null  null
APRIL lasa
Ã- Ã- Ã- Ã-
BB'-i 106
^B 5
^ â € ” r '
© Copr. 1949-1998 Hewlett-Packard Co.
Technical Information from the Laboratories of Hewlett-Packard Company
APRIL 1980 Volume 31 • Number 4
Microwave CW and Pulse Frequency Measurements to 40 GHz, by Richard F. Schneider,
Ronald systems Felsenstein, and Robert W. Offermann As radar and communications systems
reach sophisticated. higher frequencies, test equipment becomes more sophisticated.
A 400-to-1600-MHz +8 Prescaler, by Hans J. Jekat State-of-the-art technology went into
its tiny amplifier, attenuator, and binary circuits.
An Automatic Microwave Frequency Counter Test System to 40 GHz, by Larry L.
Koepke system high-performance microwave counters isn't a trivial task, but this system
does it automatically.
40-GHz Frequency Converter Heads, by Mohamed M. Sayed The heads down-convert
microwave input signals to frequencies that are more easily transmitted over coaxial cables.
A 26.5-GHz Automatic Frequency Counter with Enhanced Dynamic Range, by AH
Bologlu Here's a cost-effective counter with sensitivity in the microwatt range.
Microwave Counter Applications, by Richard F. Schneider Radar, oscillator, and general
high-frequency measurements are described.
Laboratory Notebook — A Flexible Software Development Technique, Ronald E.
Felsenstein If you have read-only memory to spare, you can use it to make changes in
long-lead-time masked ROMs that you've already ordered.
In this Issue:
40 GHz stands for 40 gigahertz, which is the scientific way to say 40,000,000,000 cycles
per second. That's a high frequency, 500 times higher than VHP television broadcast
frequencies, but still 10,000 times lower than visible light. Equipment that operates in the
4-to-40-GHz range includes police speed detection radars, military and commercial air
craft radars, line-of-sight communications systems, and satellite up and down links. Even
higher that are in use, too. There's equipment in current production that operates
up to 60 GHz, developmental systems that go up to 110 GHz or so, and experimental de
vices of the 2-300 GHz range. Two atmospheric windows at 34 and 94 GHz are of particular
interest does at are frequencies at which the earth's atmosphere appears much more transparent than it does at
other frequencies.
This month's issue and cover photo are devoted to some new high-performance microwave counter products
that help 14 this high-frequency equipment. The system described in the articles on pages 3 and 14 measures
frequencies as high as 40 GHz, whether the microwave energy is continuous or in the form of short bursts or
pulses. The counter described in the article on page 20 measures the frequencies of continuous microwave
signals up to 26.5 GHz.
Here's when question for designers of microprocessor-based equipment. What do you do when you find you have
more functions memory than you need to microprogram all the functions your equipment is supposed to have?
Why, you little more functions, of course. Many a product has been upgraded in capability for very little added
cost by spare extra ROM space. But on page 25 you'll find another suggestion for using that spare ROM. If you
save masked earlier corrections to the basic ROM, you'll be able to order masked chips much earlier for all of your ROM
except until final. that contains the corrections. That one can be an erasable ROM until all of the microcode is final.
It's a simple way to get into production faster.
-R. P. Do/an
Editor, Richard P. Dolan • Contributing Editor, Howard L Roberts • Art Director, Photographer, Arvid A. Danielson
Illustrator, Nancy S. Vanderbloom • Administrative Services, Typography, Anne S. LoPresti • European Production Manager, Dick Leeksma
H E W L E T T - P A C K A R D
1 9 8 0 .
 © H e w l e t t - P a c k a r d
© Copr. 1949-1998 Hewlett-Packard Co.
C o m p a n y
1 9 8 0
P r i n t e d
i n
Microwave CW and Pulse Frequency
Measurements to 40 GHz
A new harmonic heterodyne frequency converter plug-in
add s a u t o m a tic 40-GHz f requenc y m e a su re me n ts to th e
universal capabilities of HP's top counter.
by Richard F. Schneider, Ronald E. Felsenstein, and
Robert W. Offermann
TO BE USEFUL in the widest possible range of appli
cations, a microwave counter should be capable of
measuring the carrier frequencies of pulsed or CW
signals, and for pulsed signals, should also provide for
time interval or frequency measurement of the pulse modu
lation. The design should be optimized with wide IF bandwidths for narrow pulses and wide FM deviations, and
should have high sensitivity. Cost-effective frequency
range selection and automatic operation are essential.
A new system that operates as a plug-in to the HP 5345A
500-MHz Universal Counter1 is designed to meet these re
quirements over a carrier frequency range of 0.4 to 40 GHz.
The system, which consists of the 5355A Automatic Fre
quency Converter and the 5356A/B/C Frequency Converter
Heads, is shown in Fig. 1. It provides an effective alternative
to the complex specially assembled systems that formerly
were the only way to measure up to 40 GHz.
The 5356A/B/C Frequency Converter Heads eliminate the
need for microwave transmission lines to connect the mea
sured source to the counter. Coaxial cables, while conve
nient, cannot always be used, since two-metre coaxial lines
typically have about 10 dB loss at 18 GHz and get worse at
higher frequencies. To circumvent this, hybrid microwave
circuits in the heads down-convert incoming frequencies to
intermediate frequencies (IF) that are easily transmitted
over a 1.7-metre miniature coaxial cable to the 5355A Au
tomatic Frequency Converter Plug-in. This eliminates the
transmission line loss and effectively improves system sen
sitivity by that amount. Heads with various coaxial or
waveguide connectors can be selected to meet the mea
surement requirement (see article, page 14).
Microprocessor control in the plug-in makes operation
automatic in either pulse or CW mode. The new system uses
the single-sampler harmonic heterodyne technique.2 The
microprocessor computes the input frequency according to
the desired resolution set on the 5345A Counter front panel
Fig. 1. Model 5355/4 Automatic
Frequency Converter plug-in for
Model 5345A Counter measures
the frequencies of CW or pulsed
signals up to 40 GHz. Downconversion of the input frequency
takes place in the interchangeable
5356A/B/C Frequency Converter
Heads, eliminating the need for
high-frequency transmission lines
between the source and the
counter. External gating makes it
possible to measure the frequency
profile within a pulse.
© Copr. 1949-1998 Hewlett-Packard Co.
5355A Automatic
External Gate
5356A/B/C Head
and Driver
Fig. 2. Simplified block diagram
of the harmonic heterodyne fre
quency conversion technique for
CW and pulsed signals.
*AAD = Automatic Amplitude Discrimination
and displays the frequency on the counter's eleven-digit
The 5355A plug-in has a simplified keyboard that allows
the user to select automatic or manual CW or pulse opera
tion, to specify frequency offsets or multiplication of the
measured frequency by a constant, to display frequency
deviation, and to select the prescaler built into the plug-in.
The prescaler divides the input frequency by eight. It is
used to measure frequencies from 0.4 to 1.6 GHz; no fre
quency converter head is needed in this range. The pre
scaler has its own fused front-panel connector.
Pulse repetition frequency measurements and time inter
val measurements such as pulse width, pulse repetition
interval, pulse repetition period, and pulse-to-pulse spac
ing are made by the reciprocal-taking 5345A Counter, using
the detected IF from the 5355A Converter plug-in. The
counter mainframe also measures frequencies from 50 /uHz
to 500 MHz. The counter has a maximum time interval
resolution of two nanoseconds for single-shot intervals and
two picoseconds for time interval average measurements.
The complete microwave counter system consisting of
the 5 34 5 A Counter with the 5355A Frequency Converter
and the 5356A/B/C Heads measures any frequency from 50
fj.Hz to 40 GHz. Its sensitivity with the 5356A/B Heads is
-20 dBm from 1.5 to 12.4 GHz and -15 dBm from 12.4 to
26.5 GHz. With the 5356C Head, sensitivity is 5 dB better up
to 26.5 GHz and decreases to -10 dBm at 40 GHz. Prescaler
sensitivity is -15 dBm from 400 MHz to 1.6 GHz.
In the automatic mode, the system measures the frequen
cies of RF pulses from 100 ns to 20 ms wide at pulse repeti
tion frequencies of 50 Hz to 2 MHz. In manual mode, pulses
as narrow as 60 ns can be measured, and external gates as
narrow as 20 ns may be applied to the counter for applica
tions such as measuring the frequency profile within a pulse.
For pulsed RF-signals, the FM tolerance is 50 MHz peakto-peak for a 100-ns pulse in the automatic mode, and 80
MHz p-p for a 60-ns pulse in the manual mode. Automatic
calibration of the 5345A mainframe assures accuracy to 3
kHz in pulsed carrier frequency measurements. Resolution
is selectable to as fine as 1 00 Hz by frequency averaging. For
example, a 26.5-GHz pulse radar with a l-/xs wide pulse
could be measured with a 10-ms gate time to a resolution of
10.3 kHz and an accuracy of 43 kHz or about 2 parts in 106
(assuming no time-base error).
For CW (continuous) signals, the maximum resolution is
0.1 Hz up to 10 GHz and 1 Hz from 10 to 40 GHz. The FM
tolerance is 1 5 MHz p-p in the normal mode and 60 MHz p-p
in the special FM mode.
Harmonic Heterodyne System
The harmonic heterodyne technique has been described
in previous articles.2 Basically, a microwave sampler is
driven at programmed synthesized frequencies, as shown
in Fig. 2, until a signal occurs in the passband of the IF
amplifier, indicating that some harmonic of the sampling
frequency is mixing with the incoming microwave signal to
produce a countable IF. The microprocessor then executes
an algorithm to identify the harmonic number N and the
sign of the IF (sum or difference), and solves for the input
frequency, according to the equation:
fx = Nfs ± IF
where fx is the input frequency, N is the harmonic number,
and fs is the programmed synthesized frequency.
The harmonic number is determined by changing the
synthesized frequency slightly and measuring the change
in the IF frequency.
N =
firi -
where fIF1 = IF when fs = fj
fiF3 = IF when fs = f3
The sign of the IF in equation 1 is determined by whether
ftps is larger or smaller than frp!.
Pulse Mode
The basic design parameters of the system were derived
from the pulse requirements and the mainframe counter's
capabilities. Linear programming3 was used to optimize the
system. Seven equations in five variables were solved sub
ject to various boundary conditions, including the
minimum input frequency, the IF bandwidth, the IF guard
band, the maximum harmonic number, and the minimum
synthesizer frequency. The linear programming equations
were entered into the computer and families of solutions
were obtained for the five variables. Tradeoffs were then
made to minimize the tuning range of the synthesizer oscil
lator and optimize the IF bandwidth. Finally, a separate
computer program was derived to determine the minimum
number of frequencies required to obtain complete fre
quency coverage. The result is a set of frequency tables , one
for each frequency converter head. For example, with the
18-GHzModel 5356Ahead, only 13 synthesizer frequencies
are required.
© Copr. 1949-1998 Hewlett-Packard Co.
In the search routine the synthesizer is stepped instead of
swept. The synthesizer frequency tables are stored in a
ROM. and the synthesizer is stepped to the next frequency
in the table after waiting the longest specified pulse repeti
tion interval of 20 ms. This is repeated until a signal appears
in the IF passband of 157 to 330 MHz. Next the synthesizer
frequency is digitally incremented 4 MHz and the IF
passband is tested. If the incremented-synthesizer IF falls
outside the passband. the search routine proceeds to the
next frequency in the table. If the IPs for both synthesizer
settings are within the passband, the calculation of N and
the sign of the IF can proceed.
After the initial acquisition in the IF passband of 157 to
330 MHz, the IF can shift into the IF guard band without
affecting the measurement. The guard band extends down
to 78 MHz and up to 375 MHz. as shown in Fig. 3. If the
IF moves out of the guard band, the 5355A reacquires
the input and discards the results of any measurement
in progress.
Automatic gain control of the IF amplifier in the pulse
mode minimizes the required input signal on/off ratio and
maintains the signal-to-noise ratio. An IF detector and a
A 400-to-1600-MHz ^8 Prescaler
by Hans J. Jekat
Behind the 0.4-to-1.6-GHz input on the front panel of the 5355A
Automatic Frequency Converter is a prescaler that divides the input
frequency by eight to bring it within the range of the 5345A Counter
mainframe. The prescaler operates in both the CW and pulsed RF
modes. An arming circuit senses marginal signals to keep the
counter from miscounting. The prescaler input is protected by a fuse
that is accessible from the front panel.
Fig. 1 is a block diagram of the prescaler. The latest state-of-the-art
technology has been applied in the design of the attenuator,
amplifier, and binary circuits.
AGC Attenuator
The attenuator and the AGC (automatic gain control) circuit (see
Fig. 1) are used in both CW and pulsed RF modes. AGC is very
important in the pulsed RF mode, since unwanted signal during the
off condition of the pulse signal could cause the pulse detector to
delay its gate closing. Therefore, the AGC is set to a level where the
RF pulse can be counted, and the off portion of the pulse is then
a a
Fig. in Attenuator (I.), amplifier, and binary (r.) are housed in
TO-12 packages.
Input 400 MHz to 1600 MHz
Fig. 1. Block diagram of the 4001600 MHz +8 prescaler in the
5355/4 Automatic Frequency
© Copr. 1949-1998 Hewlett-Packard Co.
Ac Input
Fig. 3. 1.6-GHz binary (+2 cir
cuit) consists of two crosscoupled current-mode flip-flops in
a master-slave configuration.
compressed to a point below the input sensitivity, thereby preventing
noise counting and gate jitter. The attenuator contains four PIN
diodes connected in a n configuration. There are two PIN diodes in
the transmission line compared to the one that is commonly used in a
i7 attenuator. The advantage is that the off capacitance is only half as
great and therefore, the attenuation in the off condition is higher. The
tradeoff is a slightly higher input VSWR. To get good high-frequency
attenuation, the bypass capacitors are parallel-plated capacitors
with and extremely low profile of 0.13 mm. Parallel resonances and
inductive reactances are not discernible. The attenuator is packaged
in a TO-12 four-lead package (see Fig. 2).
2-to-1 600-MHz Amplifier
The amplifier used in the prescaler is constructed on a sapphire
substrate measuring 2.5 by 6.4 mm. Only two transistors and one
chip capacitor have to be mounted on the sapphire substrate. The
rest of the circuitry consists of thin-film resistors and thin-film induc
tors. The low parts count in the amplifier yields a very high reliability.
The amplifier has ±1 dB flatness and 24 dB gain, and is housed in a
TO-12 package (Fig. 2), which uses little space on the prescaler
printed circuit board.
1.6-GHz Binary
The 1.6-GHz binary (-=-2 circuit) is a monolithic high-frequency
divider circuit (see Fig. 3). Two current-mode flip-flops are crosscoupled in a master-slave configuration. Second-level current
switches control updating and latching of the flip-flops. Two input
bias control to the master flip-flop second-level current switch control
self-oscillation and bandwidth by prebiasing the data transistors.
This makes them switch faster, since they do not have to wait for the
total a swing of the slave flip-flop. This technique requires a
larger voltage swing for lower frequencies, and since the voltage
swing is limited, the low-frequency response is degraded. In other
words, we are trading off low-frequency toggle speed for highfrequency toggle speed.
By controlling the clock input bias, the binary can be pushed into
self-oscillation around 1 200 MHz. At this point, the binary is in its most
sensitive region over the entire bandwidth. The binary stops oscillat
ing and begins dividing when an RF pulse appears at the prescaler
input. In CW mode, the 1.6-GHz binary is biased to a point of no
self-oscillation. This sacrifices some sensitivity but assures that the
counter will not respond to noise. In pulsed RF mode, the binary can
be biased in the more sensitive oscillating mode since the gate signal
dictates which signal is counted.
The binary is packaged in a four-lead TO-12 package (see Fig. 2)
and the casting of the prescaler acts as a heat sink for it.
Hans J. Jekat
Hans Jekat is an electrical engineering
graduate of the Technische
Hochschule in Munich. He moved to the
U.S.A. in 1958, and since 1964 has
been with HP's Santa Clara Division. He
served as project leader for the 5300
Measuring System and designed the
5300A mainframe, the 5305A/B Count
ers, several of the MOS/LSI circuits
used in the 5300 system and other in
struments, and the prescaler and IF
amplifier for the 5355A Frequency Con
verter. His counterwork has resulted in
several patents. Hans is married, has
two sons, and lives in Redwood City,
California. For many years a trainer of show horses, he also enjoys
soccer and skiing.
© Copr. 1949-1998 Hewlett-Packard Co.
- Guard Band
Fig. 3. 5355A IF passband extends from 757.5 to 330 MHz
for the down-converted carrier frequency. The guard band
allows for worst-case FM on the carrier of 45 MHz peak. The
specification is 80 MHz peak-to-peak for pulsed signals.
Compute fs. N per fm
Search fs Table
so fiFi and fIF3
are within IFC
gate generator provide the 5345A Counter with an external
gate signal about 30 ns shorter than the pulse burst and
centered on the burst to eliminate measurement of the rise
and fall frequency transients.
The main gate circuits in the 5345A Counter have an
asymmetry that causes a small difference between the time
it takes for the gate to open and the time it takes for the gate
to close. This difference, typically about 300 picoseconds,
becomes significant when the gate is opened and closed
many times for a single measurement on narrow pulses. The
error is proportional to the intermediate frequency and in
versely proportional to the input pulse width.
To minimize the effect of this error, an automatic calibra
tion routine in the 5355A is used whenever pulse bursts
narrower than 100 /us are measured. The calibration routine
uses the synthesizer signal for a reference frequency by
switching this signal into the IF after dividing by four.
Using the gate derived from the input signal, this reference
signal is measured, and the ratio of the actual synthesizer
frequency to that measured is computed and used as a
calibration factor. To improve the accuracy of the calibra
tion factor, it is averaged for the first ten measurements after
the signal is acquired. The letter C is displayed in the left
most position of the 5345A display during these ten mea
surements to indicate that calibration is taking place.
A change in the input pulse width of more than 12%, or
loss and reacquisition of the input, will cause a new calibra
tion cycle to take place. A special operating mode disables
the calibration scheme for relative frequency measurements
where absolute accuracy is not required.
Pulse Algorithm
The flow diagram shown in Fig. 4 outlines the search,
acquisition, calibration, computation, and measurement
cycles of the pulse algorithm. The harmonic number de
termination is made with a lOO-jtxs minimum gate time. A
normal pulsed frequency measurement is made of f[F1 and
fIF3 for synthesizer frequencies fa and f3, respectively.
The harmonic number N is then computed as follows:
N =
4 MHz
Definition of Symbols: See Fig. 6, page 9
x Calibration Factor
Fig. 4. Search, acquisition, calibration, computation, and
measurement algorithm for pulse measurements.
© Copr. 1949-1998 Hewlett-Packard Co.
10 MHz From
Counter Time Base
Fig. 5. In CW mode, a pseudo
random sequence generator
modulates the VCO frequency. A
and B counters measure the two
VCO frequencies and the corres
ponding intermediate frequen
cies. This information makes it
possible to identify which har
monic of the VCO is mixing with the
input to produce the IF. The length
of the pseudorandom sequence
determines the allowable F M in the
CW mode.
A Channel
B Channel
Gate Generator
External Gate
N is continually measured during an automatic measure
For a 100-/J.S gate time, the time required for a pulse
frequency measurement is 2x100 ¿AS divided by the duty
factor of the pulse signal. Since the duty factor is the gate
width multiplied by the pulse repetition frequency, a l-/us
wide pulse with a 1-kHz repetition rate requires 200 ms to
CW Mode
Harmonic number determination in the CW mode uses a
pseudorandom sequence technique described previously4
to improve tolerance to FM. The counter's FM tolerance of
15 MHz p-p is related to the length of the sequence, which is
normally set for 22 ms. A special mode lengthens the se
quence to 360 ms for 60 MHz p-p FM tolerance.
The pseudorandom sequence is applied to the frequency
synthesizer at the VCO (voltage-controlled oscillator) in
put, as shown in Fig. 5, at a rate outside the phase-lockedloop bandwidth. The loop remains locked, but the
pseudorandom sequence modulates the VCO frequency
about 4 MHz peak-to-peak. The harmonic number and the
sign of the IF are determined by switching the counter
between the A and B channel inputs synchronously with
the 4 MHz modulation step. The pseudorandom sequence is
activated twice, once to measure the synthesizer frequency
change and once to measure the corresponding IF change.
In each case the frequency change is the difference between
the A and B counts. To determine N the IF change is divided
by the synthesizer change and the result is rounded to the
nearest integer (see next section). With this technique, only
one synthesizer is needed (many systems switch between
two synthesizers to determine N). The CW frequency reso
lution, a measure of the synthesizer noise spectral density,
is typically less than ±2 Hz at 18 GHz with a one-second
gate time.
Automatic amplitude discrimination is provided in the
CW mode by using a limiting IF amplifier and providing an
IF bandwidth that is greater than half the sampling rate.
This tends to make the average zero-crossing rate equal to
the frequency of the highest-level signal present. The
counter will measure this frequency, provided that this
signal is more than 8 dB greater than signals within 500
MHz and 20 dB greater than other signals within the fre
quency range of the 5356A and 5356B Heads.
CW Algorithm
The CW algorithm is shown in Fig. 6. Note that the syn
thesizer deviation is measured as well as the IF deviation.
This is required to determine the synthesizer deviation
accurately, since the modulation sensitivity of the VCO is
not perfectly linear. The harmonic number N is then com
puted as follows:
N =
The harmonic number is checked every ten gate times to
make sure it is correct.
The divide-by-M frequency synthesizer is phase-locked
to the 10-MHz time base of the 5345A mainframe. It oper
ates from 885.2 to 1056 MHz. A two-modulus divider, con
trolled by the microprocessor according to the frequency
table stored in the ROM, sets the frequency as shown in Fig.
7. Each converter head is coded so that when it is plugged
into the 5355A, the proper frequency table is accessed. The
VCO steps through the frequencies in a nonlinear manner.
The minimum change is 400 kHz.
The VCO output is amplified and sent to the 5356A/B/C
© Copr. 1949-1998 Hewlett-Packard Co.
fp =(fp8)- 8
Compute fs. N per fm
Search fs Table
so firi and f,F3
are within IF,
Set fs to f2
Measure Af2/4
and AfiF during
Pseudorandom Sequence
Head and to the phase-locked loop buffer amplifier. This
signal is then divided by four and applied to the micro
processor-controlled -^M two-modulus divider. The twomodulus divider permits the setting of frequencies other
than the normal integer values by switching between +40
and H-41 division factors. The phase detector has a reference
frequency input of 100 kHz derived from the 10-MHz
counter time base.
In the CW mode, the harmonic number is determined by
applying the pseudorandom sequence to the VCO input.
The pseudorandom rate is outside the loop bandwidth so
that the center frequency of the synthesizer is not perturbed.
A high-and-low-gain amplifier is used to optimize the sys
tem performance for both frequency measurement and N
determination. Low gain is used during harmonic number
determination so the maximum peak-to-peak deviation is
obtained, while high gain is used during the frequency
measurement to obtain the best spectral purity.
Elliptical filtering is used to minimize the 100-kHz
sidebands that are caused by noise on the 1 00-kHz reference
signal feeding through and modulating the VCO. Another
operational amplifier is used to condition a tuning signal
that the to a filter in the IF amplifier. This filter tracks the
synthesizer tuning and maintains the IF amplifier cutoff
frequency at about one-half the sampling frequency (syn
thesizer frequency) to minimize spurious responses.
Special precautions were taken to reduce the power-line
sidebands in the synthesizer spectrum. It was necessary to
wrap the elliptical filter inductor with a mu-metal shield
and place a sheet of transformer steel alongside the printed
circuit board casting to obtain the specified resolution.
IF Filter/Detector/Gate Generator
HP-IB Output
Display Result
CW Auto
Fig. 6. CW measurement algorithm.
Definition of Symbols for Figs. 4 and 6
Unknown prescaler frequency input
Synthesizer frequency
Unknown RF head frequency input
Down-converted IF (intermediate frequency)
Harmonic of f, that, mixed with f,, produces fIF
Keyboard-entered manual frequency
IF center passband 157.5 to 330 MHz
Value for f, from a synthesizer table
fi + 2 MHz
fi + 4 MHz
Down-converted IF when f, = ft
Down-converted IF when f, = 1j
The various IF bands are determined by three analog
filters followed by level detectors as shown in Fig. 8. The
edges of the acquisition band have individual filters and
detectors, while the guard band is determined by cascaded
high-pass and low-pass filters followed by a single detector.
All filters are five-pole Chebyshev type and all detectors
consist of a low-barrier Schottky diode and a capacitor.
The IF input is limited and has constant amplitude. The
detected level on each capacitor is compared to a reference
voltage by a high-speed voltage comparator, the output of
which gives a digital indication of the presence of a signal
in the passband of the associated filter. By designing the
filter so that the band edge frequency is several dB into the
filter's stopband, the exact cutoff frequency can be set by a
simple adjustment of the reference voltage. This allows
precise determination of the band edges without precision
trimming of the filters themselves. The outputs of the three
comparators are combined by logic that produces four data
lines from which the microprocessor can determine when a
CW or pulse signal is present in the acquisition or guard
The detection scheme is fast enough to detect the pres
ence of a valid IF signal from a single 60-ns wide burst. This
not only minimizes acquisition time in the pulse mode, but
also allows these same circuits to be used for generation of a
signal to gate the 5345A for pulse measurements. When a
pulsed-RF signal is present in the passband of one of the
filters, the output of the associated comparator approxiAPRIL 1980 HEWLETT-PACKARD JOURNAL 9
© Copr. 1949-1998 Hewlett-Packard Co.
Gain Control
100 kHz
Sequence Modulation
To 5356A/B/C
RF Head
Fig. 7. The 5355A's frequency
synthesizer generates the local
oscillator signal to down-convert
the input signal in the 5356A/B/C
Frequency Converter Head.
mates the modulation envelope of the input RF signal. By
using the comparator following the 78-to-375-MHz
bandpass filter, a gate will be obtained whenever there is a
countable signal present. However, it is not desirable to use
this detected envelope directly as a gate. This is because
pulse modulators commonly introduce a significant
amount of phase distortion in the process of turning the RF
signal on and off. Also, if the gate signal is the same width as
the burst to be counted, the timing of the gate relative to the
burst becomes extremely critical. For these reasons it is
desirable to make the gate signal narrower than the burst to
be counted, thus avoiding miscounts due to both turn-on
and turn-off distortion and to incorrect alignment between
the gate and the RF burst. This is done by using a com
parator in the detector that has two outputs, each with an
independent enable. One output is delayed and then used
to enable the other output, thus causing the leading edge of
the second output to be delayed. However, when the RF
burst ends, both outputs return to the no-signal state simul
taneously. The result is a pulse on the second output that is
narrower than the input pulse by the amount of the delay,
which in this case is 30 ns. This output is translated to the
proper levels and routed to the "gate out" connector on the
5355A's rear panel. From there, it is connected to the gate
control input on the 5345A using the cable supplied with
each 5355A. The IF signal is internally routed to the 5345A
via a delay line of the proper length so that the gate pulse is
centered in the IF burst at the 5345A's main gate flip-flop.
This timing relationship is shown in Fig. 9.
Front Panel
The 5355A has two inputs, one with a range of 400 to
1600 MHz, and the other for the removable high-frequency
RF head. A simple seven-pushbutton keyboard handles all
the measurement and diagnostic functions. Most of the user
applications are handled by three of the keys. The two
right-hand keys determine a PULSE or CW measurement,
and the bottom key selects the appropriate input.
The four remaining keys are used for more sophisticated
measurements, such as manual or offset measurements.
These measurements require keyboard-entered frequen
cies. To enter a manual frequency, the gold MAN FREQ key is
pushed to place the 5355A in gold data entry mode. In this
mode, the gold legends on the front panel apply. By using
HPR/IF (Input Select)
IF Out
Fig. 8. IF filter, detector, and gate
generation circuits.
© Copr. 1949-1998 Hewlett-Packard Co.
prove the resolution to 10 kHz. By frequency averaging,
resolution may be increased to 100 Hz. depending upon the
total measurement time, as shown in Fig. 10.
Gating errors in the 5345A. described previously in the
calibration section, cause frequency errors inversely pro-
RF Burst
Comparator Output
An Automatic Microwave Frequency
Counter Test System to 40 GHz
Enable Input
Gate Output
by Larry L. Koepke
Delayed RF Burst
Fig. 9. 5355A gate timing. The gate signal is shorter than the
detected RF pulse and centered on it. This eliminates mis
counts due to turn-on and turn-off distortion.
the two keys labeled UP and DOWN, the desired manual
frequency can be entered. For a manual measurement the
frequency entered must be within 50 MHz of the input
frequency. Pushing the gold key again restores the 5355A to
its previous measurement mode.
To enter an offset frequency, the blue OFFSET FREQ key is
pushed to place the 5355A in blue data entry mode. In this
mode, the blue legends apply. Using these keys, the sign,
mantissa, decimal point, and units of the offset frequency
can be entered. Pushing the blue key again restores the
5355A to its previous measurement mode.
Diagnostics and special functions are engaged by push
ing two keys simultaneously. Pushing CW and PULSE acti
vates one of two specialized measurement modes. When the
5355A is in CW mode, a long pseudorandom sequence is
activated so that more FM can be tolerated at the input.
When the 5355A is in pulse mode, the calibration factor is
computed continuously (normally it is computed on only
the first ten measurements).
Pushing the blue and gold keys simultaneously engages
various diagnostics and specialized modes. Each time these
two keys are pushed, two digits centered between equals
signs are displayed in the 5345A mainframe for one second.
The digits identify which mode is being activated. A total of
21 modes are available. Presently, 17 have been assigned.
All of the front-panel functions are remotely programma
ble via the HP-IB. The programming resembles pushbut
ton operation for all measurements, diagnostics, and
specialized modes. For data entry of manual and offset
frequencies, a floating-point input format is used.
Resolution and Accuracy
The resolution of a frequency measurement is directly
proportional to the gate time. For example, a I-/LIS gate time
provides a resolution of 1 MHz. To improve the resolution
on CW or repetitive pulsed RF signals, frequency averaging
is used. Averaging improves the resolution by VnT where n
is the number of samples averaged.5 Fora 1-fj.s external gate
signal and a 5345A gate time setting of 10 ms, 104 external
gates are required to complete one measurement and im-
Testing the 5343A Microwave Frequency Counter, the 5355A Au
tomatic Frequency Converter, and the 5356A/B/C Frequency Con
verter Heads for conformity to all of their specifications over their
entire frequency ranges is not a trivial task. To handle this formidable
job, its special automatic test system had to be devised. Some of its
features are:
1 . The software programs are structured to allow the operator to run
a full set of tests automatically (without operator assistance), to
select a single test to run, or to select frequencies and power
levels manually (see Fig. 1).
Special Function Keys
Set and
Read Status
of 5355A
Test 1
CW Sensititivy
and Accuracy
Test 2
Test 3
Test 4
RF Head
CW Sensitivity
and Accuracy
Test 5
RF Head
Test. 64
â € ”
RF Head
Test 7
Fig. 1. Test program structure and special function key as
signments in the Microwave Counter Test System.
* Compatible with IEEE 488-1978.
© Copr. 1949-1998 Hewlett-Packard Co.
2. The system can make numerous repetitive frequency measure
ments at different power levels automatically, freeing the test
technician to align and/or repair instruments that have failed the
automatic tests.
3. The system provides failure reports to help the test technician
locate instrument failures.
4. In the data log mode the system provides a printout of the com
plete test.
5. The operator is made aware of a failure or the end of a test by an
audible signal.
The Microwave Counter Test System is controlled by an HP 9825A
Desktop Computer using HP-IB signal sources (HP 3330B,
HP8660C, HP 8672A) to derive the frequencies of 10 Hz to 40 GHz.
Frequencies of 18.5 GHz to 26.5 GHz are derived by doubling the
8672A frequencies of 9.25 GHz to 13.25 GHz. Frequencies of 26.5
GHz to 40 GHz are derived by quadrupling the 8672A frequencies of
6.625 GHz to 10 GHz.
The system is capable of supplying CW or pulsed RF to the instru
ments under test. A 5359A Time Synthesizer is used to generate
pulses to modulate the CW signal generator outputs. HP 3331 1 B/C
Coaxial Switches used for signal switching and an HP 8495K 10-dB
Step Attenuator are controlled via the HP-IB by three HP 59306A
Relay Actuators. Two HP 436A Power Meters controlled via the HP-IB
and one HP 432C Power Meter controlled via an HP 98032A 1 6-bit I/O
Interface make the required power measurements. Frequency mea
surements on the combined 5355A Converter (listen only) and
5356A/B/C Heads are made by the 5345A Counter and output via the
HP-IB to the 9825A Desktop Computer. 5343A Microwave Frequency
Counter frequency measurements are output directly via the HP-IB to
the 9825A. A separate HP 98034A HP-IB I/O interface was used for
the instruments under test (5343Aor 5355A), so that an HP-IB failure
in one of these instruments would not affect the system instruments
under HP-IB control, which are on another 98034A HP-IB I/O Inter
face. Fig. 2 is the system block diagram.
The instrument test programs, associated data files, and special
function keys are stored on the HP 9885M Flexible Disc. When the
system is first powered up the 9825A Computer automatically loads
track from file 0; the Start program loads the special function keys from
the flexible disc and displays PRESS S.F. KEY FOR DESIRED UNIT. The
operator presses the special function key on the 9825A correspond
ing to program instrument being tested, and the instrument test program
selected by the operator is loaded from the 9885M Flexible Disc into
the 9825A memory and executed from line 0. The operator now
answers questions asked by the 9825A: do you wish to data log? yes
or no; The yes enter the date; enter the instrument serial number. The
complete test is then executed.
Larry L. Koepke
Larry Koepke has been a test and elec
tronic tooling technician with HP since
1959. Born in Rockford, Iowa, he
learned his electronics in the U.S. Army.
He assembled the test system and
wrote the test programs for the 5355A
Frequency Converter, the 5356A/B/C
Heads, and the 5343A Microwave
Counter. A resident of San Jose,
California, Larry is married, has two
daughters and one grandson, and likes
to ride horses and bicycles.
"Compatible with IEEE 488-1978.
12 MHz- 2 GHz
10 Hz- 12 MHz
Disc Drive
and Power
Unit Under
Fig. 2. Block diagram of the Microwave Counter Test System.
© Copr. 1949-1998 Hewlett-Packard Co.
Self Check
10 MHz -r
1 MHz
10 Hz
20 ns 100 ns
s 10 us 100 fis
1 ms
10 ms
External Gate Width
Fig. 10. Gafe error and resolution of frequency average mea
surements as a function of gate width.
portional to external gate width. The calibration routine
improves the accuracy about one order of magnitude. The
residual gate error, shown in Fig. 10, is independent of gate
time and may be decreased to 3 kHz for external gate widths
from about 4/xs to 100/us. Since the resolution of the calibra
tion factor is not zero and secondary 5345A main gate errors
are present, 3 kHz is the accuracy limit. However, calibra
tion is not used for external gate widths greater than 100/xs,
so the accuracy is the same as the 5345A Counter in this
Since no pulse burst standard exists, pulse accuracy mea
surements are made with a CW source with the mainframe
counter externally gated. Actual pulse measurements using
the test equipment described on page 11 typically are more
accurate than the specification. Fig. 11 shows the results of
typical pulse measurements on an 18-GHz synthesizer as a
function of pulse width.
Pulse modulation of a source causes phase modulation of
the carrier, especially during the rise and fall times of the
pulse. This can be a result of direct FM or <t>M, AM-to-FM
conversion, or frequency pulling of the source. A video
signal (feedthrough of the pulse modulation) may also be
present along with the modulated carrier, further distorting
the waveshape. Although the 5355A's adaptive gate
generator removes about 15ns from the leading and trailing
edges of the pulse, some phase modulation may remain,
especially for short pulses. Therefore, frequency accuracy
for burst measurements depends on input signal purity; any
phase perturbations that cannot be removed by the 5355A
will cause errors.
A typical CW statistical measurement of a synthesized
18-GHz source with the source and counter time bases tied
together using a 1-s gate time had a standard deviation (one
sigma) of 0.57 Hz and a mean difference of —0.08 Hz for
1000 measurements.
The 5355A can perform six measurements, two using the
prescaler input and four using the RF head input. With the
prescaler, either pulse or CW mode can be selected. With
the RF head, pulse auto, CW auto, pulse manual, or CW
manual can be selected.
The 6800 microprocessor executes these complex al
gorithms using 12K bytes of ROM and IK bytes of RAM.
With the flexibility the microprocessor allows, it was easy
to implement special self-check routines that execute
whenever the instrument is turned on. In the 5355A, the
two RAM integrated circuits are verified for data-pattern
read/write and addressability. Then the two ROM inte
grated circuits are tested via a checksum. Following RAM
and ROM tests, the synthesizer is programmed to three
known frequencies and performs three 100-jas measurments to verify each setting. Should any of these power-on
tests fail, the operator gets a unique ten-second warning
display per failure. Thereafter, the 5355A will attempt to
follow the measurement algorithm specified by the user.
The team that developed the 5355A and the 5356A/B/C
was as follows: Luiz Peregrino did most of the initial inves
tigation, systems analysis, and synthesizer design. The
MPU/HP-IB hardware was developed by John Shing.
Mohamed Sayed was responsible for the 5356A/B/C heads.
He developed the 40-GHz sampler, VCO, sampler driver,
high-pass filter, and power amplifier hybrids, and provided
the integration and testing of the heads. