution. Most can measure
and display pulse waveforms with a high degree of accuracy, providing average power,
pulse power and showing the pulse shape. More advanced instruments can measure a host
of pulse parameters such as risetime, width, overshoot, and droop. Section 7.1 of this guide
includes a discussion of the advantages using peak power meters for pulse measurements.
The need for peak measurements has expanded in recent years as digital modulation techniques have filled the wireless arena. These signals have high peak-to-average ratios, with
the highest peaks occurring relatively infrequently. This makes amplifier headroom an important parameter, since clipping the peaks will result in data loss. But since those peaks
getting clipped off don’t occur often, the impact of that clipping on the average power of
the signal can be quite small. This results in compression or clipping being rather difficult
to detect with only an average power measurement.
A peak power meter will still measure the average power accurately, but since it also continuously measures the instantaneous power, compression or clipping of the infrequent
peaks will quickly be apparent as a reduction in the peak power and the peak-to-average
ratio. Statistical methods can help to further quantify the impact of peak compression on
the signal, and can help to predict the system’s bit error rate. Statistical amplifier testing
is discussed in detail in Section 8.3 of this guide.
4.2 Choosing an RF Power Sensor
The absolute or relative power of CW signals can be accurately measured using CW diode sensors, average-responding (stacked or multipath) diode sensors, thermal sensors,
or peak power sensors. Which device you choose depends mainly on the power level and
modulation characteristics of the signal, as well as what measurement values you need to
determine. Sensor technologies were discussed in detail in Chapter 2 of this guide.
The first question is how your sensor must connect to the source being measured. Below 18
GHz, nearly all power sensors use a coaxial, type-N connectors. SMA is also used for some
35
low-cost sensors. As frequencies increase, smaller coaxial connectors are used – 3.5mm,
2.92mm, 2.4mm and 1.8mm are common sizes, providing measurements to above 60
GHz. Waveguide is another option from below 20 GHz to more than 100 GHz. Waveguide
sensors are relatively narrow band (less than one octave), so it is important to match the
sensor to the frequency band to be measured. Waveguide sensors are also more difficult to
calibrate, and are primarily limited to CW or average power measurements.
Peak Power Sensors
If the signal is always unmodulated, or if the power level never exceeds the “square law”
threshold for diode detectors (about -20 dBm), a CW Diode Sensor is an excellent choice
due to its wide dynamic range, wide RF bandwidth and true-average power detection. CW
diode sensors use high-frequency semiconductor diodes to detect the RF voltage developed across a terminating load resistor. Dual diodes are usually used to detect both the
positive and negative carrier cycles, making the sensor symmetrical, and therefore relatively insensitive to even harmonics.
CW diode sensors typically offer a lower measurement limit of about -70 dBm, and can
measure a maximum CW power of about +20 dBm before overloading. An integrated attenuator ahead of the detector assembly is sometimes used to shift this range to higher
power levels. CW diode sensors are relatively fast, offering response speeds to milliseconds
at higher power levels. As the power falls to lower level, response must be slowed considerably via filtering to yield useful results – typically around a second at -60 dBm, and even
longer to -70 dBm.
For modulated signals with peaks exceeding -20 dBm, there are several choices. Thermal
Power Sensors respond to the average power of any signal, whether CW or modulated.
Their chief drawback is that they lack the sensitivity of diode sensors – the lower measurement limit rarely extends below -30 dBm, with a maximum average power limit around
+20 dBm. However, they handle a fairly large crest factor, and can tolerate peaks well in
excess of the average power rating if the pulse width and duty cycle are short. The response speed of a thermal sensor is much slower than a diode sensor – 50ms or so even at
the highest power levels, and one second or longer below -20 dBm.
Chapter 4: Equipment Selection
36
There are also Multipath Diode Sensors that integrate multiple diode detectors and attenuating power splitters into a single, calibrated unit. These operate several pairs of detectors (usually two or three), and select the output of whichever pair is operating in its
square-law region. This has the effect of extending the true-average response of the diode
sensor to much higher power levels – and yields a device which offers nearly the dynamic
range of a CW diode sensor with close to a thermal sensor’s averaging ability. Drawbacks
include cost, slow response and noise at certain power levels, and complications due to
frequency and temperature correction differences between the detector pairs. Modern
software techniques can minimize these last two issues.
Another solution for modulated signals is the Peak Power Sensor. These offer dynamic
range between that of thermal sensors and CW diode sensors, but have extremely fast
response speeds (microseconds or less). As long as this response speed (called “video
bandwidth” and discussed in detail in Chapter 3 of this guide) is adequate, peak power sensor’s detector can faithfully follow the signal’s envelope modulation. This allows the sensor’s output to be accurately linearized and averaged by a high-speed sampler with suitable software processing. Since peak sensors continuously yield the instantaneous power
level of the signal’s envelope, a peak power meter can deliver more than just the average
power. Burst (“time gated”) power, full waveform reconstruction, pulse profile and timing,
peak power, and statistical power analysis are some of the more common measurements
provided.
4.3 Selecting a Measurement Mode
Some power meters can only handle specific types of sensors, while others offer considerable flexibility. Sensors must be used with specific measurement modes in power meters
that align with their capabilities:
•
CW or Continuous Mode: this is the basic “continuous or free-run” mode used for CW
power sensors. It returns average power of a CW signal, and will also return the aver-
age power of a modulated signal with a thermal sensor or with a diode sensor
operated within its square-law region. (see Figure 4.3.1 - Top)
•
Modulated Mode: this is a more advanced “continuous or free-run” mode. It is similar
to CW mode, but also returns limited peak information (peak, min, pk-avg ratio and
dynamic range) when a peak power sensor is used. Also may be called “free-run” mode
in peak power meters. (see Figure 4.3.1 - Top)
•
Triggered or Pulse Mode: This mode is limited to peak power sensors. Typically it
includes full pulse profiling and time-domain measurements. A signal waveform is
sometimes displayed, and the user can often select specific time intervals on the waveform to measure. (see Figure 4.3.1 - Middle)
37
•
Statistical Mode: This mode is limited to peak power sensors, and returns information
about the signal’s statistical power distribution. Sometimes these measurements are
performed as part of Modulated mode (for continuous statistical information), or as
part of Pulse Mode (yielding synchronous or gated statistical information). (see Figure
4.3.1 - Bottom)
Continuous (Modulated) Mode
Pulse (Trigger) Mode
Statistical Mode (CCDF)
Figure 4.3.1. Measurement Modes
Measurement modes are discussed in more detail in Chapter 6 of this guide, but the following guidelines may be helpful for selecting the best power sensor and measurement mode
for your signal.
Chapter 4: Equipment Selection
38
Choose a CW Diode Sensor in CW Mode for these types of measurements:
•
The signal has a low power level, below about -40 dBm.
•
The signal is CW – a simple, unmodulated RF carrier.
•
You need to measure the average power of a modulated signal whose peaks do not exceed the square-law threshold of a diode sensor (about -20 dBm).
Choose a Thermal Sensor in CW Mode for these types of measurements:
•
The signal is CW or modulated, and has an average power level that is above about
-20 dBm.
•
The signal contains a close-to-ideal pulse waveform with a narrow duty cycle and
peaks that would overload the square-law range of a CW diode or multipath sensor.
Choose a Peak Power Sensor in Modulated Mode for these types of measurements:
•
The signal has a moderate power level, above about -40 dBm.
•
This signal is continuously modulated with a video (AM or envelope) bandwidth less than about 80 MHz.
•
Signal modulation may be periodic, but only non-synchronous measurements are needed (overall average and peak power).
•
“Noise-like” digitally modulated signals such as CDMA or OFDM when only average and peak power measurements are needed. (If peak probability information is
required, consider Statistical Mode.)
Choose a Peak Power Sensor in Pulse Mode for these types of measurements:
•
Periodic or pulse waveforms with pulse power above about -40 dBm. Pulses can
be any shape.
•
Bursted signals in which power measurement must be synchronized with the
modulation.
•
Any sort of time-domain power profile or time-gated measurement is needed.
39
Choose a Peak Power Sensor in Statistical Mode for these types of measurements:
•
The signal has a moderate power level, above about -40 dBm.
•
“Noise-like” digitally modulated signals, such as CDMA (and all its extensions)
or OFDM when probability information is helpful in analyzing the signal.
•
Any modulated signal with random, infrequent peaks, when you need to know
peak probability.
4.4 Measuring Complex Modulated RF Signal: Power Meters versus Spectrum Analyzers
Instruments
A number of RF and microwave power measuring instruments have been developed to
measure a variety of signals for wireless communication, including cellular/mobile, and
commercial and Government/military RADAR applications. For simplicity, these are divided
into two categories: “tuned” and “broadband” measurement instruments.
A broadband or un-tuned power measurement uses a simple combination of a power detector and a display or recording instrument, while a tuned measurement is typically performed by a super-heterodyne, receiver-type circuit consisting of an input amplifier and/
or attenuator, local oscillator, mixer, IF filters and amplifiers, and finally, a power detector
or digitizing system. The basic receiver blocks can be combined to create instruments like
Spectrum Analyzers, Microwave Receivers, Vector Signal Analyzers (VSAs), or Cellular Radio
Test Sets.
For speed and accuracy over a wide bandwidth, the power detector/recording instrument
(power sensor/meter) combination provides either the average, or modulated carrier power of the RF signal at the input to the Sensor. Recent advances in sensor technology combined with improved computational capabilities using digital signal processing techniques
allow the Power Sensor/Meter combination to quickly and accurately measure parameters
including average power, peak and peak-to-average (PAR) power ratio, and time-domain
profiles of complex digital modulation formats used in wireless communication systems.
Although the Power Sensor/Meter instrument has no frequency selectivity, some new instruments can provide time-slotted measurements and a large color display offering great
detail for viewing the signal.
There are three typical detection circuits for modern power sensors, including thermistors,
thermocouples, and diode detectors. Each type has specific advantages and limitations,
and is discussed in detail in Chapter 2 of this guide.
Chapter 4: Equipment Selection
40
For fast rise time and wide video BW peak measurements, the RF/microwave diode detector is often the best choice. The diode detector also excels where high sensitivity and
wide dynamic range is required. Thermistor (bolometer) and thermocouple detectors can
provide accurate, true average power measurements of both CW and wideband signals for
calibration applications. (see Figure 4.4.1)
Unlike Power Meters, super-heterodyne type instruments, such as Spectrum Analyzers, offer frequency selectivity with limited bandwidth. These instruments are designed to perform relatively narrow band power measurements while placing emphasis on the spectral
power distribution. Recent progress in high speed digitizing and digital signal processing
has improved the accuracy and functionality of these instruments for the measurement of
digitally modulated signals. Some of the latest analyzers provide time-domain and statistical power analysis as well. (see Figure 4.4.2 and 4.4.3)
Uncertainty & Error Sources
There are many sources of error and uncertainty when measuring the power of any RF
signal, but these issues are further complicated by the wide bandwidth and dynamic range
of complex modulated RF/Microwave signals. For CW signals, RF power meters are the
accepted standards for accurate power measurement, but the choice is not so clear when
modulation is added. In this case, there is a temptation to use a spectrum analyzer due to
its ability to characterize the signal’s power distribution over frequency, but in addition to
considerably higher cost, there are tradeoffs involved that can often reduce the measurement accuracy.
This section will explain why a Peak Power Meter is often the most cost effective choice for
wideband, digitally modulated signals, offering unmatched speed and accuracy.
Accuracy is important for every test engineer, but there are many sources of error and
uncertainty that contribute to inaccurate power measurements. It is essential when making measurements that errors and uncertainties are understood and either corrected, or
accounted for to improve measurement accuracy.
In the RF Peak Power Meter, items contributing to errors include signal source mismatch,
power reference and detector nonlinearity, calibration factor uncertainty, and instrumentation noise and drift. Source mismatch errors are caused by impedance variation between
the RF output of the signal source and the input of the power sensor. For most measurements, mismatch is the single largest source of error and maintaining the best possible
match between the source and sensor will improve accuracy.
Sensor linearity is the next largest source of error for diode based Sensors used for most
wide bandwidth and high dynamic range measurement requirements. Two diodes in a full41
wave rectifying arrangement are used to improve sensor linearity as shown in Figure 4.4.1.
For signals below –20 dBm down to the sensor noise floor, the diode circuit has a linear
response and requires minimal linearity correction. For signals above –20 dBm the sensor
must use linearity correction techniques to compensate for diode nonlinearity.
RO
RSource
RTerm
VSource
RL
CL
Vout
RL
CL
RO
Figure 4.4.1. Dual Diode RF Power Sensor Configuration
Linearity correction tables for each RF power sensor are usually programmed into the sensor during factory calibration. These tables allow the instrument to display the correct
power reading for all input levels, even though the sensor’s response may not be perfectly
linear with applied power.
RF Signal
Low Pass
Filter
Mixer
IF Amp IF Filter
Envelope
Detector
RF Attenuator
Y
Local
Oscillator
X
Ramp Generator
for Sweep
CRT
Figure 4.4.2. Typical front-end of a Superheterodyne Spectrum Analyzer
Chapter 4: Equipment Selection
42
To correct for small sensor and meter response changes at the time of measurement, most
high quality power meters include a built-in, traceable calibration source or precision power
reference. Power sensor calibration methods are discussed in detail in Chapter 5 of this
guide. The combination of factory and field calibration processes can generally provide the
best correction for all types of sensor nonlinearities over the entire dynamic range of the
instrument to minimize errors.
Figure 4.4.2 is a diagram of a basic analog spectrum analyzer. Like the power meter, this
includes a power detector and measurement circuitry, but many additional components are
necessary in a spectrum analyzer to provide the frequency-selective functionality. These
additional components – attenuators, oscillators, mixers, filters and amplifiers all contribute to the uncertainty of the power measurement performed by the instrument. Modern
spectrum analyzers can significantly reduce errors by performing some of these functions
digitally, discussed in detail below.
The incoming RF signal is typically applied to a wideband variable input attenuator to reduce its amplitude to an acceptable level for the mixer stage that follows. Variable, or step
attenuators rarely achieve excellent RF matching over a wide frequency range, so mismatch error will be the first uncertainty seen by the incoming signal. The step attenuator is
characterized and calibrated for attenuation versus frequency at each attenuator setting,
but there can still be a fair amount of uncertainty in this process.
Additional uncertainty is caused when switching between attenuator settings, since each
setting presents a different input impedance to the user’s signal source. This results in a
mismatch loss that is different for each attenuator setting. These mismatch loss variations are calibrated out only if the complex source impedance of the signal being measured
is the same as that of the calibration signal used for attenuator characterization. If the
impedances are not identical, the actual mismatch loss at each attenuation setting will be
different from the stored calibration value, and the power reading will change each time
the attenuator range is changed.
Power meters do not have a narrow dynamic range mixer and do not require a step attenuator input. The RF signal source is applied to the power detector (and its built-in precision
termination) either directly, or through a single, fixed attenuator. In addition to being able
to maintain a very good match over a wide frequency range by removing the attenuator
switching circuitry, the power sensor presents a fixed rather than variable termination to
the RF source. While this does not eliminate mismatch loss, the loss remains constant over
the full dynamic range of the signal, greatly improving power measurement linearity.
Power meters have additional accuracy benefits over spectrum analyzers because they do
not have a heterodyne stage – the local oscillator, mixer and filter. All of these components
43
contribute to power measurement uncertainty. The mixer is a nonlinear device, and like
a diode detector, its transfer function must be carefully characterized for frequency, amplitude and temperature. A typical mixer has poor RF matching, but because it is isolated
from the RF source by the input attenuator, the mixer’s poor match is only “seen” by the
source at the analyzer’s most sensitive setting. The LO, IF and detector stages must be
similarly characterized.
For the classic spectrum analyzer, the input attenuator, multiple frequency bands, local
oscillators (LO), mixers, intermediate frequency (IF) amplifiers and filters (including the
“RBW” filter) are all analog components, and contribute to uncertainty. Modern instruments implement some stages digitally reducing the uncertainly.
Depending upon the exact design, newer digital versions as illustrated in Figure 4.4.3 have
similar attenuator and filter issues, but eliminate some LO and mixer issues. The digital
process adds other sources of error including ADC quantization error, and sample and hold
circuit jitter, but careful design can minimize or at least compensate for these effects.
The analyzer’s RBW filter may be adequate for a single frequency, continuous wave (CW)
power measurement, but for digitally modulated wide band signals, portions of the channel’s spectrum, as well as portions of the adjacent channel’s signal may or may not be
excluded by the filter, reducing power measurement accuracy.
Attn
Micro-Processor
Display
Low Pass
Filter Sampler
Input
ADC
FFT
fs
Figure 4.4.3. Digital FFT type Spectrum Analzyer
The level of measurement uncertainty is greater for a spectrum analyzer than a peak power
meter because of additional measurement stages, but for many users this accuracy tradeoff is offset by the additional functionality offered by the analyzer’s frequency selectivity.
Chapter 4: Equipment Selection
44
Benefits & Limitations of the Power Meter
Today’s RF peak power meter is both fast and accurate, while providing true power measurements of the signal in the time domain. Whether pulsed RF or complex digitally modulated waveforms, the peak power meter is designed to accurately measure peak and average power levels.
Complex modulated signals like code division multiple access (CDMA) are noise-like with
random power peaks and require a statistical approach for proper measurement. Built-in
digital signal processing capabilities allow the peak power meter to quickly perform statistical analysis on these complex signals.
The digital signal processing circuitry in Boonton peak power analyzers can continuously
acquire and process power readings at sustained sample rates up to 25 MSa/sec and triggered burst rates of 50 MSa/sec. These measurements are quickly displayed in a statistically meaningful probability density function (PDF), cumulative distribution function (CDF)
or a complementary CDF (CCDF). These functions show the number of power measurement
events at various levels (Figure 4.4.4.). The CDF and CCDF illustrate the fraction of time
transmitter crest factor is above (or below) a desired level. This can show in time when the
amplifier output is being clipped and compressed. This is useful during the design stage of
the amplifier system for size and power requirements, or to take corrective action during
operation of the amplifier to maintain optimum transmitter output power. Section 6.3 of
this guide contains an in-depth discussion of statistical power analysis.
CCDF
y
+20 Wmax Wavg Power
dbm
-30 -
-70 -
0
pr1
50
100
PROBABILITY (%)
Figure 4.4.4. Statistical occurrence of different power levels
45
Boonton meters have two adjustable markers to limit the area of measurement and can
read the power at any point across the waveform as shown in Figure 4.4.5. This feature can
be used to identify the maximum, or minimum peak power, long term average power and
PAR ratio in specific areas, usually a particular channel or time slot. Boonton instruments
are equipped with oscilloscope-like triggering capabilities to capture particular waveform
segments in signals. Section 6.2 of this guide discusses triggered and pulse measurements
with peak power meters.
Figure 4.4.5. Pulse droop on Channel 1 and Math channel showing input-to-output ratio gain change over the
pulse duration
Excellent input matching over the entire sensor frequency range will minimize mismatch
error to improve accuracy. Combined with temperature drift compensation the modern
peak power meters can accurately measure absolute power to within a fraction of a dB, and
relative power to within hundredths of a dB.
Unlike a tuned instrument, the power meter cannot provide information about the carrier
frequency or spectral content of the signal, or measure power within a specific channel
bandwidth. Rather, the peak power meter provides a cost effective way to obtain time
domain (pulse) measurements, average and peak power, and power statistics such as the
complementary cumulative distribution functions (CCDF) of complex modulated signals.
Benefits and limitations of the Spectrum Analyzer
The primary advantage of using superheterodyne instruments like Spectrum Analyzers
is the ability to limit the power measurement to a desired frequency range or a specific
channel in presence of adjacent channels. With dramatic improvements in digital signal
processing (DSP), microelectronics and linear amplifiers, DSP based Spectrum Analyzers
have improved measurement accuracy for many signal types, including the latest complex
Chapter 4: Equipment Selection
46
digital modulation formats. These units can display the waveform’s amplitude versus frequency in the band of interest, but the capability comes at a high cost, typically two to four
times that of a high performance peak power meter.
However, the frequency tuning capability in a spectrum analyzer can be a hindrance when
attempting to measure the power of a wideband signal. Figure 4.4.6 illustrates spectrum
analyzer limitations using RBW filter settings narrower than the bandwidth of the signal,
where power is added from the adjacent channel that will give an inaccurate reading. Because the measurement is band limited, some noise power is excluded from the area of
measurement to reduce accuracy.
In example (1) a CW signal can be accurately characterized with a swept
or stationary center frequency and
constant RBW setting.
In example (2), the RBW is too narrow for the signal, so the center frequency must be swept and the power
integrated to measure the total power of the signal’s entire spectrum.
This sweep results in a measurement
time much greater than performing
the same measurement with a peak
power meter.
Example (3) shows a difficult measurement. The RBW cannot be made
wide enough to encompass all three
channels, but it is not narrow enough
to totally exclude the adjacent channels. Here, it will be necessary to use
a very narrow RBW setting (and very
slow sweep) to perform the measurement.
POWER
Resolution BW (5 MHz)
1
CW
ƒ
POWER
Resolution BW (5 MHz)
2
10 MHz WiMax Signal
ƒ
POWER
5 MHz Resolution BW
1.25 MHz WiMax Signal
3
ƒ
Figure 4.4.6. The accuracy of measurement is
dependent on the Spectrum Analyzer’s resolution bandwidth (RBW) filter
47
Magnitude accuracy is also reduced by sweeping the signal at high speed to capture fast
transitioning signals. As a frequency domain instrument, it lacks the wide instantaneous
dynamic range necessary to provide meaningful statistical data for a PDF, CDF, or a CCDF,
vital information required for today’s complex digitally modulated communication signals.
Summary
Depending on the application, both broadband and tuned instruments are needed to measure the power of complex RF and microwave signals.
For pulsed power applications like RADAR, the peak power meter is the clear choice for fast,
accurate time domain measurements. Fast rise time, low duty cycle signals require a wide
dynamic range to measure large peak to average power level differences.
For communication signals such as the noise-like modulation of a CDMA signal the peak
power meter is an attractive proposition because of its wide dynamic range, wide video
bandwidth, and statistical analysis capability.
When frequency or spectral information is not required, the RF Peak Power Meter provides
the best combination of speed and accuracy over a wide range of frequencies, and at a
price well below many Spectrum Analyzers. In contrast, the spectrum analyzer provides
frequency content and selectivity that cannot be offered by power meters.
4.5 Oscilloscopes and Detectors
Before modern peak power meters were available, a system that included a crystal detector, oscilloscope, average power meter, pulse generator and assorted couplers was assembled to capture pulse waveforms from high power amplifiers used for RADAR signals.
Figure 4.5.1 is the block diagram of diode (crystal) detector system.
The CW input signal is connected to the pulse amplifier (DUT) input and pulse gated via the
connected generator for a pulsed radar output signal. The signal is passed through a directional coupler to either a dummy load, or actual antenna, and the crystal detector system.
The test signal is then split between an average power meter, and a crystal (envelope) detector connected to the oscilloscope. The CW power meter will provide an absolute average
power measurement, while the scope will provide the pulse envelope shape. The duty cycle
is calculated by dividing the pulse repetition interval by the power envelope pulse width.
The Pulse power is then calculated by dividing the average power value with the duty cycle
measurement (see Figure 4.5.2).
Chapter 4: Equipment Selection
48
Pulsed RF out
DUT
CW RF in
pulse amplifier
diode
detector
pulse
generator
pulse gate
load
coupler
splitter
CW
power
sensor
power
meter
Calibrated power value
oscilloscope
Figure 4.5.1. Typical high power pulse power measurement system using conventional methods
Pulse
Repetition
Interval
(PRI)
Power Envelope
Measured
Average
Power
(pavg)
Pulse
Width
(PW)
Duty Cycle = PW / PRI
Pulsed Power = Pavg / Duty Cycle
Figure 4.5.2. Relationships between Average Power, Pulse Power and Duty Cycle
This calculation assumes constant power during the pulse-on interval, a perfectly rectangular pulse envelope, and a constant duty cycle. Errors frequently creep into the pulse power
calculatino due to common pulse waveform anomalies like overshoot and ringing, or slow
edge transitions. Figure 4.5.3 contains several examples of distorted pulse shapes.
49
Overshoot and
Undershoot
Droop
Rise Time
Pulse
Power
Droop
Figure 4.5.3. Distorted Pulse Waveforms
A modern peak power sensor and a crystal detector circuit are illustrated in Figure 4.5.4.
The top figure is a typical single-ended detector circuit with uncertainty factors that include an uncalibrated detector with limited dynamic range and a fairly high noise floor. The
half wave rectified input cannot accurately measure asymmetretrical waveforms, and is affected by harmonic content. Matching to the RF source becomes difficult due to the parallel
effect of the output load impedance. This load is necessary to achieve fast pulse response,
and can either be the oscilloscope’s internal 50-ohm termination, or an external resistor. A
portion of this impedance appears in parallel with the detector’s input termination, which
affects the input VSWR. The effect is very small at low input levels, and becomes prounounced at high RF power inputs. Although the single ended detector can be calibrated
using simple algorithms to improve the measurement, the dual diode differential sensor
has several important advantages.
diode
detector
RF in
50 ohm
peak-detect
diode
Half wave
Rectified
scope
differential
peak-detect
RF in
Full wave
Rectified
50 ohm
dual diodes
Figure 4.5.4. Typical Oscilloscope and Power Meter Detector Configurations
Chapter 4: Equipment Selection
50
The differential system with two balanced diodes measures the fully rectified waveform.
This improves linearity and measurement response time and cancels most waveform asymmetry for true signal envelope detection. The differential configuration reduces common
mode noise, lowering the sensor noise floor while increasing dynamic range. Full-wave detection also significantly improves accuracy when measuring signals with high even harmonic content. Further improvements are made by accurately calibrating the sensor at
several power levels spanning the sensors linear range (a diode detector’s square law region). Modern instruments correct for non-linearity outside of that range, offering calibrated measurements up to the detector’s power limit (see section 3.4 on Sensor Measurement
Range). A modern two channel peak power meter offers the additional benefit of allowing
simultaneous forward and reflected power measurements illustrated in the figure below.
In Figure 4.5.5, Channel 1 is the Radar output or reference signal and channel two is the
return, or reflected power from the antenna. This value can be used to measure antenna
efficiency (return loss, S11). The extra dynamic range provided by a calibrated dual diode
sensor allows the user to measure forward and reflected power on the same instrument.
In contrast, a traditional detector feeding an oscilloscope in conjunction with the average
power meter limits dynamic range and requires splitting of the input signal to perform a
single measurement.
The two channel peak power meter provides a simpler, more convenient method to measure radar antenna return loss.
Radar Receiver
Directional
Coupler
Reflected Signal
dc
Radar
Transceiver
Antenna
Radar Transmitter
Boonton 4500B
dc
Figure 4.5.5. Power Meter Radar Antenna Return Loss Measurement
51
Radar Signal
Chapter 5: Calibration Issues
Measurements are meaningless if they are not accurate, and even with the correct equipment, performing accurate power measurements requires a valid calibration. Power meters are faced with the task of delivering a precise power reading under varying conditions.
The instrument and power sensor must both be calibrated so that the resulting power
reading closely agrees with the actual RF input power, regardless of various instrumentation and external variables.
This chapter discusses the linearity and frequency calibrations that are an integral part of
the power measurement process. Most power meters also include temperature compensation or correction – this is typically handled as part of the linearity correction process, and
therefore is not discussed separately.
5.1 Factory Open-Loop Calibration
The simplest calibration method dates back to early days of measurement, and involves
applying known signal levels and physically marking the location of the indicator needle
on the face of meter movement at each power step, then scribing a few marks between if
needed. This creates a primitive “look-up table” which compensates for the gain and linearity (shape) of the transfer function of the combined sensor and power meter.
As time went on, the meter face markings became fixed, and adjustments were performed
internally via a series of analog potentiometers, which could adjust gain and shaping for
each of the power meter’s ranges. Nonlinear networks of diodes and resistors would allow
moderately accurate correction for the curvature of the transfer function. In later years,
discrete digital circuitry replaced these nonlinear networks for the “shaping” function.
Again, this required that the sensor and meter be calibrated together as a unit.
Since most modern power meters are equipped to handle removable sensors, it now makes
sense to calibrate the meter and sensor separately so that either can be interchanged without invalidating the calibration. To calibrate the power meter, it is connected to one or
more precision DC reference voltages that simulate the output of a power sensor, and the
meter is adjusted to standardized gain values across its operating range. This process is often called DC calibration, and ensures that a particular sensor will produce the same reading
on any calibrated power meter. It may take the form of a physical adjustment of analog potentiometers, or as a digital adjustment of gain and shaping coefficients through software.
But because every power sensor is different, a power meter must also know the precise relationship between the RF input amplitude and the detected output voltage from the connected sensor. This information can range from one or two numbers to multi-dimensional
tables of calibration data. Information about the sensor’s transfer function can be characChapter 5: Calibration Issues
52
terized at the factory, and stored in a nonvolatile EEPROM in the sensor. When the sensor
is connected, this sensor calibration data is downloaded by the power meter and used as
the basis for computing the proper RF power to display for a given output signal from the
sensor. (see Figure 5.1.1)
RF Amplifier
Detector
Digitizer
Linearizer
V in
ADC
DUT
DC Signal
Amplifier
Processor
Display
Power Meter
Base Unit
LCD
P out
Figure 5.1.1. Signal Flow Block Diagram of sensor to meter to shaping to display
Some power sensor types, such as thermocouple sensors, are very linear, producing an output voltage that is directly proportional to input RF power. CW diode sensors operating
within the square-law region (below approximately -20 dBm) also share this characteristic.
Linear power detectors like this are very simple to calibrate, and for some measurements
all that is necessary is to store the “transfer gain” – just a single value of “microvolts per
milliwatt” for moderate accuracy.
As the transfer function becomes nonlinear or increased accuracy is required, more complex
equations are necessary to characterize the transfer function. Common techniques include
polynomial curve fits of segments or the entire function. If calibration points are close
enough together, even a second-order curve fit can produce excellent results.
5.2 Single, Double and Multipoint Linearity Calibrations
For best possible accuracy in the field, many power meters include the ability to perform
fine adjustments on the factory calibration by using a local reference level. There are three
basic types of field adjustment:
•
Single-point adjustment at zero input – referred to as a “Sensor Zero” or “null
adjustment,” and still uses stored data to characterize detector and instrumentation gain and detector linearity (“shaping”). No RF reference is used.
•
Two point adjustment at zero input and a mid- or full-scale power reference. Calibrate
gain and offset, but still uses stored factory linearity data for transfer function shape.
•
Multi-point adjustment at a series of power values. This process uses a series of
power steps across the sensor’s full dynamic range to fully characterize the detector’s
transfer function, and can replace or enhance the stored factory transfer function.
53
The zero adjustment is supported by nearly all instruments, and is necessary for performing low-level power measurements. For best accuracy, a zero is typically performed immediately before any measurement in the lowest 10 dB or so of a sensor’s dynamic range.
This minimizes the contribution of sensor drift and other phenomena which cannot be
characterized in the sensor’s factory calibration. The zero adjustment usually just takes a
highly-filtered reading over a relatively long time interval (often several seconds or more)
and uses this value as an offset value. The stored factory transfer function is simply adjusted up or down (offset) by the sensor zero.
Screen shot of “sensor zeroing”
The two-point adjustment consists of a sensor zero (offset adjustment) plus a gain adjustment using a fixed power reference. Many power meters include a built-in 1.00 mW
(0 dBm) RF reference output for this purpose. By adjusting the gain of the curve, small
variations in attenuation due to sensor drift and aging, and connector wear can be compensated for. This process is sometimes called a “fixed cal,” and will affect all power values
equally – by a fixed percentage in linear power (milliwatts), or by a constant number of
dBm in logarithmic units. The stored factory transfer function is still used, but is adjusted
for both slope and offset.
A multi-point adjustment further improves upon the adjustability permitted by zero and
fixed cal adjustments. This field procedure is known as a “step cal” or “Auto-Cal” in Boonton power meters. Instruments that support step calibration replace the fixed 0 dBm power reference with a precision power sweep calibrator that can generate precise, calibrated
RF levels over a wide dynamic range. The calibrator is stepped through the sensor’s entire
dynamic range, and a series of calibration values are stored for the exact connector/sensor/cable/instrument assembly in use, and at the current operating temperature.
Chapter 5: Calibration Issues
54
Sensor AutoCal sweep in progress
Depending upon how many power steps are used for the calibration sweep, the factory
transfer function can be adjusted at all points, or simply replaced. In either case, a number
of uncertainties can be reduced or eliminated by “closing the loop” in the field, discussed
more in the following section. Here are some common characteristics of performing a field
step calibration:
•
Provides both gain and offset adjustment
•
Fine-tunes the transfer function over sensor’s entire dynamic range
•
Provides an improved “current temperature” calibration beyond the factory-characterized compensation tables
•
Compensates for detector aging and degradation (temporary, small overloads, ESD or
physical shock can cause slight changes to detector transfer functions)
•
Compensates for changing losses due to connector wear
55
5.3 Field Linearity Calibration Methods
All of the field linearity calibration methods discussed so far rely on the application of a
known RF power reference to the sensor input to allow the sensor/power combination to
yield the most accurate readings possible. However, when looking at the sources of potential inaccuracy and drift, it should be apparent that there are a number of stages between
the input RF signal and the digitized detector value, and all of these must be accounted for
when the system is calibrated.
Traditionally, power meter base units and sensors are separately calibrated as discussed
above in Section 5.1, and then the field calibration (zero, fixed cal, or step cal) is used to
calibrate out the small errors that occur when a sensor is mated to a power meter, as discussed in Section 5.2. The basic function of a power sensor is generally to convert RF power
to DC voltage, while the function of a power meter is to convert that DC to a meaningful
power reading.
The power sensor is factory characterized with one or more linearity tables that describe
its transfer function, or in some cases, deviations from a stored “default” transfer function.
Depending on the sensor, these tables may have multiple inputs to allow compensation
for temperature and frequency, both of which can strongly affect the shape of the curve.
But in all cases, the sensor’s calibration tables describe how its DC output (sometimes DC
chopped to AC) relates to the RF input.
The power meter’s job is simple: measure the sensor’s DC output, and use the appropriate
transfer function to linearize, or “shape” the value in milliwatts or dBm. This requires that
the power meter know exactly what voltage the sensor is outputting, and this is where the
calibration of the power meter base unit is required. Usually, this calibration is performed
by connecting a precision DC source to the meter’s sensor input, and calibrating at one or
more voltage points. This way, when the sensor is connected, the entire system will be in
a calibrated state.
However, this process is not perfect – there can be small losses due to cables and connectors between the sensor and power meter, noise offset, as well as drift in the power
meter’s analog stages. This drift and uncertainty is usually referred to as “instrumentation
uncertainty” (see Chapter 9 of this guide), and will add to the drift and uncertainty of the
detector itself.
Chapter 5: Calibration Issues
56
This uncertainty can be reduced by any of three methods:
•
“Closed Loop” field calibration with a known RF level applied to the detector input
(sensor input connector)(see Figure 5.3.1)
•
Field calibration with a known DC level injected immediately after the detector (see
Figure 5.3.2)
•
Digitizing the signal immediately after the detector, and transmitting digital data
to the base unit
The first method is discussed in the preceding sections. This typically uses a precision RF
power reference or calibrator that is built into the power meter. When this technique is
used, the entire measurement path is inside the calibration loop, and the overall accuracy
becomes dependent upon the accuracy of the RF calibrator. The advantages of RF calibration are that connector and detector changes will be calibrated out, and that a faulty
or blown sensor will be immediately apparent. In the case of step calibration, the entire
dynamic range of the sensor can be fully calibrated in the field. This technique usually produces the highest obtainable measurement accuracy. The chief disadvantage is that the
sensor must be disconnected from the source and connected to the calibrator every time
field calibration must be performed. This can be inconvenient in some automated systems.
RF Amplifier
Power Sensor
SP3T RF Switch
Cable
DUT
Calibration
Source
Calibration
Control Line
Figure 5.3.1. Block Diagram of the Calibrator Method
57
Power Meter
Base Unit
The second method bypasses and disconnects the detector and injects a precision DC level
into the signal chain in place of the detector output. This leaves the input connector and
detector itself out of the calibration loop. However, since it is easier to generate a stable,
precise DC voltage than an equally stable and precise RF power level, some of the uncertainty due to not calibrating the entire signal chain is offset by a reduction in uncertainty
of the calibration source. The primary advantage of this arrangement is that the sensor
can remain connected to the device under test during calibration. And of course, the disadvantage is that drift or malfunction due to connector or detector aging or damage will
go unnoticed. Since factory-generated linearity data must be used for shaping, detector
linearity changes cannot be compensated for. This reduces accuracy somewhat compared
to a closed-loop RF calibration.
The third method is used by most USB sensors, and reduces the need for field calibration
because there are relatively few analog stages between the detector and the digitizer.
When a digitizing sensor is factory calibrated, the entire analog signal chain is present, and
since this circuitry is within the same housing and the sensor output is digital data, it is possible to more accurately characterize the system at the factory. There are still errors that
can crop up – not all USB ports deliver the same power supply voltage to the sensor, many
are notorious for injecting large amounts of EMI which can cause false readings, and all the
same RF problems exist (connector and detector aging or damage).
It is possible to combine techniques, and perform field linearity calibration on a digitizing
sensor. Most allow (or require) at least a zero adjustment and the operating software often
allows a fixed calibration using a known RF reference source. As of the publication date
of this guide, there are no digitizing sensors that permit full power sweep field calibration,
and all rely on stored factory linearity data.
Digitizer
RF Amplifier
DC Switch
SP3T
ADC
DUT
DC Reference
Signal
Calibration
Control Line
Power Meter
Base Unit
Figure 5.3.2. Block Diagram of DC Method
Chapter 5: Calibration Issues
58
5.4 Frequency Response Correction
The task of a power sensor is simple: convert RF to a measureable and known DC level
across a broad range of carrier frequencies. However, detectors are not perfect, and there
are always minor variations in the sensor’s output as frequency changes. Most sensors are
fairly flat at lower frequencies, and begin to experience increasing response deviations as
the input frequency goes up. Matching at the input RF connector and within the sensor’s
detector assembly begin to play a role, and the output of the detector itself eventually falls
off.
The good news is that the variation is generally small (typically no more than a few dB),
and if the frequency is known, it is possible to compensate for the frequency response
deviations. To accomplish this task, power sensors are factory calibrated at a series of frequency points to generate a table of correction values. These values are most commonly
referred to as “Effective Efficiency” values (in percent), or “Calibration Factors” (in dB), and
are supplied to the user.
Setting frequency automatically computes the correct calfactor from sensor’s table
In modern sensors, the table is stored in an EEPROM within the sensor, and the power meter
can automatically load and apply the appropriate factor based on a user set frequency. If
the operating frequency falls between table entries, an interpolation or curve fit is used.
In some sensors, the frequency correction data is part of a multi-dimensional table that
simultaneously performs linearity, frequency and temperature compensation.
There are two common methods of generating basic Calibration Factors. The first is a direct comparison with a “gold standard” power sensor that has a NIST traceable frequency
calibration. In this technique, a leveled signal source is first swept or stepped through each
desired calibration frequency, and the leveled power reading from the NIST traceable power
sensor is recorded at each calibration frequency. That sensor is then replaced by the sensor
to be calibrated, and the sweep is repeated. The ratio at each frequency point between the
resulting reading and the corresponding reading from the reference sweep is the calibration factor. (see Figure 5.4.1)
59
High Performance
Signal Generator
Precision
Leveled
Signal
Calibration
Standard
Power Meter
Base Unit
Power
Sensor
DUT
Power
Sensor
Figure 5.4.1. Diagram showing leveled source feeding first “standard” sensor, then DUT
The second method uses a terminated average thermistor sensor (reference), precision
power splitter, and a signal source. The generator supplies a CW signal at a specific frequency to the precision power splitter and the reference power meter value is used to correct for
any level variations from the signal source or splitter. (see Figure 5.4.2)
IEEE-488 Bus
Reference
Power Meter
High Performance
Signal Generator
Power
Meter
DUT
Precision Power Splitter
Power Sensor
Figure 5.4.2. Diagram of Power-Splitter based sensor frequency calibration setup
Chapter 5: Calibration Issues
60
The center point of the splitter is a constant voltage point and the SWR at the splitter
output port is dominated by that side arm of the splitter. Low splitter SWR minimizes
mismatch error in the transfer of calibrations. In addition, the stability of the thermistor
mount means that the combination can be accurately calibrated and will perform well with
wide variations in generator characteristics. It also supplies transfer calibration factor accuracies of 1.2% to 2.5% (RSS) across the frequency range of the system.
The mistake that many customers make when attempting to perform their own transfer
calibration comes from their belief that the output level of a signal generator is accurate
and its VSWR is very good. The leveling accuracy of most generators is ±1 dB and VSWR
rarely better than 1.35:1 over a broad frequency range. RF sweepers are even worse, often having leveling inaccuracies of up to ±3 dB. That is why Calibration Factors cannot
be checked by using just the output of a generator. They must be checked by one of the
methods described above. Assembling and maintaining a traceable, automated sensor
calibration station based on the power-splitter method is now within the reach of many
calibration labs thanks to systems manufactured by Tegam.
61
Chapter 6: RF Power Analysis
RF power measurement can be divided into several broad categories:
•
Continuous Power Measurement. The measurement process is continuous or freerunning, there are no pauses in acquisition and results are delivered as they are measured. In almost all cases, the primary measurement is simply the average power.
Other parameters such as peak power may also be measured.
•
Triggered Pulse or Burst Acquisition. Typically the measurement process is discontinuous, and synchronized with the signal in some way. Once initiated, acquisition runs
for a period of time then stops, and the result is delivered. The measurement result is
generally a power-versus-time representation of the signal, and may include advanced
time and power analysis.
•
Statistical Power Analysis. A relatively new method of analyzing power in which a
large population of power samples is acquired and analyzed to determine the frequency of occurrence of a particular power level or range. The samples may be acquired
continuously (free-run), or synchronized with the signal in some way (triggered or
gated).
Following is a detailed discussion of each of these power measurement modes. All power
meters support at least one of these methods, and more advanced peak power meters support them all. Which mode is most appropriate will depend primarily on the signal being
measured, but is also affected by what parameters of that signal are important to know.
Typical Modulated Mode displays
Chapter 6: RF Power Analysis
62
6.1 Continuous Power Measurement
This is the most common power measurement method, and its operation can be likened to
a digital multimeter. In the most basic continuous mode, the sensor is connected to a CW
or modulated signal, and the power meter will simply display the average power of that
signal. The display can either be numeric, with selectable resolution or graphical – most
often displayed as a bargraph or meter scale. Units are most often dBm or watts.
CW and average diode sensors, thermal sensors and other types of low-bandwidth trueaverage power sensors all operate in continuous mode, often called “CW Mode” for these
sensor types, since operation is tailored towards CW signals. Modulated signals may also
be accommodated, but care must be taken to stay within the square-law region if a CW
diode sensor is used. Modern instruments perform continuous mode measurements with a
low-noise, low-bandwidth, high-resolution analog channel to process and digitize the sensor output, and all the shaping and corrections are performed by a microprocessor to yield
an accurate power value.
Peak power sensors may also be used in continuous mode, sometimes referred to as “Modulated Mode” since the peak sensor allows more flexibility and capability with modulated
signals. The circuitry is somewhat different – a high-bandwidth amplifier and high-speed
digitizer follow the detector, and the digitized samples are linearized and then averaged
together to yield the average power. Additional parameters such as peak and minimum
power can be made available with suitable processing. But the basic operating mode is the
same – the power is continuously measured and displayed.
Continuous mode instruments generally measure power at a relatively slow rate – usually
from about 10 Hz to a few kHz, and then apply a time integration filter to reduce noise
further at low signal levels. This filter can also be useful for reducing the displayed power
fluctuations caused by modulation. For periodic signals, it is beneficial to set the averaging time to be equal to an integer number of modulation cycles. Chapter 8 of this guide
discusses the effects and benefits of signal filtering.
In addition to a numeric power display, continuous mode measurements may also be presented in a graphical format showing power versus time, subject to the limitations of the
measurement rate. Most useful is a “strip chart” type display, in which the power is shown
as a scrolling line. In some instruments, this display can span from seconds to hours, which
can be very useful for observing a drifting signal, especially when used in conjunction with
appropriate signal filtering.
Another common feature of continuous mode measurements is ratiometric measurement.
Most power meters allow the user to store a “reference level” from which a power ratio is
computed and displayed. Most often the display is in dBr, with 0.00 dBr representing a
63
power equal to the reference level. For dual-channel power meters, it is usually possible to
display the ratio between the two channels as well. This can be useful for computing gain
or attenuation. When used with a directional coupler, the ratio between the channels is
equal to the return loss of the signal passing through the coupler.
The primary display may be the actual sensor signal amplitude, offset amplitude, or a
mathematical function (ratio, sum or difference) of the sensor signal and the signal from a
second sensor or stored reference.
6.2 Triggered and Pulse Analysis
For periodic or pulsed signals, it is often necessary to analyze the power for a portion of the
waveform, or a certain region of a pulse or pulse burst. The following is a brief review of
power measurement fundamentals:
Unmodulated Carrier Power. The average power of an unmodulated carrier consisting of
a continuous, constant amplitude sinewave signal is also termed CW power. For a known
value of load impedance ZLOAD, and applied voltage VRMS, the average power is:
P = VRMS² / ZLOAD watts
Power meters designed to measure CW power can use thermoelectric detectors which respond to the heating effect of the signal or diode detectors which respond to the RMS voltage of the signal. With careful calibration, accurate measurements can be obtained over a
wide range of input power levels.
Modulated Carrier Power. The average power of a modulated carrier which has varying
amplitude can be measured accurately by an average or CW type power meter with a thermoelectric detector, but the lack of sensitivity will limit the range. Diode detectors can be
used at low power levels that are within the square-law response region (no peaks higher
than about -20 dBm). At higher power levels the diode responds proportionally to voltage
rather than power, and significant error in the average power reading will result.
Pulse Power. Pulse power refers to power measured during the on time of pulsed RF signals (see Figure 6.2.1). Traditionally, these signals have been measured in two steps: (1)
average-responding sensors (thermoelectric or square-law diode types) measure the average signal power, (2) the average reading is then divided by the duty cycle to obtain pulse
power, PPULSE :
PPULSE = PAVERAGE / Duty Cycle
where Duty Cycle = Pulse Width / Pulse Period
Chapter 6: RF Power Analysis
64
Pulse power computed in this way provides useful results when applied to ideal, periodic,
rectangular pulse waveforms, but is inaccurate for pulse shapes that include distortions,
such as overshoot or droop, or if pulse period and width are not perfectly uniform. (see
Figure 6.2.1)
v(t)
(a)
t
v(t)
(b)
t
Figure 6.2.1. Ideal (a) and distorted (b) pulsed RF signals
Peak Power. Peak power meters perform power measurements in a manner which overcomes the limitations of the pulse power method and provides both peak power and average power readings for all types of modulated carriers. Fast responding diode detectors
track the RF envelope to produce a wideband video signal which is sampled at a high bandwidth and data rate by the peak power meter. The sampled detector points are accurately
converted to instantaneous power in watts on an individual basis using stored calibration
information.
Once the samples have been converted to linear power, the mean value of all or a subset
may be computed to yield the true average power without restriction to the diode detector’s square-law region. A time-domain reconstruction of the signal’s envelope can be created by assembling the samples in sequence into a display buffer. For repetitive signals,
this process can make use of equivalent time or interleaved sampling techniques to yield
time resolutions greatly exceeding the sample rate when synchronized by an internal or ex65
ternal trigger signal. Repetitive signals also permit synchronous filtering (trace, or “video”
filtering) of the resulting waveform – discussed in more detail below.
Peak power meters often refer to this measurement mode as “Triggered” or “Pulse Mode”.
Operation is similar to a modern digital storage oscilloscope – power samples are stored in
a circular memory buffer until a trigger signal is received. The samples, with the desired relationship to the trigger signal, are then selected and processed to obtain a power-versustime trace.
The trigger signal can either be a separately applied “external” pulse, or can be generated
by the RF signal’s amplitude crossing a defined threshold in either direction. The trigger
source, level and polarity are programmable, and oscilloscope-like settings such as trigger
delay time and trigger holdoff are often available.
Early peak power meters did not have memory – they used a variable delay generator, and
sampled the input signal at some defined period after the trigger to yield “power at a time
offset”. To reconstruct the waveform, the delay time was incremented in small steps in
each succeeding trigger, and the resulting array of power points assembled in order on a
display. This technique permitted wide measurement bandwidth using the slow A/D converters of that time period.
Modern peak power meters acquire the signal at very high conversion rates – typically
many MHz – and large acquisition buffers permit display of both pre- and post-trigger portions of the waveform. For each triggered sweep, data acquisition into the circular buffer
is restarted, and it runs at high speed until a trigger edge has been detected and all posttrigger samples acquired. At this point, acquisition stops, using the buffer is processed and
displayed before another trace is restarted.
Rising edge of a Peak Power Envelope
Chapter 6: RF Power Analysis
66
Figure 6.2.2. Screen shot of under-sampled rising edge on left and well-defined, over-sampled edge on right
Triggering. Like a DSO, there are a number of trigger options. Auto-trigger modes will
force a trigger when no edges are detected, but will synchronize with the signal once the
edges appear. A “peak-to-peak” trigger mode can be chosen to automatically set the trigger level based on the input signal. The most advanced peak power meters have complex
trigger generators which can do trigger arming and qualification on counted signal events
or time delays. Some also have programmable “fence” and exclusion intervals to assist
triggering on burst type signals.
Display timebases range down to the few ns/div range, limited generally by the video
bandwidth and time resolution of the instrument. Time resolutions of 200ps or better are
available on the fastest peak power meters, and are critical to accurate waveform reconstruction and pulse measurement. (see Figure 6.2.2)
Programmable time markers (cursors) are often available, and may be positioned on any
portion of the displayed trace to mark regions of interest for detailed power analysis. Cursor measurements generally include the power at each marker, as well as a series of parameters for the interval defined between the two markers – usually at least the average and
peak power. This is very useful for examining the power during a RADAR pulse or digital
communication burst when only the central region of the pulse is of interest. By adjusting
trigger delay and other parameters, it is possible to measure the power of specific timeslots
of TDMA signals such as GSM and EDGE.
Trigger holdoff allows burst synchronization even if there is more than one edge in the
burst which may satisfy the trigger level. Simply set the holdoff time to slightly shorter
than the burst’s repetition interval to guarantee that triggering occurs at the same point
in the burst each sweep.
67
Figure 6.2.3. Screen shots showing cursor measurements and statistical distributions
For periodic waveforms, automatic measurement of waveform parameters is available in
pulse mode. Once a stable, periodic signal is detected, the instrument automatically locates the waveform transitions, and calculates a number of pulse parameters such as pulse
frequency, width, duty-cycle, rise and fall times, top and bottom powers, pulse on power,
overshoot, and average power a full cycle.
Trace averaging is available, and generally required for performing low-level measurements
due to the wideband noise of peak power sensors. Unlike Continuous Mode, in which the
measurement bandwidth is simply reduced by averaging more samples, Pulse Mode uses
synchronous averaging, or “Video Averaging”. This averages each acquisition sample with
other samples at exactly the same time offset relative to the trigger, effectively averaging
each trace with the previous traces while maintaining time alignment. Video averaging
works best for periodic waveforms. A detailed description of this process and its benefits
is discussed in Chapter 8 of this guide.
Pulse mode of advanced peak power meters often include the same sorts of features found
in advanced digital oscilloscopes including deep memory, waveform zoom, trace storage
and recall, and mathematical functions between traces.
6.3 Statistical Power Analysis
For pulsed and periodic waveforms, the signal’s power envelope can be reconstructed
and analyzed in the time domain to provide a considerable amount of useful information.
However for continuously-modulated signals or periodic signals with noise-like modulation
within bursts or packets, it becomes difficult or impossible to trigger from the signal itself,
or to extract useful information in the time domain. Simple, continuous-mode processing
can yield the average and sometimes peak power, but often there is more information to
be gained by alternate acquisition and processing methods.
Chapter 6: RF Power Analysis
68
Many modern communication signals fall into this category due to their noise-like, digitally modulated data formats; CDMA, OFDM and various forms of QAM are some examples.
For these signals, statistical power analysis often makes more sense than time-domain
analysis. When statistical analysis is performed, power samples are acquired and analyzed
by how frequently each power value occurs rather than precisely when it occurs. A large
sample population is acquired either asynchronously or synchronously, and then sorted by
power into “bins” to yield a histogram. The more samples in the population, the finer the
histogram resolution.
Statistical Power Analysis is best for signals with the following characteristics:
•
Moderate signal level above about -40 dBm.
•
Digitally modulated signals, especially “noise like” formats such as CDMA (and its
extensions) or OFDM when probability information is helpful in analyzing the signal.
•
Any signal with random, infrequent peaks, when you need to know peak probability.
Statistical Presentations. There are several common presentations for viewing statistical
power measurements: Histogram, PDF, CDF and CCDF are a few viewing options for each.
•
A power histogram is the simplest – the samples are sorted into equal-width bins of
any convenient size. The bin divisions are most often logarithmic in power, with each
bin ranging an equal number of dBm. Bin depths (maximum count values) can be up
to or exceeding 32 bits (4 billion counts).
•
The PDF, or Probability Density Function is essentially a continuous function that is
similar to an infinite-resolution (zero bin width) histogram. The PDF cannot directly return absolute signal measurements, but its shape yields a qualitative indication of the
power distribution. A multi-level signal such as QAM will show up as a distinct “hump”
at each power level, and may be useful for visualizing system linearity.
•
The CDF, or Cumulative Distribution Function is the integral of the PDF. Its value
is monotonic and increases from 0.0 to 1.0, and represents the probability that the
power is at or below that point. Textbook representations will have the independent
variable (in this case power) on the X axis and probability on the Y axis. A CDF value
of 0.0 will be found at the minimum power point, and a CDF value of 1.0 will be at the
maximum (absolute peak) power.
69
•
The CCDF, or Complementary Cumulative Distribution Function (sometimes shown
as “1-CDF”) is simply the arithmetic inverse of the CDF, representing the probability
that the power is at or above that point. A CCDF value of 0.0 will be found at the maximum power point, and a CCDF value of 1.0 will be at the minimum power. The CCDF
presentation is more often used for power analysis because peak power is generally of
more interest than the minimum power, and it is more convenient to expand about the
origin.
Amplitude
15.8 dBm
12.2 dBm
Amplitude
PDF
3 dBm
CCDF
CDF
0%
Increasing Occurrence
100%
0%
Decreasing Occurrence
-3.6 dBm
100%
Figure 7.2.3. Common statistical diagrams including a PDF, CDF and CCDF
Frequent
Events
Increasing
Probability
Infrequent
Events
Increasing Crest Factor
Normalized CCDF plot showing probability on the Y axis versus normalized power on the X axis (dual channel overlay)
Chapter 6: RF Power Analysis
70
4500B gated histogram with power on the Y axis. Pulse waveform in upper windows showing time gate cursors,
and gated PDF (high resolution histogram) in lower window. For this view, it is more meaningful to present power
on the Y axis using the same scaling in both windows.
Figure 6.3.1. Tabular CCDF view
The CCDF is most often presented graphically using a log-log format, with log power (dBm)
on the X axis and log probability on the Y axis. For analyzing communication signals, it
may be helpful to normalize the power to the average power, so the X axis represents the
number of dB above or below the average power level (now scaled in relative dB, or dBr).
This is useful because it is sometimes only the shape of the CCDF that is important, and not
its absolute power value. In this case, a power value of 0 dBr is the average power, and a
CCDF value of 0.0 represents the peak-to-average power ratio.
Some power meters invert the X and Y axes to maintain the power measurement convention of displaying amplitude on the Y axis. This can be particularly useful when presenting a
histogram or PDF along with a corresponding time-domain trace. Both orientations present
the same information.
71
Tabular, or single-point CCDF values are also common, and often power (whether relative
or absolute) is treated as the dependent variable (see Figure 6.3.1). For example, a “0.01%
CCDF power level” indicates the peak-to-average threshold in which only 0.01% of the
power samples (a probability of 1e-4) fall above. A typical CDMA signal might have a
0.01% CCDF value of 8.7 dBr, indicating that only one of every 10,000 power points will be
more than 8.7 dB above the average power.
These measurements are useful because the expected shape of the CCDF can be easily
computed based on the data patterns and modulation format, or a CCDF for an undistorted
“gold” signal can be acquired, analyzed and stored. The CCDF of the signal may then be
measured at various points in the signal chain – for example immediately after a power
amplifier. When the curves are overlaid, signal distortion will be immediately evident by a
difference in the shapes. For example, mild peak compression is apparent in Figure 6.3.2,
which shows the output CCDF (yellow trace) falling off at a steeper rate than the input CCDF
(blue trace) as the peak-to-average ratio increases.
Figure 6.3.2. Dual CCDF: amplifier input on Channel 2 (blue, right), and amplifier output on Channel 1 (yellow, left)
The cursors in this example are located at 0.001% (or 1e-5) probability, and show an input
CCDF ratio of 10.862 dB and output CCDF ratio of 9.540 dB. The two curves are normalized to the average power of each, so the difference in these two CCDF values (1.332 dB)
indicates the degree of compression of the peaks that occur only once every 100K samples.
More frequent peaks will experience less compression (CCDF curves are closer together at
higher %CCDF points), and the rarer and higher-amplitude peaks will experience even more
compression. Absolute limiting (clipping) would be shown by the CCDF becoming vertical.
The use of statistical power analysis for amplifier testing is discussed in more detail in Section 8.3.
Chapter 6: RF Power Analysis
72
Figure 6.3.3. 4500B screen shot of Gated CCDF showing the pulse waveform of a WLAN frame, and time gates in
the upper window, and the signal’s CCDF curve in the lower window. The time gates in this example have been positioned to include only the payload portion of a data frame, and exclude the interval that the pulse is off, as well
as its initial training sequence.
Gated statistics. Many modern signals are transmitted in bursts or timeslots, but also use
noise-like digital modulation formats that will benefit from statistical analysis. For these
protocols, it is beneficial to acquire the statistical population synchronously – that is triggered or gated externally or by the signal so that the acquired samples only represent particular intervals in time. GSM-EDGE is one example – it is most useful to acquire data only
during the payload portion of the timeslot. WiFi and other signals have similar requirements – often a burst begins with some sort of training sequence that will skew the distribution in undesirable ways if samples from that interval are included in the population.
(see Figure 6.3.3)
High-end peak power analyzers offer the ability to operate in two modes simultaneously.
The instrument is set up for a basic, pulse mode acquisition so that a triggered waveform
is visible. All triggering features of the power meter may be used. Cursors are then positioned on the waveform to indicate the interval over which to perform statistical analysis. Only power samples occurring within this interval are included in the population. All
samples outside the interval, such as the burst “off” time, rising and falling edges, and
data references such as pilots and training sequences are discarded, and will not skew the
distribution.
Statistical Acquisition Size. One weakness of statistical power analysis is that the power
bins used to acquire data are not of infinite depth. The faster the data points are acquired,
the sooner the bins will fill up, and eventually the maximum count limit is reached and acquisition can no longer proceed. At this point, a decision must be made on how to handle
the situation. (see Figure 6.3.4)
73
The simplest option is to consider the acquisition “complete” and stop adding new samples
to the population. Depending whether statistical measurements will be immediately required, it may be more appropriate to clear the population and restart acquisition of a new
population. The down side of this “flush and restart” technique is that there is a short period at the start of each acquisition when the population is small and the statistical resolution will be extremely coarse. If a CCDF measurement at a very small probability is needed,
there may not be enough samples for a statistically valid return value.
To remedy this, a portion of the distribution may be discarded, or “decimated”. All power
bins are scaled by an equal value, for example 0.5. This is computationally simple, and fast
to perform – the binary count values in each bin are right shifted by one bit. Once all the
count bins have been halved, acquisition can continue. The elegance of this technique is
that the shape of the statistical distribution is not affected except at the very rarest power
values, where binary roundoff error or truncation becomes significant.
The effect can be considered a CCDF “filter,” in which the distribution is most heavily
weighted by recent events, and older events eventually are decimated off the bottom. The
time it takes this to happen is proportional to the count at which decimation occurs, and
inversely proportional to the sample rate.
Many signals can yield meaningful CCDF results with several to a few tens of megasamples,
so it is possible to make use of this decimation process for less than a “full” bin. Peak power meters may include the option to terminate, restart or decimate statistical acquisition
after a user-defined population size (“Terminal Count”) or time interval (“Terminal Time”)
has been reached. (see Figure 6.3.4)
Figure 6.3.4. Terminal count options
Chapter 6: RF Power Analysis
74
Chapter 7: Power Measurement Applications
This chapter discusses several common power measurement applications:
•
Low Duty Cycle Measurements – a discussion of why the conventional AveragePower
/ DutyCycle method for computing pulse power is inaccurate for narrow duty cycle
waveforms, and the advantages of using peak power measurement techniques. (see
Figure 7.1.4)
•
Measuring Modern Communication Signals – considers the special needs when measuring digitally-modulated signals of modern wireless networks. The modulation formats used often require special techniques to yield accurate and meaningful results.
•
Using Power Meters for EMC Testing – EMC testing is receiving a great deal of attention. RF emissions are, and have always been important, but increasing emphasis is
now being placed on equipment’s immunity from various types of interference. As the
signals used for EMC testing have become more complex, traditional methods of instrumenting the test environment and test results have given way to new techniques
requiring the use of peak power measurements.
7.1 Low Duty-Cycle Pulse Measurements
Why the average power measurement is not adequate! A prime concern for specialized
tube amplifiers like TWTA’s, magnetrons, and klystrons is the amplitude, quality and stability of the device’s output power. These devices are designed for high power RADAR,
particle accelerators, and Magnetic Resonance Imaging devices (MRI). They must provide
consistent pulsed linear power to either a large antenna with low return loss, or to large
powerful magnets that create similar power transfer issues. These amplifiers require accurate peak power measurement for safety and optimum performance. A common characteristic of these applications is the use of a low duty cycle pulse for measuring a small size
at a far distance for radar applications, or accurately controlling atom size particle position
for physics research.
Before peak power meters were available, the pulse power was computed indirectly from
an average power measurement performed with an average-responding power meter. The
pulse power was calculated by dividing the average power (PAVG) by the duty cycle. The
duty cycle is often a known characteristic and is calculated by dividing the pulse width of
the power envelope by the pulse repetition interval. (see Figure A 7.1.1)
75
This computation assumes constant peak power and does not take into consideration factors such as overshoot and ringing. This is why it is called pulse power and not peak pulse
power (see Figure 7.1.2). The pulse must be repetitive, rectangular, and of constant duty
cycle to make an accurate calculation. For more information on this topic see Section 4.5.
Pulse
Repetition
Interval
(PRI)
Power envelope
Measured
Average
Power
(Pavg)
Pulse
Width
(PW)
Duty Cycle = PW / PRI
Pulsed Power = Pavg / Duty Cycle
Figure 7.1.1. Pulse Power Computation
Overshoot and
Undershoot
Droop
Rise Time
Droop
Pulse
Power
Figure 7.1.2. Common pulse waveform distortions that can affect the accuracy of pulse power computations
Power Envelope
RF Carrier
Figure 7.1.3. Envelope of a wide duty cycle pulse waveform is simple to measure with most average-responding
power meters
Chapter 7: Power Measurement Applications
76
One advantage of using an average power meter is the ability to measure over a wider
dynamic range than a peak power meter, but this assumes the signal envelope is a perfect
rectangle. This advantage disappears for narrow duty cycle signals. (see Figure 7.1.3)
In a RADAR or MRI system, the RF or microwave carrier is sent in short bursts over long
periods to provide a signal to measure over long range and small object size. Simple range
finding radars use pulse modulation, while some doppler radars use a continuous wave
tone. Pulse modulation requires the carrier to be switched on and off in synchronization
with an external pulse signal and is not modulated like a communication signal. The envelope of the pulse waveform is extracted from the demodulated carrier in the receiver (see
Figure 7.1.3).
A CW signal has the same average and peak power value and can be measured provided
the value is within the sensors dynamic range. A narrow duty-cycle pulsed signal can have
a significantly lower average value outside of the sensors dynamic range which the peak
value falls within that range.
To perform an accurate measurement of pulse power using the duty cycle method, it is
necessary to accurately measure the signal’s average power. This requires maintaining
the average power well above the sensor’s noise floor. At the same time, the power sensor
must be able to handle the highest power peaks while the pulse is on, or the sensor will
produce erroneous readings or burn out.
Most average power sensors can accommodate peaks that are 10 dB to 20 dB above their
maximum average power ratings, so pulse waveforms with relatively wide duty cycles can
be measured without challenge (see Figure 7.1.3). But for signals with duty cycles narrower than about 1%, the dynamic range of the average-responding power sensor is eroded
by the need to maintain both the peak power and the average power within the sensor’s
operating, and measureable range.
Duty Cycle = 0.001%
1us
100ms
Figure 7.1.4. Narrow Duty Cycle pulse envelope
77
Simply stated: As the duty cycle of a power envelope is decreased, typically below 1%, the
average power reading is further away from the actual peak power delivered and requires
larger dynamic range sensors. (see Figure 7.1.4)
The following example, calculation shows duty cycle and dynamic range computations for a
periodic waveform with a one microsecond pulse repeating at a 10 Hz rate.
Duty cycle calculation: (1.0e-6 / 0.1) = 0.00001, or 0.001%
Conversion to dB = 10 x Log10(0.00001) = -50 dB
Therefore, measuring the pulse power of this signal requires a power meter with at least 50
dB of dynamic range. For a typical thermal sensor with 22 dB peak headroom (+42 dBm),
this means the signal’s average power must remain at least 50 dB below its peak rating,
or -8 dBm. These types of sensors have a noise floor of about -25 dBm, so the signal can
vary by no more than 17 dB before its peak burns out the sensor or its average falls below
the noise floor. For accurate measurements, the signal should remain 6 to 10 dB above the
noise floor, further degrading the dynamic range.
A peak power sensor with its wide dynamic range is often a better solution for measuring
this type of signal for several reasons.
1. The pulse shape is not always rectangular and will contribute errors when calculated
using an average power sensor measurement in the pulse power calculation.
2. The dynamic range of an average power sensor is reduced in proportion to the duty
cycle because the noise is integrated into the measurement with a long PRI and a short
pulse width.
3. A fully calibrated good quality peak sensor has a dynamic range of 70 dB and capable
of measuring a 50 dB peak to average ratio without affecting the measurement.
Figure 7.1.5 shows a comparison of Duty Cycle vs. Peak Pulse Power for three common
power sensor types. The sensors used for this comparison are an average-responding thermocouple sensor, an average-responding diode sensor (operating in the diode’s square-law
region below -20 dBm), and a peak power diode sensor. The maximum and minimum
power values for each sensor represent their total dynamic range capability.
Chapter 7: Power Measurement Applications
78
Thermal, Average Diode, and Peak Power Comparison
50
40
30
Peak Pulse Power (dBm)
20
10
Thermal Detector Power Limit
0
Thermal Detector Noise Floor
-10
Uncalibrated Diode Detector
-20
Power Limit
-30
Uncalibrated Diode Detector
-40
Peak Sensor Noise Floor
Noise Floor
-50
Peak Sensor Power Limit
-60
-70
100
10
1
0.1
0.01
0.001
0.0001
0.00001
Duty Cycle (%)
Figure 7.1.5. Thermal, Average Diode, and Peak Power comparison
Note for the two average-responding sensors, that the usable dynamic range becomes
narrower as the duty cycle decreases. However, this is not the case for the peak sensor.
Although it can handle short peaks up to 10 dB above its +20 dBm average power rating,
its measureable pulse power rating does not rise with narrowing duty cycle due to the +20
dBm upper limit of its calibrated measurement range. But on the lower side, the noise floor
does not increase as the duty cycle narrows because the peak sensor is able to trigger and
measure only during the “on” portion of the pulse, and discard measurements at all other
points in time.
This means that the usable dynamic range of a peak power sensor remains constant with
duty cycle, unlike average responding sensors. The pulse waveform in our example above
can be measured while the signal varies over a 60 dB dynamic range, and still remain substantially above the sensor’s noise floor to produce stable and accurate measurements.
The shading of Figure 7.1.5 provides additional insight into how the available dynamic
range of average-responding sensors is reduced when measuring narrow duty pulses. This
is shown in Figure 7.1.6.
Note that both thermal and average diode sensors run out of dynamic range for duty cycles
narrower than about 0.003%. The operating area of the thermal sensor can be extended in
some applications by taking advantage of the fact that its peak handling capability is considerably higher than its average power rating. This is dangerous, however, as the source
79
may be able to generate considerably more power than the sensor can safely handle. The
user must depend upon the signal’s duty cycle remaining narrow enough to limit the average power to a safe value. If the signal’s duty cycle increases, the average power will
increase accordingly.
The peak power sensor does not suffer from duty-cycle limitations, and its operating dynamic range is wide no matter what the waveform’s duty cycle. The true peak power of the
signal is measured directly, and even single pulse events may be easily measured.
50
40
Thermal Sensor
“Extended” Area
30
Peak Pulse Power (dBm)
20
10
Peak Power Sensor
Dynamic Range
0
-10
-20
Avg Diode Sensor
Dynamic Range
-30
-40
-50
-60
-70
100
10
1
0.1
0.01
0.001
0.0001
0.00001
Duty Cycle (%)
Figure 7.1.6. Dynamic range of Thermal, Average Diode and Peak Power sensors versus duty cycle
7.2 Statistical Analysis of Modern Communication Signals
The latest wireless communication formats like DVB, DAB, WiMax, WLAN, and LTE cellular
use OFDM modulation schemes with multiple carriers to transmit digital information. OFDM
is a multi-carrier modulation scheme with a high crest factor to transmit large amounts of
data. The introduction of digital transmission technology has made it necessary to deal
with power peaks up to 20 dB above the average value. RF power components must be
suitably specified to handle the expected voltage peaks and avoid break down, or flash
over. The crest factor, i.e. the ratio of the peak value to the average or RMS value, must be
determined to correctly specify these components. The peak power of several interconnected transmitters can reach more than one hundred times the thermal or average power
level. The selection of the RF power components for the transmission system (antenna
combiners, coaxial lines and antennas) cannot be based solely on the thermal or average
power. Short voltage spikes that occur rarely are critical when determining the required
size and power handling capability of the RF components.
Chapter 7: Power Measurement Applications
80
Crest factor measurements over 12 dB (PPEP/PAVG) are difficult to make with repeatable results. To properly accommodate for these high crest factors a single peak measurement
is not adequate and statistical analysis should be used. Low amplitude communication
signals with high crest factor, although important when considering BER, are of greater
concern for their contribution to system damage. The high voltage associated with large
power peaks can produce flashover, or a standing arc in the transmitter system and destroy components. Statistics are an important tool for measuring these rare events when
concerned with property and personal damage. Because the instantaneous power values
are sorted by magnitude rather than their time of occurrence, they are counted and not
averaged. This process can run for a very long time, limited only by available memory, or
can run indefinitely if decimation is applied. It is invaluable for characterizing events, such
as the maximum peak power of an OFDM signal that might occur once a day. Capturing
pulse data using statistics provides additional insights not easily observed when capturing
amplitude vs. time measurements.
Figure 7.2.1 shows a real world CCDF (complementary cumulative distribution function) of
an HDTV, OFDM modulated signal. Trace A shows a 100% AM modulated sine wave, which
has “squared off” CCDF showing a peak-to-average ratio of 3 dB at both low and high probabilities. This is due to the highly predictable periodic waveform. The OFDM signal shown
in Trace B has a peak-to-average ratio of about 15dB, and follows a Rayleigh distribution
pattern. Trace C shows white, Gaussian noise for reference, which has a theoretically infinite peak-to-average ratio, here shown to be about 17 dB for at a probability level of 10-12.
10
A = CCDF of sine-wave
0
B = OFDM envelope C = white noise
-2
10
probability p
10-4
-6
10
10-8
A
10-10
B
C
-12
10
-14
10
0
2
4
6
8
10
12
14
16
18
20
k in dB above average
Figure 7.2.1. CCDF of a sinusoidal signal (carrier approach), an OFDM envelope signal and for white noise
81
For the AM modulated signal, it is clear that very few samples must be acquired to measure
the signal’s peak power, while the digital OFDM signal has a high peak-to-average ratio
with peaks that occur very infrequently. For these types of signals, the power meter must
acquire many samples over a period of time to accurately characterize the power distribution of the signal.
The following section will explain the statistical distributions and how they can be used on
a modern peak power meter. Figure 7.2.2 is an illustration of the PDF, or probability density function of a 16 QAM modulated digital signal. Levels, 1, 2, and 3 represent the three
distinct power levels of a 16 QAM signal.
16 QAM Constellation Diagra
16 QAM Power Histogram
Amplitude
CD
Level 3
0011
0111
11
+3V
1011
+2V
Level 2
AB
-3V
10
+1V
0110
01
0010
00
-2V
1010
10
-1V
+1V
-1V
0101
0001
01
1001
-2V
Level 1
0000
0100
00
W
CD
# of occurrences
6 QAM Power Histogram
16 QAM Constellation Diagram
Power Level Three
CD
0011
0111
+3V
11
1011
Power Level Two
1111
+2V
AB
+1V
0110
01
0010
00
-3V
0001
-2V
Power Level One
10
-1V
+1V
-1V
0101
1110
11
1010
10
01
1001
+2V
AB
+3V
1101
-2V
0000
0100
W
00
1000
1100
CD
Figure 7.2.2.
# of occurrences
Chapter 7: Power Measurement Applications
82
1000
+2V
For communication purposes it is often desirable to view the maximum power value in a
CCDF, or complementary cumulative distribution function. This is accomplished by integrating the PDF to create a CDF, or cumulative distribution function as illustrated in Figure
7.2.3. The final CCDF, complementary cumulative distribution function is calculated by subtracting the CDF from one. (1 - CDF = CCDF). This distribution has the highest peak values
(which have the lowest probability of occurrence) displayed at the upper left corner of the
graph.
Amplitude
15.8 dBm
12.2 dBm
Amplitude
PDF
3 dBm
CCDF
CDF
0%
Increasing Occurrence
100%
0%
Decreasing Occurrence
-3.6 dBm
100%
Figure 7.2.3. PDF, CDF, and CCDF diagram
The CCDF presentations illustrated in Figure 7.2.4 show probability on the X axis and absolute power in dBm on the Y axis. Note that the X axis is using linear scaling, so there is
rising “tail” displayed at the highest power levels which occur increasingly infrequently.
Figure 7.2.4. Two CCDFs of a WLAN signal showing the benefits of gating (right) versus free-running (left) acquisitions. The gated CCDF excludes the low-power “off” and low-crest-factor preamble section.
83
In the first example, the continuous, free-running acquisition process will gather samples
during both the active and inactive (off) signal intervals, distorting the CCDF. Note that the
power quickly falls off above about 62% probability, indicating the signal is spending more
than 1/3 of its time at low or “off” power levels. In a time-slotted or bursted signal such as
WLAN, this is expected, since the transmitter is turned off between signal packets.
To provide a more meaningful CCDF on time-synchronous signals, advanced peak power
meters such the Boonton 4500B included a time-gated statistical mode, which allows statistical acquisition only during selected portions of a waveform. This ability permits exclusion of off time, and preamble, and results in a CCDF which accurately reflects the power
distribution during the more random “payload” portion of the frame. The second example
in Figure 7.2.4 shows the same WLAN burst in the time domain in the upper trace window
with time-gate cursors being used to define the region of interest for CCDF analysis. The
CCDF in the lower trace window is computed only for the WLAN payload, and is no longer
skewed by low-crest-factor preamble and the off interval between bursts.
Note that the X axis in both examples uses linear scaling of probability, which results in a
rising “tail” as the CCDF curve approaches zero probability (the absolute peak power). This
tail is the region of the most interest to RF engineers, and using logarithmic scaling for the
CCDF straightens out the “tail” as it approaches the axis. This is often a more useful display
format, as it expands the very low probability events of interest. These are the regions
where signal compression can begin to affect the bit error rate of digitally modulated communication signals. The two screen shots below in Figure 7.2.5 use log scaling of the X axis.
Figure 7.2.5. Time gating the payload portion of the signal improves peak power statistical measurement accuracy by excluding the off interval between frames via a synchronous trigger and time gate cursors. The left-hand
display shows the CCDF for the preamble, which has a lower peak-to-average ratio (3dB), and has a relatively flat
CCDF. The right-hand display shows the CCDF for only the payload, which has a much higher peak-to-average ratio
(9.5dB), resulting in much greater stress on the signal chain.
Chapter 7: Power Measurement Applications
84
Limiting the time gates to the preamble area illustrates the low peak power values, and
constant power envelope. This is in contrast to the data section of the signal and provides
a qualitative view of inaccuracy. Statistics are important when measuring modern communication signals because of their high peak-to-average ratio or crest factor. The crest factor
is an important signal parameter and can be calculated and displayed in a CCDF. Power
domain statistics are ideal for noise-like signals like LTE and WLAN. Their non-periodic
characteristics are difficult to evaluate in the frequency domain using a spectrum analyzer,
or in the time domain using an oscilloscope.
The CCDF is often presented in a normalized, log-log display format, with power values
shown in dB relative to the signal’s average power. This is helpful for comparing the CCDF
at different points in the signal chain, as a particular signal should have a well-defined
CCDF regardless of its absolute power level. And while measurement instrumentation often displays signal amplitude on the Y axis, the textbook CCDF is typically shown with log
probability on the Y axis and normalized power on the X axis. This rotated, normalized
presentation is now becoming more common in power meters as well.
The next example in Figure 7.2.6 shows the CCDF of a WCDMA signal displayed on a Boonton 4542 peak power meter. The CCDF is displayed using the rotated and normalized presentation format – with log probability on the Y axis and crest factor (normalized peak
power) on the X axis. The left end of the Y axis is 0 dBr, which corresponds to the signal’s
long-term average power. The theoretical maximum peak occurs at 0% probability, which
is undefined on the log presentation. The trace intersects the X axis (0.0001%, or 10-6
probability) at a crest factor of about 15dBr. This means that only one sample out of every
one million would be expected to exceed the average power by more than 15dB. As the
probability is decreased further, it should be apparent that the crest factor will continue to
increase by a small amount, but from the slope of the CCDF near the bottom of the screen,
it appears that several more decades would add no more than 1dB.
The Boonton 4540 series peak power meters include a dual CCDF feature that allows comparison of the input and output power distributions of an RF device such as a power amplifier. This comparison in Figure 7.2.7 shows the crest factor deviation between the input
(Ch2, blue, on the right) and output (Ch1, yellow, on the left) of a signal amplifier. Because
the signal being amplified is an actual communication signal, it contains all the frequencies
and power levels of interest and operates the amplifier over its entire dynamic range. The
CCDF is more useful than a simple, crest-factor measurement as it quantifies the amount of
compression at various probability levels.
This allows designers to evaluate their amplifier performance using its intended signal type
rather than a CW tone to estimate the performance using a figure of merit like the 1dB
compression point. If the amplifier has been built into a receiver and a baseline BER value
85
for an operating system is known, the BER and CCDF can be correlated on the physical layer
before the receiver is assembled for production. Section 8.3 of this guide contains an indepth discussion of the use of dual-channel statistical power measurement for RF amplifier
testing.
Frequent
Events
Increasing
Probability
Infrequent
Events
Increasing Crest Factor
Figure 7.2.6. Rotated and Normalized CCDF display with Log-Log scaling
Frequent
Events
Increasing
Probability
Infrequent
Events
Increasing Crest Factor
Figure 7.2.7. Dual CCDF “input/output” display shows the output has a reduced crest factor, indicating signal
compression
Chapter 7: Power Measurement Applications
86
7.3 Using Power Meters for EMC Testing
The complexity of modern digital equipment has caused EMI/EMC susceptibility testing to
become increasingly important. Many EMC standards have been created including MILSTD-461, IEC 61000, ISO 11451 Automotive, EN 50, and FCC part 15 that provide specific
guidelines for EMC and EMI test methodologies. Early standards required a CW carrier, or
single tone with constant modulation as the disturbance test signal. In January of 2010,
the IEC committee approved the 61000-4-4-am1 (ed. 2) amendment allowing the use of
burst testing on devices. Amendment 1 defines an impulse (spike frequency) of 100kHz
and Edition 2 requires burst testing with either the traditional 5kHz spike or the new
100kHz spike frequency. The burst test emulates real world RF interference emitted by
base station communication amplifiers and ground based RADAR antennas. This section
will illustrate how a peak power sensor can replace a single diode detector in a field probe
to measure pulse power, improve repeatability and increase dynamic range of the power
measurement.
Historical Background. Prior to the late 19th century, the primary sources of electromagnetic disturbance were lightening strikes and sun spots, but the growing popularity of electrical and radio equipment in the early 20th century generated the first artificial forms of
interference from electrical powered equipment and competing radio transmitter towers
around the world. This competition led to the creation of international regulatory agencies,
like the FCC. This trend continued in the 1940s with the adoption of high power industrial
switch devices that caused coal mine explosions, automobile and airplane fuel station fires,
and electrical grid outages. During the 1950s and 60s, ISM (industrial, scientific, & medical) unlicensed frequency bands were allocated by the FCC which permitted the generation
of relatively high-power RF signals. Because emission in these bands was uncontrolled,
a variety of interference issues were created due to sideband harmonic and broadband
emissions. The impact of this interference resulted in the need to create new standards
and laws to regulate these emissions. With the advent of digital circuitry in the 1970s,
faster switching speeds had increased emissions and lower circuit voltage requirements
increased susceptibility. The 1980s brought an increasing use of mobile communications
and broadcast media channels creating pressure on the available spectrum space. Regulatory requirements for smaller band allocations demanded increasingly sophisticated EMC
design methods. Although digital signals are often less susceptible to interference than
analog systems, their operation at lower power levels gives up some of this immunity.
These issues have created the need for increasingly complex EMC/EMI testing.
Electromagnetic Compatibility (EMC) is a branch of electrical science that studies the unintentional generation, propagation and reception of electromagnetic energy from electromagnetic interference, EMI. An Emission is intentional or unwanted electromagnetic
energy produced by a source, which may couple into other devices. Susceptibility or immunity is the inability or ability of a piece of electronic equipment, referred to as the victim, to
87
Radiative
Source
Victim
Inductive
Capacitive
Conductive
Figure 7.3.1. Electromagnetic Coupling paths
operate correctly in the presence of nearby emissions or other electromagnetic interference
signals. Electromagnetic compatibility is achieved through addressing both emission and
susceptibility aspects of an electronic device.
The diagram in Figure 7.3.1 shows the four different types of electromagnetic coupling:
radiative, inductive, capacitive, and conductive. The primary type of coupling discussed in
this application note will be radiative, in which a signal radiates through space as an electromagnetic wave with no physical connection or coupling between the source and victim.
The purpose of immunity testing is to emulate the effect of real world RF interference upon
your electronic device or system. One example would be the automotive CANBUS system
used for wired, digital communication between electronic subsystems in a motor vehicle.
These systems are often used to monitor and control important performance and safety
parameters of the vehicle including engine operation, and acceleration, braking, and the
steering/stability systems, so their ability to operate correctly under all foreseeable conditions of electrical interference is crucial to passenger safety. Rigorous RF immunity has
become a mandatory part of the automotive design process as well as most other systems
where any sort of malfunction could result in injury or damage to people or property.
Immunity testing is performed in a large anechoic chamber for isolation from external RF
interference while testing the “EUT”, short for “Equipment Under Test”. One important
requirement for the tests is to apply a simulated interference signal with an accurately
known amplitude. RF field strength is typically measured and characterized during or prior
to the test using one of two techniques: the closed-loop method and the substitution
method. While each method has advantages, the IEC standard to which the test is being
performed will often determine the method that must be used to instrument the signal’s
amplitude.
Chapter 7: Power Measurement Applications
88
The Closed Loop Method requires an RF field probe positioned in front, or on top of the EUT
during susceptibility testing (see block diagram in Figure 7.3.2). The signal generator’s
output power is adjusted at each of the specified frequency steps across the test band to
achieve the desired RF field strength in the anechoic chamber. The word “probe” can have
two meanings. One is the field probe assembly inside the test chamber and the other is
a commonly used term for an average diode detector circuit. The average diode detector
is a component of the field probe assembly and measures RF power via a coaxial cable for
purposes of this discussion.
The average diode detector in the field probe does not accurately measure the field
strength of a modulated RF signal, so correction factors must be applied to the probe readings to account for the signal’s dynamic behavior. A CW signal can be used to estimate the
power being delivered, but an additional correction factor must be applied to account for
the modulation applied during the actual testing. This correction is adequate for simple
AM modulation, but is often insufficient for the narrow duty spikes required by today’s
test standards.
The simple diode detector can be replaced by a peak power sensor to accurately measure
the interference signal’s true amplitude even in the presence of modulation. A peak power
sensor can follow a signal’s power envelope and yield the true average and peak power,
provided the envelope bandwidth remains within the maximum video bandwidth rating of
the sensor and power meter. A good peak power sensor is calibrated for increased dynamic
range, and temperature compensated. Using a peak power sensor eliminates the need to
apply modulation corrections when a pulsed or modulated interfering signal is used rather
than a CW source. In cases where the modulating waveform is complex or a narrow duty
pulse, a peak power sensor becomes mandatory, as it is impossible to accurately correct
these types of waveforms for nonlinearity due to modulation when using a conventional
diode sensor or probe. These limitations are discussed in detail in Chapter 3 of this guide.
Anechoic Test Chamber
Radiating
Antenna
EUT
Field
Probe
Signal
Generator
Figure 7.3.2. The “Closed Loop” Method Diagram
RF Radiation Absorbing Material
Closed Loop Method
89
EUT
Test Set
Power
Detector
The Substitution Method uses an RF field probe to characterize and calibrate the RF field
strength in the anechoic chamber before the EUT is placed inside (see block diagram in Figure 7.3.3). The field strength is adjusted for each frequency step across the band and the
EUT is positioned in the test environment. This method does not require field monitoring
during the test and is referenced by some EMC test standards. While not required, a probe
is often used during the test run just to monitor the RF field. This direct feedback assures
Anechoic Test Chamber
good system performance.
Anechoic Test Chamber
Radiating
Antenna
Radiating
Antenna
Field
Probe
Field
Probe
Signal
Generator
RF Radiation Absorbing Material
Signal
Generator
RF Radiation Absorbing Material
Power
Detector
Power
Detector
Anechoic Test Chamber
Anechoic Test Chamber
Radiating
Antenna
EUT
EUT
Test Set
Radiating
Antenna
EUT
EUT
Test Set
Signal
Generator
RF Radiation Absorbing Material
Figure 7.3.3. The Substitution Method Diagram
Signal
RF Radiation Absorbing Material
Substitution Test Method
Generator
Substitution
Testuse
Method
Both the closed loop and substitution
methods
a high power signal generator connected to a radiating antenna for repeatable RF signal transmission while testing the EUT.
The closed loop method requires the field probe during testing, while the substitution
method only states it can be used to improve measurement quality. In either case, it can
be beneficial to use calibrated peak power sensor rather than the average diode detector
in most field strength probes. This eliminates multiple calibrations, modulation correction
factors, and the temperature compensation associated with the average diode detector,
and provides both peak and average information about the interfering field’s characteristics. Without knowing these values, it is impossible to be certain that the EUT is operating
in the intended interference environment.
Chapter 7: Power Measurement Applications
90
Chapter 8: Performance Tips
A straightforward power measurement under “good” conditions is generally not a challenge – simply hook up a power meter and read the results. However many applicationrelated issues can make the process more challenging. This chapter discusses several topics
on improving the performance of your power measurement:
•
Reducing measurement noise – a discussion of power measurement noise, and common reduction techniques of simple filtering and synchronous averaging.
•
Optimizing ATE Performance – a review of computer controlled testing, and how the
entire system may be automated for increased throughput. Describes the common
“sequential” ATE approach in which the controller steps through a sequence of steps
many times to perform a multi-point measurement. Following this is a discussion of a
modern alternative using preprogrammed source sweep control and buffered acquisition in the power meter to significantly reduce test duration.
•
Amplifier Testing Statistical Techniques – explores a new approach for RF amplifier
linearity testing using statistical analysis of a modulated power sweep rather than the
traditional method of applying a series of CW tones to measure IP1 and IM3. The new
method promises both increased test speed and enhanced compression characterization for estimating in-system performance of an amplifier.
8.1 Reducing Measurement Noise
It is often necessary to perform power measurements across a wide dynamic range, and
sensors may become challenged by signal power levels at both extremes. As discussed in
Chapter 3, the top of the range usually has hard limits – both the average and peak power
must remain within a safe (for the sensor) and calibrated range to be measured accurately.
The lower end of the range, however, is more difficult to define.
All measurements must be made in the presence of noise – it is a part of every electrical
measurement, including RF power. Unless the amplitude of the signal to be measured is
considerably greater than the noise amplitude, that noise will add to the measured signal,
and can result in considerable measurement uncertainty, as discussed in Chapter 9.
One convenient aspect of noise is that it is random. The finer details of noise are well understood by Boonton Electronics’ sister brand, Noisecom, who specializes in components
and instruments designed to generate or analyze various types of electrical noise. Because
of this randomness, and its Gaussian distribution, it is possible to reduce the effect of noise
while maintaining the characteristics of the signal one wishes to measure.
91
As generated, power measurement noise is fairly wideband, and has a roughly “white”
characteristic, in which the noise has constant power per Hertz. If the bandwidth is halved,
so is the average noise power. When the noise bandwidth is greater than the bandwidth
of the signal to be measured, it is possible to filter the noise bandwidth reducing the total
amount of noise.
tio
POWERdBm
N
dB
+10
S/
Ra
er
ow
eP
s
i
No
20
WLAN Signal
(20 MHz)
0dbm
-10
GSM Signal
(200 KHz)
-20
-30
CW Signal
(1 Hz)
-40
-50
200
1Hz
1KHz
10KHz
20
100kHz
1MHz
10MHz
100MHz
ƒHz
Graph of average noise power versus noise bandwidth, showing how BW may be reduced to improve S/N
CW signals represent an extreme case. The bandwidth of a CW signal is effectively 0 Hz,
so it is possible to filter the signal and noise together to sub-1 Hz bandwidth without affecting the accuracy of the measurement. Most power meters provide an averaging filter
that averages a number of readings over a defined time interval to yield a filtered result.
Increasing the averaging time setting will reduce the measurement noise at the expense of
settling time to signal level changes.
Unfortunately, there are other types of noise present besides Gaussian noise, which cause
increased filtering to reach a point of diminishing returns. A one-second filter time is generally appropriate for -60 dBm with a CW diode sensor, but increasing the filter to ten
seconds will not reduce noise by a factor of ten and permit the same accuracy at -70 dBm.
Most CW and average power meters include an “auto-filter” setting, in which the instrument chooses the filter time constant based on the measured power level. This can cause
extended settling times when the power level is changed, so a good recommendation is
that if you know your expected power level, set the filter manually for best performance.
Chapter 8: Performance Tips
92
The averaging filter of most CW and average power meters, as well as peak power meters
operating in continuous, free-run, or “modulated” mode is most often a sliding sample,
or “window” filter, in which power readings are performed at a relatively fast rate, and
then averaged together as needed. The filtered power reading at any given time is simply
the average of the last “n” samples, where “n” is the filter time divided by the internal
acquisition rate. Unweighted averages are the most common results, but other types of
weightings are sometimes used to help accommodate fluctuating signals such as pulse
waveforms. By continuously adding new samples and discarding old ones, the filter output
can be recomputed and updated at the acquisition rate.
When measuring a wideband signal with a peak power meter, filtering out the noise can
become more difficult. In peak power meters, it is often necessary to maintain measurement bandwidths of tens of MHz or more. Clearly, it is not possible to apply filters with millisecond or second time constants without also filtering out desirable signal characteristics.
Take advantage of the redundancy of repetitive signals. Any periodic waveform can be
synchronously filtered by averaging together multiple periods of the waveform. As long
as the individual periods or waveform events can be made to align closely in time, it is possible to average together even hundreds or thousands of individual cycles or sweeps, and
produce a filtered waveform with significantly less visible noise and no degradation of the
waveform’s amplitude and profile.
Effect of trace averaging for reducing noise on periodic pulse waveform. These screen shots both show one lowlevel (-39dBm) pulse of a periodic waveform. The left trace has no averaging, and its min-to-max ratio is about 2.4
dB, representing the effective peak-to-peak measurement noise. Increasing averaging to 1024 in the right trace
reduces peak-to-peak noise during the pulse to less than 0.2dB, as well as bringing the noise floor down from about
-53dBm to below -70dBm.
93
Synchronous waveform averaging is often called “trace averaging” or simply “averaging”
in peak power meters operating in triggered or “pulse” mode. Sometimes the term “video
averaging” is used, since it is familiar to spectrum analyzer users, but this can be confusing
in power meters.
The source of this confusion is that some peak power meters offer an additional method for
reducing noise – they are able to reduce the video bandwidth of the sensor or power meter
input circuitry. This technique will not introduce errors as long as the video bandwidth of
the measurement remains above the video bandwidth of the signal. In most cases, this
bandwidth reduction offers only a modest reduction in noise, but it may make the difference between a signal that can and can’t be measured, and it can be used in conjunction
with filtering or averaging.
One other technique that can improve noise in peak power meters is simply to sample
faster. If the power meter is already sampling at or above Nyquist rate, this method will
not yield significant improvement. But if it is undersampling and has excess bandwidth, a
faster sample rate will tend to filter the noise without impacting the signal.
8.2 Optimizing ATE Performance
Like most of the instruments available today, modern Peak Power Meters usually include
high-speed connectivity such as GPIB, LAN and/or USB interfaces to allow their operation
to be controlled remotely by a PC or system controller, and therefore to be integrated into
automated test equipment (ATE) systems. This functionality extends the instruments’
capabilities far beyond manual front panel operation since other devices can be combined
with the power meter to provide a centralized location not only to control them but to collect and display data from multiple instruments as well. The intent here is to provide an
overview of the additional functionality that can be obtained from an automated measurement application of power meters and an overview of some of the PC tools that facilitate
instrument programming.
The ‘glue’ that ties instruments to a PC consists of a PC I/O interface such as the VISA I/O
library. VISA is short for “VXIplug&play Systems Alliance”, and that group is now part of
the IVI Foundation. With VISA, the physical interface used on the PC can be its built-in serial, Ethernet LAN or USB interface, a PCI-based GPIB interface card, a USB-to-GPIB interface
cable or a combination of these I/O interfaces. The VISA library includes all the low-level
drivers and functions required to use any of these interfaces for instrument control. This
allows a portable, and bus-independent programming model which frees the developer
from needing to implement a low-level communication channel for each instrument and
bus in the system.
Chapter 8: Performance Tips
94
The last item the developer must select is a development environment to create the automated software. The .NET IDE (integrated development environment) allows the choice of
Visual Basic, C Sharp, or C++ for code development of automation software. There are also
iconic programming environments such as Agilent VEE or National Instruments LabView,
which allow users to create measurement automation applications without extensive programming experience. These, along with add-on libraries, utilities, and example programs
can make it easier for the non-programmer engineer to develop automation software for
his application.
Instrument-to-Host Communication. Most instruments today adhere to the SCPI (Standard Commands for Programmable Instruments) software standard which defines a common interface language for communication between instruments and computers. SCPI
commands consist of ASCII text messages in a defined syntax and can be implemented
in any PC host programming environment. The net result of conforming to this standard
is that instruments with similar functions accept the same commands to execute those
functions minimizing or eliminating software changes as instruments are replaced in test
systems.
By definition all SCPI commands are sent as ASCII text strings, however measurement data
returned by an instrument can be transferred either as ASCII text (default case) or in a
binary format which is a more compact method of data transfer. The binary data transfer
mode is configured by the host and is the most efficient manner of moving data when both
the host and instrument use the IEEE floating point format for three reasons:
•
Processing time for the Binary-to-ASCII conversion on the instrument side is eliminated
•
Processing time for the ASCII-to-binary conversion on the host side is eliminated
•
The number of bytes per value is reduced due to more efficient packing and the elimination of delimiters
The default ASCII data transfer mode is the simplest to implement since all the formatting
is performed transparently by the VISA library. It is also simplest for troubleshooting, since
the bus transactions may be easily monitored and logged using readily available tools and
utilities. But when reduced I/O traffic or data transfer speed become issues, the binary
transfer mode can improve system efficiency with a little more effort at programming time.
95
Measurement Automation. There are several compelling reasons to take the effort to
automate the testing process:
•
It provides consistency of results by eliminating possible errors from front panel manual operation
•
Timing of system events and measurements can be made totally consistent
•
Higher measurement throughput provides either reduced test time or the ability to
collect more data in the same amount of time
•
Automation facilitates large volume production testing where test time and accuracy
are paramount
•
Collecting results from multiple devices for processing and display on the PC allows for
new ways of information display beyond the capability of each stand alone device
•
Instruments can acquire data at a faster rate than their display update rate which can
be bypassed via PC control to take advantage of this acquisition rate
•
It is simpler to record and archive measurement results for quality assurance or regulatory purposes
Historically, RF power measurement in ATE systems has been performed via a query-response protocol. ASCII command and data transfer were used in both directions, either
via the SCPI or proprietary protocols. However recent advances in power measurement,
coupled with increased user requirements have now made it necessary to examine newer
solutions for increased system throughput. Modern power meters, especially peak power
meters, can have very high bandwidths and measurement rates, and their performance
can far exceed the capability of a manual or query-response process to record and view
the results.
In a typical ATE system with several instruments connected to the PC host, there is a possibility that the overall speed and efficiency of the measurement process will degrade due
to the amount of I/O traffic while communicating with the devices in the system. That is
especially true when ATE systems attempt to mimic the manual testing method of ‘measure then record’. That technique is still valid for most automated applications and will still
be faster and more consistent when controlled by a PC, but it is only as fast permitted by
the entire sequential measurement cycle. An automated, sequential measurement cycle
typically contains the following steps for each measurement that must be performed:
Chapter 8: Performance Tips
96
One Loop of Sequencial Measurement Cycle
Step 1 Configure signal source
Controller configures the signal source for next step in sweep: set power, frequency,
attenuation, etc.
Step 2 Configure power meter
Controller configures the power meter for any expected changes in the source signal: set frequency, range, filtering, etc.
Step 3 Delay for signal settling
Controller waits for a pre-programmed or adaptive time delay to allow the source
to settle before starting measurement acquisition.
Step 4 Initiate new measurement
Controller commands power meter to flush last reading and begin acquiring fresh
measurement data.
Step 5 Acquire measurement data
Power meter acquires measurement data and stops when complete. This step may
or may not run concurrently with the next (data processing) step.
Step 6 Process measurement data
Power meter processes the acquired data to yield a single measurement result.
May run concurrently with previous step (measurement acquisition).
Step 7 Notification of “measurement ready” state
Power meter notifies host that a new reading is ready via either a one-way notification or bidirectional polling process.
Step 8 Return measurement result
Controller reads back the new measurement from instrument and saves the result.
Step 9 Next measurement in sequence
Repeat steps 1 - 8 for each point to be measured.
97
Step 7, “Signal measurement ready” may consist of either a polling activity, in which the
host continuously polls the power meter to obtain the current measurement status, or via
a notification process, such as a power-meter-to-host “interrupt”, which notifies the host
when a measurement is ready. The IEEE-488 has a “Service ReQuest” (SRQ) signal line
dedicated to this purpose, and many newer remote control protocols support or emulate
this functionality. Once the host has received confirmation that a new result is ready, it
can read and record the result as needed. The software design process of an ATE system
needs to be aware and take advantage of this built-in functionality of current instruments
to optimize their operation in a remotely controlled application.
In many systems, some steps of the measurement sequence can be combined to reduce the
amount of bus traffic. For example, steps 4 through 8 are often replaced by a single queryresponse cycle, in which the power meter is told to “start a fresh measurement, stop when
done, process, and return the result”. Once the command has been given, the controller
simply waits for the power meter to return a finished result. The SCPI language allows for
this sort of control via the READ query. This eliminates the need for a polling or interruptdriven scheme for obtaining measurement status at the expense of tying up the remote
control bus while the host is waiting for the power meter to respond.
The Modern Approach. Instruments today have enough intelligence built in to allow portions of this cycle to proceed without controller intervention. For example, a signal generator can usually be programmed to perform a pre-configured power or frequency sweep
with a single command. Likewise, a power meter may be commanded to acquire and buffer a series of readings, and return all of these values at once as a delimited list or binary
data block. Sending a large number of values via a single bus transaction is considerably
more efficient than sending them one at a time. In some cases, even the device under test
may perform under “self control”, sometimes via a built-in test or diagnostic mode. One
example of this is in cellular handset test, in which the handset be commanded to sequence
through a programmed series of transmit power steps to test its power control subsystem.
Chapter 8: Performance Tips
98
Entire Buffered Sweep Measurement Cycle
Step 1 Configure and arm power meter
Controller configures the power meter to acquire and buffer a sequence of measurements to match the expected sweep.
Step 2 Configure and initiate signal source sweep
Controller configures the signal source to perform a full sweep: set start and end
power, frequency, attenuation, sweep rate/duration, etc, then initiates the sweep
(see Figure 8.2.1).
Step 3 Acquire sweep measurement data into power meter buffer
Power meter acquires measurement data for entire sweep and stops when complete.
Step 4 Notification of “sweep ready” state
Power meter notifies host that sweep is ready via either a one-way notification or
bidirectional polling process.
Step 5 Return measurement sweep result
Controller reads back all buffered readings from sweep and saves the result.
If suitable source control and measurement buffering is available, the entire measurement
process can consist of a single sequence of host transactions to initiate the process, and
final host step to read back all of the results. All other steps are managed and timed within
the source and power meter, and can proceed without controller intervention. The challenge of this approach is the difficulty in synchronizing the acquired measurement data
with the sweep progress of the source signal. Without direct communication between the
source and power meter, it can be difficult
to Ramp
determine what power step the source is on.
Power
15
10
dBm
5
dBm
0
-5
-10
-15
Time (0 to 10 seconds)
Figure 8.2.1. Power Ramp / Time-based power sweep
99
The simplest method of synchronizing the two is to use time. The source is programmed to
perform a linear sweep over a known time interval or at a known sweep rate, and the power
meter is programmed to acquire data over exactly this same time interval. If both processes are begun in synchronization, and the source sweep rate and power meter recording
rate are both accurately known, then it is not difficult to surmise the source sweep state
based on the position of a reading in the array of returned measurements. This method
works well with continuous and CW sweeps.
Another synchronization method is to use triggered acquisition, such as a hardware connection between the signal source and power meter. Many signal generators have a synchronization output that is set to a logic state to indicate when each step in a pre-programmed
sweep is occurring. Usually this signal is not asserted until the source is stable, which helps
to address the need to program a signal settling delay. In this case, this synchronization
pulse from the signal generator can then be used directly as a trigger command to the
power meter, instructing it to record a reading each time it receives the pulse. Attention
must be paid to timing and polarity – the source sweep must be programmed to occur at
a rate appropriate for the power meter to complete a measurement at each desired point
(see Figure 8.2.2).
dBm
Power Steps
15
10
5
0
-5
-10
-15
-20
-25
Time (20 trigger events)
Figure 8.2.2. Power Sweep and power meter measurement of a single pulse within the sweep
Chapter 8: Performance Tips
100
Sometimes it may be helpful to control the source sweep externally. In this case, an external pulse generator or custom hardware generates sync pulses which are sent to both
the source and the power meter. Each time a pulse is received, the source will step and the
power meter will acquire a new reading. The necessary time delays for settling and acquisition can usually be programmed into the power meter, although it is also possible to use
the pulse generator such a way as to step the source on the rising edge of the sync pulse,
and acquire a fresh measurement on its falling edge.
Another option on peak power meters is to use the signal source itself for synchronization. When the signal to be measured is a periodic or pulsed waveform, the power meter’s
internal trigger system may be used. The power meter is configured to initiate a sweep on
the leading edge of each pulse. With suitable trigger delay settings to allow pre-trigger
acquisition, the entire pulse can be acquired and processed, and its average power stored
as a reading. Control of the source can provide both the “start sweep” synchronization, and
the “acquire a new reading” signal (see Figure 8.2.3).
For instance, a GSM handset may be programmed to transmit a power sweep that tests
each level of its internal power-control attenuator. The attenuator may have 128 levels,
and the attenuation value is incremented for each GSM frame, with the handset transmitting for once timeslot in the frame. In GSM, a timeslot is 577 microseconds, and a frame
of eight timeslots is 4.615 milliseconds, so the phone will transmit a pulsed signal with a
217 Hz period.
The power meter is instructed to trigger on the signal’s rising edge for each timeslot, and
perform and record an interval average power measurement over the active portion of the
timeslot (typically the middle 550 microseconds of the pulse). A trigger holdoff setting of
about 4.5 ms should be used to ensure the power meter does not retrigger before the next
timeslot begins, and the trigger level will need to be set low enough that even the lowest
pulse amplitude will trigger a sweep.
To initiate the sweep, the power meter is configured and armed, and the source commanded to begin a sweep of 128 frames. This sweep will take 128 Frames / 217 Hz = 590
milliseconds to complete, and at the end of the sweep, the controller simply queries the
power meter to retrieve the measurement buffer containing all 128 readings. If carefully
optimized, the entire process should take less than one second.
All this is possible due to the inclusion of fast “buffered measurement modes” in modern
power meters. In these modes, the instrument acquires a series of measurements into its
memory buffer with little or no processing. If needed, the data may be post-processed
after the acquisition cycle is completed. This buffered acquisition mode is setup and executed under PC control. These results can be collected in a free run mode or by a trigger
101
Figure 8.2.3. Command sequence to set up and read the measurement buffer over the entire power sweep of 128
pulse levels
event. Data rates in the range up to 1000 readings per second are typical for the free run
mode, but the triggered mode rate may be less due to the latency time to re-arm for the
next trigger event. Buffer sizes of 1 million readings or more per channel are typical. For
buffers that are implemented as FIFO’s (First In First Out), this data collection technique
can continue beyond the time it takes to fill the buffer since FIFO’s can be read out and
sent to the host while still being filled. For example, reading the buffer when it is half filled
in most cases allows continuous measurement acquisition as read and write pointers are
updated as the FIFO is accessed. This is true as long as the time to read a fraction of the
buffer (which frees that area of the buffer) is less than the time to fill the entire buffer. If
necessary, binary data transfer can be used to reduce data transfer times.
For the CW (continuous wave) mode up to 1 million readings (typical) can be stored at
pre-programmed rates as high as 1000 readings per second for a specified number of readings (and therefore a known time duration). The data can then be transferred to the host
PC after all the readings have been stored in the buffer. This allows the user to trade off
between time resolution, measurement duration, and data set size. This mode is best for
continuous modulation formats where the power is stepped at periodic time intervals. In
this mode the power meter is free-running and readings are stored in the buffer at specific
time intervals.
For the power ramp case, the test condition is to linearly increase the power from -10 dBm
to +10 dBm over a ten second duration with a measurement rate of 20 readings per second
for a total of 200 readings which get stored every 0.5 seconds. The results are read by the
host PC every time 20 readings are stored in the buffer. The acquisition stops after the 200
readings have been collected. The read rate of the buffered results by the PC is much faster
Chapter 8: Performance Tips
102
than the rate the readings are stored in the buffer which means that this mode could be
used for long term monitoring of the power sweep over multiple power sweep cycles with
either a wait time to reset the power level or directly sweep up and down continuously. In
other words, the number of readings to acquire can be as large as the maximum size of the
buffer and the power meter will continuously update the read and write locations of the
buffer during the measurement process.
For the pulsed signal types, triggered pulse measurements can also be buffered to record
power sweeps for discontinuous formats such as GSM or TDMA. Each pulse trigger stores
a reading in the buffer which is usually the average power between two on screen cursors.
Modern power meters have powerful triggering features that allow synchronization with
most types of pulse and burst modulation formats including an external trigger event.
Buffered measurements in the pulse mode can also be used with a continuous power sweep
for the case where a trigger event is used to indicate when the power change has properly
settled in the device under test. Another case where a continuous power sweep can be
used is for devices that accept continuous power as an input and provides pulsed output
power. For these cases absolute timing of the synchronization pulses is not critical – they
only need to be repeatable. GSM timing generally has pulse repetition periods longer than
3 milliseconds should not present a problem, but signals with pulse periods less than about
3 to 4 milliseconds may skip pulses due to the trigger re-arm time which can range from 1
to 15 milliseconds or more, depending upon timebase and trigger settings.
For the above cases the data can be graphed either as output power vs. input power or
power vs. time where time is obtained from the reading rate or the trigger rate. If the trigger is an external event then the horizontal axis of the graph can represent any parameter
step change that the trigger represents such as temperature steps or distance increments.
Another application well suited for an ATE system is testing of high gain amplifiers particularly pulsed output amplifiers. A typical ATE system for this application would consist
of a two channel peak power meter, a signal generator (or sweeper) and possibly a pulse
generator to acquire and display pulsed input power, output power and gain for a given frequency range. In this configuration, the PC controls the frequency and power level from the
signal generator and collects power readings to display power or gain vs. frequency. The
frequency axis is constructed either by trigger events that step the frequency output of
the sweeper or by measuring a frequency to voltage output of the sweeper at each power
reading during the frequency sweep. This ATE configuration is very close to the functionality of a scalar analyzer. If the trigger event that yields the frequency can be associated
with other events such as time, angular or linear position, temperature, or other then the
measured power can be graphed or correlated to that event.
103
As an example of high speed power meter data acquisition Figure 8.2.4 represents a 200
point timed power sweep from -10 dBm to +10 dBm for an amplifier showing a nominal
gain of 15 dB and going into compression above +12 dBm.
20
18
Output Power (dBm)
16
14
12
10
8
6
4
2
0
-10
-5
0
5
10
Input Power (dBm)
Figure8.2.4. Typical Amplifier Gain sweep
8.3 Communication Amplifier Testing
Digital communication amplifiers for LTE and WiMAX signals require a broad frequency
band and wide dynamic range to accommodate complex modulation schemes. Traditional
figures of merit like the 1 dB compression point and the third order intercept for testing linearity will not be sufficient to account for peak-to-average ratios in excess of 15 dB. Peak
power meters like Boonton’s 4500 series provide amplifier communication designers with
an alternative method for testing amplifier linearity. This section will explain the value of
statistics for measuring the peak to average ratio of these complex digital signals.
The 1 dB compression point of an amplifier is defined as the output power at which the device’s gain drops by 1 dB from its small-signal value. This is typically known as P1dB or CP1.
To measure the compression point, a single CW tone (carrier) from an RF signal generator
is supplied to the input of the amplifier, and ratio of the output to input power is measured
to yield the amplifier’s small-signal gain. The input amplitude is then gradually increased
until that measured ratio decreases by 1 dB, representing 1 dB gain compression. This
figure is commonly used as a reference point for the beginning of amplifier non-linearity,
and is approximately equal to the maximum useable peak output power for the amplifier.
Figure 8.3.1 illustrates two historical methods for evaluating amplifier linearity with a CW
input signal. The 1 dB compression point and the third-order intercept are two figures of
merit that provide designers with assumptions about amplifier performance. In the past,
using a CW signal to measure the behavior of a narrow band amplifier was common, but for
modern broad band devices this is an unnecessary limitation.
Chapter 8: Performance Tips
104
Multiple tones can be substituted for the CW test signal, or a modulated or noise signal
with frequency components across the amplifier’s entire operating band can be used. For
communication amplifiers that typically amplify signals with a large peak-to-average power ratio (crest factor), the average input power must be reduced sufficiently so that the
expected peaks rarely or never saturate the device. Depending upon the amplifier’s target
application, designers typically estimate an amplifier’s maximum average output power
will be 6 to 12dB lower than its 1dB compression point. This estimate will allow digital
signals with a high crest factor to saturate the amplifier only occasionally to maintain an
acceptable BER.
Output
Pwr
Single Tone
Output
Pwr
Two Tones
1P3
1dB comp pt.
1dB/dB
3dB/dB
1dBCP
Input Pwr
IM3
Input Pwr
Figure 8.3.1. The 1 dB compression point and third-order intercept figures of merit
The 3rd order intercept point is an additional figure of merit using the third order harmonic product of two mixed tones. This is commonly referred to as IP3, short for the third
order intercept. The third order harmonic product of two mixed tones is referred to as IM3,
or the third order inter-modulation product. This is the beginning of a line having a 3:1
slope that intersects with the 1:1 slope of input vs. output power at IP3, or the third order
intercept point. These values are shown in Figure 8.3.1.
The 1 dB compression point and IP3 are two figures of merit used to estimate the spurious
free dynamic range (SFDR) of the amplifier. The SFDR is calculated by selecting a point 1/3
the Y-axis distance below the third order intercept value with respect to the thermal noise
limit of a 50 ohm resistor, which is -174 dBm/Hz. The signal’s useful dynamic range, or
UDR is equal to the SFDR minus the measurable noise floor of the amplifier (NF). The UDR
is calculated by selecting a point 1/3 the distance in dB from P3 to the amplifier noise floor
assuring that IM3 products will be below that noise floor. The noise floor can be estimated
or measured using noise figure. These assumptions may artificially limit system dynamic
range (see Figure 8.3.2).
105
P3
IP3
10dB
P1dB
1/3
3dB
2/3
UDR
1dB
SFDR
third harmonic
Power
Out
NF
Thermal Noise
Power In
PIM3
Figure 8.3.2. IP3 and IM3 figure of merit, and SFDR of an amplifer
Amplifier compression test system. A simple test to measure the 1 dB compression point
is shown in Figure 8.3.3 below, and requires the following equipment: a CW Signal source,
dual-channel power meter, two suitable power sensors, a directional coupler, test amplifier,
power supply, and all the necessary cable connections. For higher power systems, some
form of additional signal attenuation may be required to maintain signal levels within the
range that can be measured by the power sensors.
Signal Source
Power Meter
14.7 dBm
Reference
Sensor
TP1
Coupler
TP2
Measurement
Sensor
DUT
15Y
Bias
Supply
Figure 8.3.3. Amplifier Compression Test System Block Diagram
Chapter 8: Performance Tips
106
Part 1: Measure the 1 dB compression point of the DUT
The 1 dB compression point of an amplifier can also be measured using a peak power meter
in average mode as illustrated in Figure 8.3.4. Measure the input, and output of the amplifier simultaneously at low-to-medium output power, and compute the small-signal gain.
The left screen shot shows the input power on Channel 1, and output power on Channel 2.
The gain is the ratio of Channel 2 to Channel 1, and is shown in the lower right of the bottom window. At this operating level (2 dBm output power), the measured gain is 20.026
dB.
The right side of figure 8.3.4 showsthe input power has been increased until the automatically computed gain value in the lower window has fallen by 1 dB from its small-signal
value on the left side. The output power at this point is measured to be 15 dBm, which
would be the P1dB figure-of-merit for this amplifier.
Figure 8.3.4. Two-Channel measurements showing amplifier gain (Channel 2 to Channel 1 ratio) dropping by 1dB
Using statistics to measure the peak to average ratio. Unlike simple, average power
measurements using CW tones, statistical analysis can be used to compare the peak to
average ratio of the signal in percent with respect to the total signal time. The peak power
values are sorted, or binned according to their magnitude. They can be divided by the average power and displayed in log-log plot in dBr as crest factor. A common communications
display is the CCDF, or complementary cumulative distribution function. The CCDF in Figure
8.3.5 shows how long a particular crest factor is being transmitted as a percentage of total
signal time. 0 dBr is equal to the average power and 0% time is equal to maximum crest
factor. In this example, a crest factor probability of 0.0001% occurs at 15 dBr with respect
to the average.
107
Frequent
Events
Increasing
Probability
Infrequent
Events
Increasing Crest Factor
Figure 8.3.5 CCDF of a modulated communication signal
Demonstration signal. The time-domain screen capture shown Figure 8.3.6, below is a
noise-like wireless digital communication signal that is difficult to measure using a spectrum analyzer or oscilloscope. The power meter is an effective tool for measuring the crest
factor of these signals using statistical analysis, however a time-domain display does not
provide much information about peak power events that occur very infrequently. Statistical analysis is a much more useful technique which makes it possible to characterize the
probability of occurrence of rare peak (high crest factor) events. Since amplifiers and other
devices are most stressed by these peaks, that is where signal compression or clipping will
be most likely to result in a loss of data. A meaningful statistical measurement requires
a large number of samples and Boonton power meters can collect millions of peak power
measurements in seconds.
Figure 8.3.6. A noise-like CDMA communication signal displayed in the time domain
Chapter 8: Performance Tips
108
Frequent
Events
Increasing
Probability
Infrequent
Events
Increasing Crest Factor
Figure 8.3.7. Dual CCDF of typical CDMA signal
Using CCDF distributions to compare amplifier input and output. The screen shot in Figure 8.3.7 contains two CCDF distributions. The transmitter signal in blue is the reference,
or input channel and the yellow signal is the output of the DUT. This allows direct comparison of input vs output using the normalized CCDF of the peak to average ratios. This view
allows comparison of the signal over the entire dynamic range of the amplifier using the
particular modulation format and frequencies of interest.
The graph button on a Boonton 4540 allows the user to toggle between the 1 dB compression point, and the dual CCDF display. The difference between a narrow band figure of
merit and the statistical display is illustrated by the peak to average ratio deviation, well
before the 1 dB compression point is reached. The crest factor is not measured using the 1
dB, or 3rd order intercept point figures of merit. This is clearly shown in Figure 8.3.8.
The statistical methods used in this discussion can infer BER qualities before the entire
receiver circuit has been assembled. This can save valuable design, and compliance testing
time.
Toggle between
screens using the
graph button
Figure 8.3.8. Difference between conventional (peak and average) and CCDF statistical analysis
109
Chapter 9: Measurement Accuracy
RF power meters have long been accepted as accurate measurement standards, but just
how accurate are the power measurements they yield? Calculating the accuracy of average and peak RF power measurements requires more than just a glance at a specification
sheet. While many users ask “how accurate is the meter,” the metrologist more properly
asks “what is the uncertainty of this particular measurement?” Uncertainty is a quantifiable measurement of exactly where measurement errors are, or more importantly, may be
present.
This chapter will examine the various sources of measurement uncertainty, describe how
these uncertainties are combined to yield a single uncertainty value, and show two sample
uncertainty calculations for typical power measurement scenarios.
9.1 Introduction to Uncertainty
RF power measurement accuracy depends on a variety of factors derived from the measuring instrument, device under test (often abbreviated “DUT”), characteristics of the signal
being measured, instrument settings, and environmental factors. Calibration, signal frequency, level and modulation, source and load mismatch, and noise all play an important
role in the determining total uncertainty.
Each of these factors adds its own contribution to the uncertainty – some large and some
negligible. These “uncertainty terms” can be combined mathematically to yield a single
uncertainty value for a particular power measurement. However, many of the values can
vary considerably from one measurement application to the next, even with identical
equipment, so there is never a single “data sheet value” for power measurement accuracy.
Uncertainty Values. When combining uncertainty values, it is important that all numbers
be in the same units. Uncertainty is typically given as a fraction or percentage, with 0.0%
indicating that term does not contribute any error to the reading. Sometimes, uncertainty
values are provided as a logarithmic value, often in the form of “± x.xx dB”. The following
formulas may be used to convert between percent uncertainty and dB tolerance:
U% = (10(UdB/10) -1) × 100
and
UdB = 10 × Log10(1 + (U% / 100))
Worst Case Uncertainty. Uncertainty values for each term are usually specified as “worstcase” values – that is the measurement uncertainty due to that particular item will never
be greater than the specification. In some cases, “typical” values are also given, and can be
used to better understand the characteristic of that term.
Chapter 9: Measurement Accuracy
110
The “combined worst-case uncertainty” approach is a very conservative method for calculating accuracy where the worst-case values of each individual uncertainty term are added
together. The formula for worst-case measurement uncertainty is:
UWorstCase = U1 + U2 + U3 + U4 + ... UN
But typically most of the uncertainty terms are independent of one another, so the probability of them all existing at worst-case conditions simultaneously is extremely small. For
this reason, a more realistic approach known as the “root-sum-of-squares” (RSS) technique is generally used to combine the terms to yield a single, expected uncertainty value. To compute combined uncertainty with the RSS technique, each uncertainty term is
squared, the squares are added together, and the square root of the resulting summation
is calculated.
URSS = √ ( U12 + U22 + U32 + U42 + ... UN2 )
Uncertainty Distributions. A problem with the basic RSS method is that it does not account for the fact that the distribution of errors for each term may have different shapes
within the worst-case uncertainty bounds. Certain types of errors will vary with a normal,
or Gaussian distribution, with most existing within a narrow range, and fewer errors for
larger error values. The worst-case limits will typically be set several standard-deviations
away, and often some method of guaranteeing conformance will exist. Other errors may
vary linearly within a set of bounds, yielding a rectangular distribution, with equal probability of any error within the limit bounds, large or small. Other distribution shapes such
as “U-shaped” are possible as well – often resulting from normal distributions that are “cut
off” or forced within a range by some adjustment process.
To account for these varying probabilities of these error distributions, the worst-case uncertainty values for each term may be scaled, or normalized by an appropriate constant
to adjust for that term’s probability distribution or shape. Once the worst-case values are
normalized in this way, the RSS process can yield more meaningful result.
The distribution shape is a statistical description of how the actual error values are likely to
vary from the ideal value. Three main types of distributions are normal (Gaussian), rectangular, and U-shaped. The “K” multipliers for each type of distribution are:
111
Distribution Multiplier
K
Normal
√ (1/4) = 0.500
Rectangular
√ (1/3) = 0.577
U-shaped
√ (1/2) = 0.707
The formula for calculating RSS measurement uncertainty from worst-case values and distribution shape scale factors is:
URSS = √ [ (U1K1)2 + (U2K2)2 + (U3K3)2 + (U4K4)2 + ... (UNKN)2 ]
where U1 through UN represent each of the worst-case uncertainty terms, and K1 through
KN represent the normalizing multipliers for each term based on its distribution shape.
This calculation yields what is commonly referred to as the combined standard uncertainty,
or UC, with a level of confidence of approximately 68 %. To gain higher levels of confidence
the Expanded Uncertainty is often employed. Using a statistical coverage factor of two will
provide an expanded uncertainty with a confidence level of approximately 95%:
UEXPANDED = 2 × UC
This is generally accepted method within the RF power measurement industry.
9.2 Power Measurement Uncertainty Contributions
As outlined above, there are a number of factors that contribute to uncertainty in RF peak
power measurements. This section will define the major factors. It should be noted that
there is no standard set of defined uncertainty terms, and different instrument manufacturers may group or name the terms differently. The following list describes the terms
that are used for computing uncertainty for power measurements performed with Boonton
Peak and CW power meters and sensors.
Instrument Uncertainty. This term represents the amplification and digitization uncertainty in the power meter, as well as internal component temperature drift. Cable and
connector loss between the sensor and power meter may also be included. In most cases,
this is very small, since absolute errors in the circuitry are calibrated out by field calibration
processes such as sensor zero and fixed calibration, or sensor step calibration (“AutoCal”),
leaving only relative linearity errors. Instrument Uncertainty is typically a datasheet value.
Calibrator Level Uncertainty. This term is the uncertainty in the calibrator’s output level
for a given setting for calibrators that are maintained in calibrated condition. The value of
is term is typically a single datasheet value for fixed-output (0 dBm) power references, and
will vary with level for variable-output calibrators.
The value to use for calibration level uncertainty depends upon the sensor calibration technique used. If AutoCal was performed, the calibrator’s uncertainty at the measurement
power level should be used. For sensors calibrated with the “FixedCal” method, the calibrator is only used as a single-level source, and you should use the calibrator’s uncertainty
Chapter 9: Measurement Accuracy
112
value at the FixedCal level, 0 dBm for most sensors. This may make FixedCal seem more
accurate than AutoCal at some levels, but this is usually more than offset by the reduction
in shaping error afforded by the AutoCal technique.
For sensors that are not field calibrated, the calibrator uncertainty term may be neglected.
Calibrator Mismatch Uncertainty. This term is the mismatch error caused by impedance
differences between the calibrator output and the sensor’s termination. It is calculated
from the reflection coefficients of the calibrator (rCAL) and sensor (rSNSR) at the calibration
frequency with the following equation:
Calibrator Mismatch Uncertainty = ±2 × rCAL × rSNSR × 100 %
The calibrator reflection coefficient is typically a specification for the calibrator or RF power
reference output, and may be provided as a VSWR. If so, the corresponding reflection coefficient may be looked up in the reference table in Section 10.2 of this guide, or computed
with the following formula:
Reflection Coefficient (rCAL) = (VSWR -1) / (VSWR +1)
The sensor reflection coefficient, rSNSR is frequency dependent, and is usually provided as
a sensor datasheet value. If it is listed as a VSWR value, the equation above may be used
to convert.
Source Mismatch Uncertainty. This term is the mismatch error caused by impedance differences between the measurement source’s output and the sensor’s termination. For
many measurements, this is the single largest error term, and care should be used to ensure
the best possible match between source and sensor. Both the sensor and DUT contribute
to this term, as does the mating of the two connectors. The source mismatch uncertainty
value is calculated from the reflection coefficients of the source (rSRCE) and sensor (rSNSR) at
the measurement frequency with the following equation:
Source Mismatch Uncertainty = ±2 × rSRCE × rSNSR × 100 %
The source reflection coefficient is a characteristic of the RF source under test, and will usually vary with frequency, and sometimes level. It must be supplied, measured or estimated.
The sensor reflection coefficient, rSNSR is frequency dependent, and is usually provided as a
sensor datasheet value.
113
Both the source and sensor reflection coefficients may be computed from their corresponding VSWR values using the formula above if required.
Sensor Shaping Error. This term is sometimes called “linearity error” and is the residual
non-linearity in the measurement after the sensor’s output has been linearized and scaled
to a power value by the calibration and shaping processes.
Calibration is typically performed at discrete level steps and is extended to all levels via
a curve-fitting technique. The sensor shaping error is close to zero at these calibration
points, and increases in between due to imperfections in the curve-fitting algorithm.
An additionally component of sensor shaping error is due to the fact that the sensor’s
transfer function may not be identical at all frequencies. The published shaping error includes terms to account for these deviations.
If your measurement frequency is close to your field calibration frequency and you are
measuring at a level close to one of the field linearity calibration points, it is probably acceptable to use a value lower than the published uncertainty in your calculations.
Sensor Temperature Coefficient. This term is the error that occurs when the sensor’s temperature has changed significantly from the temperature at which the sensor calibration
was performed. It is usually specified in dB/degree C or similar units, and may be a single
value or level dependent. In this case, the term’s value must be computed based on the difference between the current measurement temperature and the calibration temperature,
and then scaled to a percent uncertainty.
Note that when the measurement is performed at the exact same temperature as field
calibration, this term will cancel to zero.
Sensor Noise. This term is the uncertainty contribution by the intrinsic noise that is a
part of all power measurements. It is not necessarily generated within the sensor itself,
but rather is noise from the entire measurement chain. For convenience, however, the
measurement noise is usually specified as input referred noise, and is the amount of power
(usually in dBm or nanowatts) which appears to be present at the sensor input.
Sensor noise is an apparent power quantity, and may be reduced by filtering or averaging
as discussed in Section 8.1 of this guide. The noise usually approximates bandlimited white
Gaussian noise, and therefore its amplitude (and uncertainty contribution) may be reduced
by narrowing the measurement bandwidth.
Chapter 9: Measurement Accuracy
114
Sensor datasheets often give the effective noise at a certain level of filtering. For continuous-mode measurements, the degree of filtering may be specified as an integration
time or reading averaging factor. For Pulse-mode measurements, a video averaging figure
is often used. Regardless of the method, it is important to understand the relationships
between the acquisition mode, sample rate, and filtering to ensure that the correct noise
power value is used.
The uncertainty due to sensor noise depends upon the ratio of the noise to the signal power
being measured:
Noise Uncertainty = ± Sensor Noise (in watts) / Signal Power (in watts) × 100 %
The noise rating of a particular power sensor may be found on the sensor datasheet. It may
be necessary to adjust the sensor noise for more or less filtering or averaging, depending
upon the application. As a general rule (within a decade of the datasheet point), noise is inversely proportional to the filter time or averaging used. Noise error is usually insignificant
when measuring at high levels (25 dB or more above the sensor’s minimum power rating).
Sensor Zero Drift. This term is the reading uncertainty due to long-term change in the
zero-power reading that is not a random, noise component. Increasing filter or averaging
will not reduce zero drift. For low-level measurements, this can be controlled by zeroing the
meter just before performing the measurement.
The uncertainty due to zero drift depends upon the ratio of the drift to the signal power
being measured:
Zero Drift Uncertainty = ± Sensor Zero Drift (in watts) / Signal Power (in watts) × 100 %
Like sensor noise, Zero drift error is usually insignificant when measuring at 25 dB or more
above the sensor’s minimum power rating. The drift specification sometimes indicates a
time interval such as one hour. If the time since performing a sensor Zero or AutoCal is very
short, the zero drift is greatly reduced.
Sensor Frequency Calibration Factors. Sensor frequency calibration factors, or “calfactors” are used to correct for sensor frequency response deviations. These calfactors are
characterized during factory calibration of each sensor by measuring its output at a series
of test frequencies spanning its full operating range, and storing the ratio of the actual
applied power to the measured power at each frequency. During measurement operation,
the power reading is multiplied by the calfactor for the current measurement frequency to
correct the reading for a flat response.
115
Calfactors are also called “efficiency factors,” and can be provided as either a percent or a
dB correction value. They are usually factory generated, but like any calibration, include
a number of uncertainties. These uncertainties occur due to both standards uncertainty,
and measurement uncertainty in the calibration process, and will be different for each frequency calibrated. Both worst case and RSS uncertainties are provided for the frequency
range covered by each sensor, and are listed on the sensor datasheet and in the Boonton
Electronics Power Sensor Manual.
If the measurement frequency is between sensor calfactor entries, the most conservative
approach is to use the higher of the two corresponding uncertainty figures. It is also be
possible to estimate the figure by linear interpolation.
If the measurement frequency is identical to the field calibration frequency, a calfactor
uncertainty value of zero may be used, since any absolute error in the calfactor cancels
out during field calibration. At frequencies that are close to the calibration frequency, the
calfactor uncertainty is only partially cancelled out during calibration, so it is generally acceptable to take the uncertainty for the next closest frequency, and scale it down.
9.3 Sample Uncertainty Calculations
The following examples shows calculations for two measurement scenarios using the Boonton Model 4540 RF Power Meter. The first is a CW example using the 51075 CW sensor,
a general-purpose, 18 GHz CW Dual-Diode power sensor. The second example shows the
same instrument performing a measurement with a 57518 power sensor, a typical choice
for mid-bandwidth modulated signals such as CDMA. Additional information on these
Boonton products may be found in Chapter 11 of this guide.
The figures used in these examples are meant to show the general technique, and do not
apply to every application. Some common sense assumptions have been made to illustrate
the fact that uncertainty calculation is not an exact science, and requires some understanding of your specific measurement conditions. These are in spirit with the RSS method,
which is itself a method for estimation.
Typical Example #1: Boonton 4540 RF Power Meter with 51075 CW Power Sensor
Measurement Conditions
Source Frequency
10.3 GHz
Source Power
-55 dBm (3.16 nW)
Source VSWR
1.50 (reflection coefficient = 0.20) at 10.3 GHz
AutoCal Source
Internal 50 MHz Calibrator
AutoCal Temperature
25°C
Current Temperature
25°C
Chapter 9: Measurement Accuracy
116
In this example, assume that an AutoCal was performed on the sensor immediately before
the measurement. This will reduce certain uncertainty terms, as discussed below.
Step 1: The Instrument Uncertainty figure for the 4540 Series is ± 0.20 %. Since a portion
of this figure includes temperature drift, the instrument uncertainty is ± 0.10 %, or half the
published figure.
UInstrument = ± 0.10%
Step 2: The Calibrator Level Uncertainty for the power meter’s internal, 50 MHz calibrator
may be read from the calibrator’s specification. It is ± 0.105 dB, or ± 2.45 % at a level of
-55 dBm.
UCal Level = ± 2.45%
Step 3: The Calibrator Mismatch Uncertainty is calculated using the formula in the previous
section, using the internal 50 MHz calibrator’s published figure for rCAL and calculating the
value rSNSR from the VSWR specification on the 51075 sensor datasheet.
rCAL = 0.024 (internal calibrator’s reflection coefficient at 50 MHz)
rSNSR = (1.15 - 1) / (1.15 + 1) = 0.070
UCalMismatch = ± 2 × rCAL × rSNSR × 100 %
= ± 2 × 0.024 × 0.070 × 100 %
= ± 0.34%
(calculate reflection coefficient of 51075, max VSWR = 1.15 at 50 MHz)
Step 4: The Source Mismatch Uncertainty is calculated using the formula in the previous
section, using the DUT’s specification for rSRCE and calculating the value rSNSR from the VSWR
specification on the 51075 sensor datasheet.
rSRCE = 0.2 (DUT source reflection coefficient at 10.3 GHz)
rSNSR = (1.40 - 1) / (1.40 + 1) = 0.167
(calculate reflection coefficient of 51075, max VSWR = 1.40 at 10.3 GHz)
USourceMismatch = ± 2 × rSRCE × rSNSR × 100 %
= ± 2 × 0.20 × 0.167 × 100 %
= ± 6.68%
117
Step 5: The uncertainty caused by Sensor Shaping Error for a 51075 CW sensor that has
been calibrated using the AutoCal method can be assumed to be 1.0%, as per the discussion in the previous section.
UShapingError
= ± 1.0%
Step 6: The Sensor Temperature Drift Error depends on how far the temperature has drifted
from the sensor calibration temperature, and the temperature coefficient of the sensor. In
this example, an AutoCal has just been performed on the sensor, and the temperature has
not drifted at all, so we can assume a value of zero for sensor temperature drift uncertainty.
USnsrTempDrift
= ± 0.0%
Step 7: This is a relatively low-level measurement, so the noise contribution of the sensor
must be included in the uncertainty calculations. We’ll assume default filtering. The signal
level is -55 dBm, or 3.16 nW. The RMS noise specification for the 51075 sensor is 30 pW,
from the sensor’s datasheet. Noise uncertainty is the ratio of these two figures expressed
as a percentage.
UNoise Error = ± Sensor Noise (in watts) / Signal Power (in watts)
= ± (30.0e -12 / 3.16e -9) × 100 %
= ± 0.95%
Step 8: The Sensor Zero Drift calculation is very similar to the noise calculation. For sensor
zero drift, the datasheet specification for the 51075 sensor is 100 pW, so we’ll take the
liberty of cutting this in half to 50 pW, since we just performed an AutoCal, and it’s likely
that the sensor hasn’t drifted much.
UZero Drift = ± Sensor Zero Drift (in watts) / Signal Power (in watts)
= ± (50.0e -12 / 3.16e -9) × 100 %
= ± 1.58%
Step 9: The Sensor Calfactor Uncertainty is calculated from the uncertainty values in the
Boonton Electronics Power Sensor Manual. There is no entry for 10.3 GHz, so we’ll have
to look at the two closest entries. At 10 GHz, the calfactor uncertainty is 4.0 %, and at 11
GHz it is 4.3 %. These two values are fairly close, so we’ll perform a linear interpolation to
estimate the uncertainty at 10.3 GHz:
UCalFactor = [ ( F - F1 ) × (( CF2 - CF1 ) / ( F2 - F1 )) ] + CF1
= [ ( 10.3 - 10.0 ) × (( 4.3 - 4.0 ) / ( 11.0 - 10.0 )) ] + 4.0
= 4.09%
Chapter 9: Measurement Accuracy
118
Step 10: Now that each of the individual uncertainty terms has been determined, we can
combine them to calculate the worst-case and RSS uncertainty values: U (± %), (U×K)2
( %2 )
Uncertainty Term
UWorstCase
“K”
(U×K)2
(± %)
Distribution Multiplier
( %2 )
1. instrument
0.10
0.500
0.0025
2. calibrator level
2.45
0.577
1.9984
3. calibrator mismatch
0.34
0.707
0.0578
4. source mismatch
6.68
0.707
22.305
5. sensor shaping error
1.00
0.577
0.3333
6. sensor temperature drift
0.00
0.577
0.0000
7. sensor noise
0.95
0.500
0.2256
8. sensor zero drift
1.58
0.577
0.8311
9. sensor calibration factor
4.09
0.500
4.1820
Total Worst Case Uncertainty
± 17.19 %
Total sum of squares
Combined Standard
RSS
29.936 %2
±5.47%
RSS Uncertainty UC
Expanded RSS Uncertainty
±10.94%
U (coverage factor k = 2)
(±0.45 dB)
In this example, the two largest contributions to total uncertainty are the source mismatch
and the sensor CalFactor. Also note that the expanded uncertainty is approximately onehalf the value of the worst-case uncertainty. This is not surprising, since the majority of
the uncertainty comes from just two sources. If the measurement frequency was lower,
these two terms would be reduced, and the expanded uncertainty would probably be less
than half the worst-case. Conversely, if one term dominated (for example if a very low level
measurement was being performed, and the noise uncertainty was 30%), the expanded
uncertainty value would be expected to approach the worst-case value. The expanded
uncertainty is 0.45 dB.
119
Typical Example #2: Boonton 4540 RF Power Meter with 57518 Peak Power Sensor
Measurement Conditions
Source Frequency
900 MHz
Source Power
13 dBm (20 mW)
Source VSWR
1.12 (reflection coefficient = 0.057) at 900 MHz
AutoCal Source
Boonton Model 2530 1 GHz Calibrator
AutoCal Temperature
38°C
Current Temperature
49°C
In this example, we will assume that an AutoCal was performed on the sensor earlier in the
day, so time and temperature drift may play a role in the uncertainty.
Step 1: The Instrument Uncertainty figure for the 4540 Series is ± 0.20 %. Since it has
been a while since AutoCal, we’ll use the published figure.
UInstrument = ± 0.20%
Step 2: The Calibrator Level Uncertainty for the Model 2530 External 1 GHz calibrator may
be calculated from the calibrator’s specification. The 0 dBm uncertainty is 0.065 dB, or
1.51 %. To this figure, we must add 0.03 dB or 0.69 % per 5 dB step from 0 dBm. 13 dBm
is 2.6 5 dB steps (13/5) away from 0 dBm. Any fraction must always be rounded to the
next highest whole number, so we’re 3 steps away.
UCalLevel = ± (1.51 % + (3 × 0.69 %))
= ± 3.11%
Step 3: The Calibrator Mismatch Uncertainty is calculated using the formula in the previous
section, using the 1 GHz calibrator’s published figure for rCAL and calculating the value rSNSR
from the VSWR specification on the 57518 sensor datasheet.
rCAL = 0.091 (internal 1 GHz calibrator’s reflection coefficient)
rSNSR = (1.15 - 1) / (1.15 + 1) = 0.070
UCalMismatch = ± 2 × rCAL × rSNSR × 100 %
= ± 2 × 0.091 × 0.070 × 100 %
= ± 1.27%
(calculate reflection coefficient of 57518, max VSWR = 1.15 at 1 GHz)
Chapter 9: Measurement Accuracy
120
Step 4: The Source Mismatch Uncertainty is calculated using the formula in the previous
section, using the DUT’s specification for rSRCE and calculating the value rSNSR from the VSWR
specification on the 57518 sensor datasheet.
rSRCE = 0.057 (DUT source reflection coefficient at 900 MHz)
rSNSR
= (1.15 - 1) / (1.15 + 1) = 0.070
(calculate reflection coefficient of 57518, max VSWR = 1.15 at 0.9 GHz)
USourceMismatch = ± 2 × rSRCE × rSNSR × 100 %
= ± 2 × 0.057 × 0.070 × 100 %
= ± 0.80%
Step 5: The uncertainty caused by Sensor Shaping Error for a 57518 peak sensor is 4 % at
all levels, from the sensor’s datasheet. But since we’re measuring at 900 MHz, which is very
close to the 1 GHz AutoCal frequency, we’ll assume that the frequency-dependent portion
of the shaping error becomes very small, and we’ll estimate that 2 % remains.
UShapingError
= ± 2.0%
Step 6: The Sensor Temperature Drift Error depends on how far the temperature has drifted
from the sensor calibration temperature, and the temperature coefficient of the sensor. In
our case, we are using a temperature compensated sensor, and the temperature has drifted
by 11 degrees C (49° C – 38° C) from the AutoCal temperature. We will use the equation in
the previous section to calculate sensor temperature drift uncertainty.
USnsrTempDrift = ± (0.93 % + 0.069 % /°C)
= ± (0.93 + (0.069 / 11.0)) %
= ± 1.69%
Step 7: This is a high-level measurement, and the noise contribution of the sensor is negligible, but we’ll calculate it anyway. Use the meter in modulated mode with default filtering.
The signal level is 13 dBm, or 20 mW. The “noise and drift” specification for the 57518
sensor is 50 nW, from the sensor’s datasheet. Noise uncertainty is the ratio of these two
figures, expressed as a percentage.
UNoise&Drift 121
= ± Sensor Noise (in watts) / Signal Power (in watts)
= ± (50.0e - 9 / 20.0e - 3) × 100 %
= ± 0.0003%
Step 8: A separate Sensor Zero Drift calculation does not need to be performed for peak
sensors, since “noise and drift” are combined into one specification, so we’ll just skip this
step.
Step 9: The Sensor Calfactor Uncertainty needs to be interpolated from the uncertainty
values in the Boonton Electronics Power Sensor Manual. At 1 GHz, the sensor’s calfactor
uncertainty is 2.23 %, and at 0.5 GHz it is 1.99 %. Note, however, that we are performing
our AutoCal at a frequency of 1 GHz, which is very close to the measurement frequency.
This means that the calfactor uncertainty cancels to zero at 1 GHz, as discussed in the
previous section. We’ll use linear interpolation between 0.5 GHz and 1 GHz to estimate a
value. 900 MHz is only 20 % (one fifth) of the way from 1 GHz down to 500 MHz, so the
uncertainty figure at 0.5 GHz can be scaled by one fifth. UCalFactor
= 1.99% × (900 - 1000) / (500 - 1000)
= 1.99% × 0.2
= ± 0.40%
Step 10: Now that each of the individual uncertainty terms has been determined, we can
combine them to calculate the worst-case and RSS uncertainty values:
Uncertainty Term
UWorstCase
“K”
(U×K)2
(± %)
Distribution Multiplier
( %2 )
1. instrument
0.20
0.500
0.0025
2. calibrator level
3.11
0.577
3.2201
3. calibrator mismatch
1.27
0.707
0.8062
4. source mismatch
0.80
0.707
0.3199
5. sensor shaping error
2.00
0.577
1.3333
6. sensor temperature drift
1.69
0.577
0.9509
7. sensor noise & drift
0.00
0.500
0.0000
8. sensor calibration factor
0.40
0.500
0.0400
Total Worst Case Uncertainty
± 18.43 %
Total sum of squares
Combined Standard
RSS
6.6729 %2
±2.58%
RSS Uncertainty UC
Expanded RSS Uncertainty
±5.17%
U (coverage factor k = 2)
(±0.22 dB)
Chapter 9: Measurement Accuracy
122
From this example, different error terms dominate. Since the measurement is close to the
calibration frequency, and matching is rather good, the shaping and level errors are the
largest. Expanded uncertainty of 5.17% translates to an uncertainty of about 0.22 dB in
the reading.
It should be noted that measurement uncertainty calculation is a very complex process, and
the techniques shown here are somewhat simplified to allow easier calculation. For more
complete information, the following publications may be consulted:
1. “ISO Guide to the Expression of Uncertainty in Measurement”
©1995, International Organization for Standardization, Geneva, Switzerland
ISBN 92-67-10188-9
2. “U.S. Guide to the Expression of Uncertainty in Measurement”
© 1996, National Conference of Standards Laboratories, Boulder, CO 80301
ANSI/NCSL Z540-2-1996
123
Chapter 9: Measurement Accuracy
124
Section 3
Power Measurement Reference
This section is a collection of reference material useful to RF and microwave
engineers. The first chapter includes amplitude measurement conversions,
return-loss, reflection coefficient, VSWR conversions and forward and reverse power tables. The second chapter includes a list of peak and average
power meters, including available sensors.
125
Chapter 10: Reference Tables
10.1 Amplitude Measurement Conversions
This is a chart of conversions for commonly used RF amplitude measurement units. dBm,
watts, volts, etc, including conversion equations.
dBm
dBW
Watts
Volts
Amps
dBV
dBmV
into 50Ω
into 50Ω
into 50Ω
into 50Ω
90.00
60.00
1.00 MW
7.07 kV
141 A
76.99
136.99
80.00
50.00
100 kW
2.24 kV
44.7 A
66.99
126.99
70.00
40.00
10.00 kW
707 V
14.1 A
56.99
116.99
60.00
30.00
1.00 kW
224 V
4.47 A
46.99
106.99
50.00
20.00
100 W
70.7 V
1.41 A
36.99
96.99
40.00
10.00
10.0 W
22.4 V
447 mA
26.99
86.99
30.00
0.00
1.00 W
7.07 V
141 mA
16.99
76.99
29.00
-1.00
794 mW
6.30 V
126 mA
15.99
75.99
28.00
-2.00
631 mW
5.62 V
112 mA
14.99
74.99
27.00
-3.00
501 mW
5.01 V
100 mA
13.99
73.99
26.00
-4.00
398 mW
4.46 V
89.2 mA
12.99
72.99
25.00
-5.00
316 mW
3.98 V
79.5 mA
11.99
71.99
24.00
-6.00
251 mW
3.54 V
70.9 mA
10.99
70.99
23.00
-7.00
200 mW
3.16 V
63.2 mA
9.99
69.99
22.00
-8.00
158 mW
2.82 V
56.3 mA
8.99
68.99
21.00
-9.00
126 mW
2.51 V
50.2 mA
7.99
67.99
20.00
-10.00
100 mW
2.24 V
44.7 mA
6.99
66.99
19.00
-11.00
79.4 mW
1.99 V
39.9 mA
5.99
65.99
18.00
-12.00
63.1 mW
1.78 V
35.5 mA
4.99
64.99
17.00
-13.00
50.1 mW
1.58 V
31.7 mA
3.99
63.99
16.00
-14.00
39.8 mW
1.41 V
28.2 mA
2.99
62.99
15.00
-15.00
31.6 mW
1.26 V
25.1 mA
1.99
61.99
14.00
-16.00
25.1 mW
1.12 V
22.4 mA
0.99
60.99
13.00
-17.00
20.0 mW
999 mV
20.0 mA
-0.01
59.99
12.00
-18.00
15.8 mW
890 mV
17.8 mA
-1.01
58.99
11.00
-19.00
12.6 mW
793 mV
15.9 mA
-2.01
57.99
10.00
-20.00
10.0 mW
707 mV
14.1 mA
-3.01
56.99
9.00
-21.00
7.94 mW
630 mV
12.6 mA
-4.01
55.99
8.00
-22.00
6.31 mW
562 mV
11.2 mA
-5.01
54.99
7.00
-23.00
5.01 mW
501 mV
10.0 mA
-6.01
53.99
6.00
-24.00
3.98 mW
446 mV
8.92 mA
-7.01
52.99
Chapter 10: Reference Tables
126
dBm
dBW
Watts
Volts
Amps
dBV
dBmV
into 50Ω
into 50Ω
into 50Ω
into 50Ω
5.00
-25.00
3.16 mW
398 mV
7.95 mA
-8.01
51.99
4.00
-26.00
2.51 mW
354 mV
7.09 mA
-9.01
50.99
3.00
-27.00
2.00 mW
316 mV
6.32 mA
-10.01
49.99
2.00
-28.00
1.58 mW
282 mV
5.63 mA
-11.01
48.99
1.00
-29.00
1.26 mW
251 mV
5.02 mA
-12.01
47.99
0.00
-30.00
1.00 mW
224 mV
4.47 mA
-13.01
46.99
-10.00
-40.00
100 µW
70.7 mV
1.41 mA
-23.01
36.99
-20.00
-50.00
10.0 µW
22.4 mV
447 µA
-33.01
26.99
-30.00
-60.00
1.00 µW
7.07 mV
141 µA
-43.01
16.99
-40.00
-70.00
100 nW
2.24 mV
44.7 µA
-53.01
6.99
-50.00
-80.00
10.0 nW
707 µV
14.1 µA
-63.01
-3.01
-60.00
-90.00
1.00 nW
224 µV
4.47 µA
-73.01
-13.01
-70.00
-100.00
100 pW
70.7 µV
1.41 µA
-83.01
-23.01
-80.00
-110.00
10.0 pW
22.4 µV
447 nA
-93.01
-33.01
-90.00
-120.00
1.00 pW
7.07 µV
141 nA
-103.01
-43.01
10.2 Return Loss / Reflection Coefficient / VSWR Conversions / Fwd-Rev Power
Return Loss
VSWR
Reflection Coefficient
(dB)
127
Thru Power
Reflected Power
(%)
(%)
0.0
INF
1.000
0.00
100.00
1.0
17.391
0.891
20.57
79.43
2.0
8.724
0.794
36.90
63.10
3.0
5.848
0.708
49.88
50.12
4.0
4.419
0.631
60.19
39.81
5.0
3.570
0.562
68.38
31.62
6.0
3.010
0.501
74.88
25.12
7.0
2.615
0.447
80.05
19.95
8.0
2.323
0.398
84.15
15.85
9.0
2.100
0.355
87.41
12.59
10.0
1.925
0.316
90.00
10.00
11.0
1.785
0.282
92.06
7.94
12.0
1.671
0.251
93.69
6.31
13.0
1.577
0.224
94.99
5.01
14.0
1.499
0.200
96.02
3.98
15.0
1.433
0.178
96.84
3.16
16.0
1.377
0.158
97.49
2.51
Return Loss
VSWR
Reflection Coefficient
(dB)
Thru Power
Reflected Power
(%)
(%)
17.0
1.329
0.141
98.00
2.00
18.0
1.288
0.126
98.42
1.58
19.0
1.253
0.112
98.74
1.26
20.0
1.222
0.100
99.00
1.00
21.0
1.196
0.089
99.21
0.79
22.0
1.173
0.079
99.37
0.63
23.0
1.152
0.071
99.50
0.50
24.0
1.135
0.063
99.60
0.40
25.0
1.119
0.056
99.68
0.32
26.0
1.106
0.050
99.75
0.25
27.0
1.094
0.045
99.80
0.20
28.0
1.083
0.040
99.84
0.16
29.0
1.074
0.035
99.87
0.13
30.00
1.065
0.032
99.90
0.10
31.00
1.058
0.028
99.92
0.08
32.00
1.052
0.025
99.94
0.06
33.00
1.046
0.022
99.95
0.05
34.00
1.041
0.020
99.96
0.04
35.00
1.036
0.018
99.97
0.03
36.00
1.032
0.016
99.97
0.03
37.00
1.029
0.014
99.98
0.02
38.00
1.025
0.013
99.98
0.02
39.00
1.023
0.011
99.99
0.01
40.00
1.020
0.010
99.99
0.01
Chapter 10: Reference Tables
128
10.3 Wireless and Radar/Microwave Bands
Frequency
Band
Waveguide
3 - 30 MHz
HF
N/A
30 - 300 MHz
VHF
N/A
300 - 1000 MHz
UHF
WR-2300, WR-2100, WR-1500, WR-1150
1 - 2 GHz
L
WR-1000, WR-770, WR-650, WR-430
2 - 4 GHz
S
WR-430, WR-340, WR-284, WR-229
4 - 8 GHz
C
WR-229, WR-187, WR-159, WR-137
8 - 12 GHz
X
WR-112, WR-90
12 - 18 GHz
Ku
WR-62
18 - 26.5 GHz
K
WR-51, WR-42
26.5 - 40 GHz
Ka
WR-28
30 - 50 GHz
Q
WR-22
40 - 60 GHz
U
WR-19
50 - 75 GHz
V
WR-15
60 - 90 GHz
E
WR-12
75 - 110 GHz
W
WR-10
90 - 140 GHz
F
WR-8
110 - 170 GHz
D
WR-6
The “WR” number is the inside width of the waveguide in hundredths of an inch (mils / 10). The waveguide’s
inside height is typically half this dimension.
Example: WR-15 waveguide inside measurement is 0.150W x 0.075H
10.4 Sensor Cable Length Effects
When wide bandwidth peak power sensors are used with long sensor cables, the bandwidth and risetime is impacted due to cable loss at high frequencies. Ordinarily, the cable
rolls off the highest frequencies quite severely as its length is increased. The input circuit of
the power meter may be compensated for longer cables to reduce this effect.
This compensation is optional for certain Boonton peak power meter models, and is strongly recommended if extended-length cables are used. Using a standard length cable with
a compensated channel will result in significant overshoot and increased peak-to-average
display.
129
To calculate the new risetime specification for a sensor, input board and cable combination;
the square root of the sum of the squares of the cable and sensor are used, as shpwn in the
formula below.
Risetime = √ (Cable Risetime2 + Sensor Risetime2)
The following table shows the cable risetime effect for various standard cable lengths. Use
the equation above to compute composite risetime with a particular sensor. Note the table
includes columns showing the effect of using extended-length cables on uncompensated
(standard configuration) power meter input channels, and length-compensated (special
order) power meter inputs.
For example, if a 59318 sensor is used with a 10 ft cable on a Model 4500B equipped with
a 10 ft compensated input channel, the resulting risetime is computed as follows:
59318 risetime: 10ns
Cable Length
Cable Risetime
Cable Risetime
(uncompensated)
(compensated input)
5 ft (1.5m)
(standard cable: no effect)
(standard cable: no effect)
10 ft (3.0m)
55 ns
15 ns
20 ft (6.1m)
140 ns
40 ns
25 ft (7.6m)
180 ns
50 ns
50 ft (15.2m)
400 ns
75 ns
10 ft cable risetime: 15ns
Total risetime = √ (102 + 152)
= √ 325
= 18 ns
Chapter 10: Reference Tables
130
Chapter 11: Boonton Solutions
Boonton Electronics has been a leader in RF power measurement for more than 30 years.
This chapter contains a guide to popular Boonton Electronics power measurement solutions.
11.1 4240 Series RF Power Meter
The 4240 Series of CW RF power meters provides the high speed measurement capability needed in a
production environment, as well as the simplicity of operation required for bench top use. It provides
very accurate measurements from -70 dBm to +44 dBm (sensor dependent) and has a rapid display
update rate for tuning applications. The easy to read LCD displays both channels simultaneously with
numeric and bar graph information. The 4240 Series has a 5 digit resolution and can display the value
in either logarithmic or linear units. The 4242 two channel model allows the simultaneous comparison of multiple inputs during testing and in difference and ratio measurements. The 4240 Series is
compatible with all Boonton CW diode, thermocouple, and waveguide sensors from 10 kHz to 40 GHz.
Standard IEEE-488 GPIB and RS232 ports allow convenient interface with an ATE system. The SCPI
command set, or an available LABVIEW driver allow simple integration with an existing ATE system.
Quick Features:
• -70 dBm to +44 dBm, depending on sensor
• Automatically loads sensor data
• 90 dB dynamic range, depending on sensor
• Simple software control via SCPI language
• 10 kHz to 40 GHz measurement range
• 50 MHz step calibrator
• Single or dual-channel display
• IEEE-488 and RS-232 interfaces standard
• >200 measurements per second
• HP437, HP438, and Boonton 4220A/4230A emulation
131
11.2 4530 Series RF Power Meter
The 4530 Series RF Peak Power Meter can make Peak, CW Power and RF Voltage measurements at high
speed from 10 Hz to 40 GHz (sensor dependent). Boonton’s 4530 Series RF power meters combine the
accuracy of a laboratory-grade instrument with the speed required for production test. They employ
proprietary measurement techniques that accurately measure digitally-modulated signals. Whether
you’re measuring CW power or the peak power of WCDMA or HDTV signals, Boonton’s single-channel
Model 4531 and dual-channel Model 4532 are the logical choice for high volume production test. The
4530 provides seamless CW power measurement over its broad dynamic range without the interruptions and nonlinearities caused by range changes required by lesser power meters. Our thermal and
peak-power sensors never need range switching and even our CW diode sensors with 90 dB dynamic
range use only two widely overlapping ranges. The 4530 displays periodic and pulse waveforms in
graphical format, and a host of automatic measurements characterize the time and power profiles of
the pulse. As with all measurement modes, the graph display includes complete pan and zoom ability,
and can present the data in CDF, CCDF or distribution (histogram bar) formats.
Quick Features:
• Frequency Range: 10 kHz to 40 GHz
• Dynamic Range: Peak Power >60 dB and CW Power 90 dB
• Synchronous/Asynchronous Triggering
• Effective sampling rates up to 50 MSamples/sec
• Dual-channel statistical measurements (CDF/PDF)
• Modulation bandwidth to 20 MHz
• GPIB and RS232 standard interface with SCPI / RS232 commands
• LABVIEW Drivers available
Chapter 11: Boonton Solutions
132
11.3 4540 Series RF Power Meter
The Boonton model 4540 Series RF Power Meter is the instrument of choice for capturing, displaying and analyzing RF power in both the time and statistical domains. Applications include pulsed RF
signals such as radar, TDMA and GSM, pseudorandom or noise-like signals such as CDMA, WLAN and
WiMAX. The 4540 Series is a single or dual channel RF Power Meter that can measure modulated or CW
signals using peak and average Boonton Power sensors. The 4540 Series offers Pulse, Modulated/CW,
and Statistical operating modes, making it well suited for all requirements of R&D, manufacturing and
control operations. Single channel versions (4541) and dual channel versions (4542) are available.
The 4540 Series RF Power Meter offers an impressive detailed representation of measured signals.
This instrument is equipped with an “Auto set” feature. This feature analyzes incoming signals and
presets the instrument’s timing and trigger settings in a way that allows for immediate measurements.
Quick Features:
• Three operating modes: Pulse, Statistical, and Modulated/CW
• High Bandwidth Wide Dynamic Range Sensors
• Intuitive User Interface
• 4” color LCD display
• 200 ps time resolution
• 7 ns rise time
• Video bandwidth up to 70 MHz
• 17 default presets plus storage for 25 user defined presets
• Statistical analysis including CCDF
• Text view of up to 14 out of 28 parameters per channel simultaneously
(power / voltage, time, statistics, channel math)
• Effective sampling rate of up to 5 GSamples/second for repetitive signals
• GPIB, USB and LAN interfaces
133
11.4 4500B RF Peak Power Analyzer
The 4500B Peak Power Analyzer has brought the performance of peak power analyzers to a new peak
and is changing the way the industry views and analyzes RF data. The Boonton Model 4500B is the
instrument of choice for capturing, displaying, analyzing and characterizing RF power in both the time
and statistical domains. Applications include pulsed RF such as RADAR, TDMA and GSM, pseudorandom or noise-like signals such as CDMA and WLAN and modulated time slotted signals such as GSMEDGE and TD-SCDMA. Peak power sensors are available which feature <7 nsec rise time (typical video
bandwidth up to 65 MHz) and dynamic range of 70 dB (pulse mode) or 80 dB (modulated mode).
These sensors have been optimized for use with the 4500B and are ideal for measuring RADAR, 3G
and future 4G wireless systems which use complex modulation such as OFDM. The 4500B features
optional probability density functions (PDF) and cumulative distribution functions (CDF, CCDF) to accurately characterize noise-like RF such as CDMA, HDTV and WLAN. These statistical functions build and
analyze a very large population of power samples continuously at a rate of up to 25 MHz or triggered
up to 50 MHz on two channels simultaneously. These functions are fast, accurate and allow the measurement of very infrequent power peaks for a user-defined population size or acquisition interval.
Quick Features:
• 8.4” TFT color LCD display
• Displays up to 4 measurement channels, 2 memory channels and 1 math channel simultaneously
• 100 ps time base resolution, 10 GSa/Sec effective sample rate
• Video bandwidth greater than 70 MHz, typical risetimes to 5ns (sensor dependent)
• Automatic peak-to-peak, delay-by-time and delay-by-events triggering
• Flexible triggering and greater than 80 dB dynamic range (sensor dependent)
• GPIB, USB and LAN
• Text view of 15 time and power measurements per channel
• Envelope, persistence and roll mode displays
• Gated CCDF and PDF with log display (optional) at acquisition rates up to 50 MSa/s
• Continuous statistical analysis of power (optional) at acquisition rates up to 25 MSa/s
• Familiar user interface
• Peak Power Sensors available with high video bandwidth, fast rise time, and wide dynamic range
Chapter 11: Boonton Solutions
134
11.5 Boonton CW and Peak Power Sensors
CW Sensor table:
Model
Frequency Range
Dynamic Range1
Impedance Connector
Overload Rating
Maximum SWR
Pulse / Continuous
Frequency
SWR @ 0 dBm
Wide Dynamic Range Dual Diode Sensors
51075A
50 ohm
500 kHz to
-70 to +20 dBm
18 GHz
1 W for 1 µs
500 kHz to 2 GHz
1.15
300 mW
2 GHz to 6 GHz
1.20
6 GHz to 18 GHz
1.40
10 W for 1 µs
500 kHz to 2 GHz
1.15
3W
2 GHz to 6 GHz
1.20
N (M)
51077A
50 ohm
500 kHz to
-60 to +30 dBm
18 GHz
6 kHz to 18 GHz
1.40
100 W for 1 µs
500 kHz to 2 GHz
1.15
25 W
2 GHz to 6 GHz
1.20
6 GHz to 18 GHz
1.40
1 W for 1 µs
10 MHz to 2 GHz
1.15
2 GHz to 4 GHz
1.20
4 GHz to 18 GHz
1.45
N (M)
51079A
50 ohm
500 kHz to
-50 to +40 dBm
18 GHz
N (M)
51071A
10 MHz to
50 ohm
26.5 GHz
-70 to +20 dBm
300 mW
K (M)
51072A
50 ohm
30 MHz to
-70 to +20 dBm
40 GHz
1 W for 1 µs
300 mW
18 GHz to 26.5 GHz
1.50
30 MHz to 4 GHz
1.25
4 GHz to 38 GHz
1.65
38 GHz to 40 GHz
2.00
15 W for 1 µs
10 MHz to 30 MHz
1.25
300 mW
30 MHz to 16 GHz
1.18
16 GHz to 18 GHz
1.28
150 W for 1 µs
10 MHz to 2 GHz
1.10
10 W
2 GHz to 12.4 GHz
1.18
12.4 kHz to 18 GHz
1.28
K (M)
Thermoscouple Sensors
51100(9E)
50 ohm
10 MHz to
-20 to +20 dBm
18 GHz
N (M)
51200
10 MHz to
50 ohm
18 GHz
N (M)
135
0 to +37 dBm
Special Purpose Dual Diode Sensors
51011 (EMC)
10 kHz to 8 GHz
50 ohm
-60 to
1 W for 1 µs
10 kHz to 2 GHz
1.12
+20 dBm
200 mW
2 GHz to 4 GHz
1.20
4 GHz to 8 GHz
1.40
N (M)
51011 (4B)
100 kHz to 12.4 GHz
50 ohm
-60 to
1 W for 1 µs
100 kHz to 2 GHz
1.12
+20 dBm
300 mW
2 GHz to 4 GHz
1.20
4 GHz to 11 GHz
1.40
N (M)
51013 (4E)
100 kHz to 18 GHz
50 ohm
-60 to
1 W for 1 µs
+20 dBm
300 mW
N (M)
51015 (5E)
100 kHz to 18 GHz
50 ohm
11 GHz to 12.4 GHz
1.60
100 kHz to 4 GHz
1.30
4 GHz to 10 GHz
1.50
10 GHz to 18 GHz
1.70
-50 to
10 W for 1 µs
100 kHz to 1 GHz
1.07
+30 dBm
2W
1 GHz to 2 GHz
1.10
2 GHz to 4 GHz
1.12
4 GHz to 12.4 GHz
1.18
12.4 GHz to 18 GHz
1.28
N (M)
51033 (6E)
100 kHz to 18 GHz
50 ohm
-40 to
10 W for 1 µs
100 kHz to 1 GHz
1.07
+33 dBm
2W
1 GHz to 2 GHz
1.10
2 GHz to 4 GHz
1.12
4 GHz to 12.4 GHz
1.18
N (M)
51078
100 kHz to 18 GHz
50 ohm
-20 to
100 W for 1 µs
+37 dBm
7W
N (M)
12.4 GHz to 18 GHz
1.28
100 kHz to 4 GHz
1.15
4 GHz to 12 GHz
1.25
12 GHz to 18 GHz
1.40
Diode Average Sensor (for use with 4530, 5230, 4230, 4240, 4540 **)
51085
50 ohm
N (M)
500 kHz to 18 GHz
-30 to
100 W for 1 µs
500 kHz to 4 GHz
1.15
+20 dBm
5 W (*)
4 GHz to 12.4 GHz
1.20
12.4 GHz to 18 GHz
1.25
1 Models 4731, 4732, 4231A, 4232A, 4300, 4531, 4532, 5231, 5232, 5731, 5732
(*) For 51085 Peak Power - 1kW peak, 5μs pulse width, 0.25% duty cycle. For 51085 CW Power - 5W (+37dBm)
average to 25°C ambient temperature, derated linearly to 2W (+33dBm) at 85°C.
(**) Note: 4540 frequency starts at 1 MHZ. For other meters please contact Boonton.
Chapter 11: Boonton Solutions
136
Peak Sensor Table
Model
Dynamic Rating
Overload Rating
Sensor Response
Maximum SWR
I m p e d a n c e (Low Bandwidth)
Frequency Range
Peak Power Range
Pulse / Continuous
Fast Risetime
Slow Risetime
Frequency
Connector
CW Power Range
(Bandwidth)
(Bandwidth)
SWR @ 0 dBm
Int. Trigger Range
For use with models 4500B and 4540
57006
0.5 to 6 GHz
-50 to +20 dBm
1 W for 1 µs
< 7 ns
<10 µs
50 ohm
(0.05 to 6 GHz)
-60 to +20 dBm
200 mW
(70 MHz
(350 kHz)
59318
0.5 to 18 GHz
-24 to +20 dBm
1 W for 1 µs
< 10 ns
50 ohm
(0.05 to 18 GHz)
-34 to +20 dBm
200 mW
(50 MHz
-40 to +20 dBm
N (M)
1.25
<10 µs
0.05 to 2 GHz
1.15
(350 kHz)
2 to 16 GHz
1.28
16 to 18 GHz
1.34
typical)
-10 to +20 dBm
N (M)
0.05 to 6 GHz
typical)
59340
0.5 to 40 GHz
-24 to +20 dBm
1 W for 1 µs
< 10 ns
<10 µs
0.05 to 4 GHz
1.25
50 ohm
(0.05 to 40 GHz)
-34 to +20 dBm
200 mW
(50 MHz
(350 kHz)
4 to 38 GHz
1.65
38 to 40 GHz
2.00
-10 to +20 dBm
K (M)
typical)
For use with models 4400, 4500, 4400A and 4500A Analyzers. Model 4530 with 1 GHz Calibrator Model 2530
-24 to +20 dBm
1 W for 1 µs
< 15 ns
<200 ns
0.05 to 2 GHz
1.15
50 ohm
-34 to +20 dBm
200 mW
(35 MHz
(1.75 MHz)
2 to 16 GHz
1.28
N (M)
-10 to +20 dBm
16 - 18 GHz
1.34
56318
0.5 to 18 GHz
typical)
-24 to +20 dBm
1 W for 1 µs
< 15 ns
<200 ns
0.05 to 2 GHz
1.15
50 ohm
-34 to +20 dBm
200 mW
(35 MHz)
(1.75 MHz)
2 to 4 GHz
1.20
K (M)
-10 to +20 dBm
4 to 18 GHz
1.45
18 to 26.5 GHz
1.50
56326
0.5 to 26.5 GHz
-40 to +20 dBm
1 W for 1 µs
< 100 ns
<300 ns
0.05 to 2 GHz
1.15
50 ohm
-50 to +20 dBm
200 mW
(6 MHz)
(1.16 MHz)
2 to 6 GHz
1.20
N (M)
-27 to +20 dBm
6 to 16 GHz
1.28
16 to 18 GHz
1.34
56518
0.5 to 18 GHz
For use with models 4400, 4500, 4400A and 4500A Analyzers. Model 4530
57518
0.1 to 18 GHz
-40 to +20 dBm
1 W for 1 µs
< 100 ns
< 10 µs
0.05 - 2 GHz
1.15
50 ohm
(0.05 to 18 GHz)
-50 to +20 dBm
200 mW
(6 MHz)
(350 kHz)
2 - 16 GHz
1.28
16 - 18 GHz
1.34
-27 to +20 dBm
N (M)
57540
0.1 to 40 GHz
-40 to +20 dBm
1 W for 1 µs
< 100 ns
< 10 µs
0.05 to 4 GHz
1.25
50 ohm
(0.05 to 40 GHz)
-50 to +20 dBm
200 mW
(6 MHz)
(350 kHz)
4 to 38 GHz
1.65
38 to 40 GHz
2.00
-27 to +20 dBm
K (M)
For use with models 4500, 4400 and 4530
-24 to +20 dBm
1 W for 1 µs
< 150 ns
< 500 ns
0.03 to 2 GHz
1.15
50 ohm
-34 to +20 dBm
200 mW
(3 MHz)
(700 kHz)
2 to 6 GHz
1.20
N (M)
-10 to +20 dBm
6 to 18 GHz
1.25
56218
137
30 MHz to 18 GHz
For use with models 4500 and 4400
56526
500 MHz to
-40 to +20 dBm
1 W for 1 µs
< 100 ns
< 300 ns
0.03 to 2 GHz
1.15
50 ohm
26.5 GHz
-50 to +20 dBm
200 mW
(6 MHz)
(1.16 MHz)
2 to 4 GHz
1.20
4 to 18 GHz
1.45
18 to 26.5 GHz
1.50
K (M)
-27 to +20 dBm
11.6 Most Popular Peak Sensors
Boonton has several popular peak power sensors and the following three are the most
popular & widely used:
59340 Peak Power Sensor:
Boonton offers a very fast 59340 Peak Power Sensor for frequencies up to 40 GHz. Wireless technology is becoming increasingly demanding for test and measurement equipment. Highly dynamic signals,
often with noise-like signal characteristics, fast switching carriers, and pulsed signals need to be analyzed accurately. Developers and engineers want reliable measurement tools to analyze such intricate
waveforms. The Boonton 59340 is the latest and fastest Peak Power Sensor operating up to 40 GHz.
Highlighted features:
• Optimized for Boonton’s high end Peak Power Meter series 4500B and 4540.
• Dynamic input power ranges from -24 dBm to +20 dBm in Peak mode and
-34 dBm to +20 dBm in CW mode
• Frequencies up to 40 GHz
• Rise time of less than 10 ns
• Bandwidth up to 50 MHz
Chapter 11: Boonton Solutions
138
57006 Peak Power Sensor:
Boonton offers another popular and very fast Peak Power Sensor, the 57006, for frequencies up to 6
GHz. This peak power sensor is optimized for Boonton Power Meter series 4500B and 4540. It provides dynamic input power ranges from -50 dBm to +20 dBm in Peak mode and -60 dBm to +20 dBm
in CW mode. Measuring RF carrier frequencies up to 6 GHz, it has a fast 7 ns risetime response and a
video bandwidth of 70 MHz. This sensor is also ideal for statistical measurement with high dynamic
signal content due to its high speed. Frequency range is dependent upon which bandwidth the sensor
is set to. It ranges from 0.5 GHz to 6 GHz in high bandwidth mode and from 50 MHz to 6 GHz in low
bandwidth mode.
Highlighted features:
• Frequencies up to 6 GHz
• Low Freq Radar Application
• Communications Applications
• Bandwidth up to 65 MHz
• High dynamic range (-60 to +20 dBm)
• 7 ns rise time (5 ns typical)
• Work with Boonton 4500B and 4540 Series
59318 Peak Power Sensor:
59318 Peak Power Sensor is for frequencies up to 18 GHz. This peak power sensor is also optimized for
Boonton Power Meter series 4500B and 4540. It provides dynamic input power ranges from -24 dBm
to +20 dBm in Peak mode and -34 dBm to +20 dBm in CW mode. With frequencies of up to 18 GHz,
it has a rise time of less than 10 ns and a bandwidth of up to 50 MHz. Frequency range is dependent
upon which bandwidth the sensor is set to. Usually it ranges from 0.5 GHz to 18 GHz in high bandwidth mode and from 50 MHz to 18 GHz in low bandwidth mode.
Highlighted features:
• Frequencies up to18 GHz
• High Frequency Communication Radar and Amplifier Testing
• 10 ns rise time (8 ns typical)
• 44 dB dynamic range in Peak Mode
• Bandwidth up to 50 MHz
• Work with Boonton 4500B and 4540 Series
139
25 Eastmans Rd
Parsippany, NJ
United States
Tel:
+1 973 386 9696
Fax:
+1 973 386 9191
www.boonton.com
Follow us on:
WTGinnovation
Wireless Telecom Group
blog.wtcom.com
WTGinnovation
B/REFGUIDE/1111/EN
Note: Specifications, terms and conditions are
subject to change without prior notice.
© Copyright 2011
141
All rights reserved.
Scan this Code to
Download a Digital Copy
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement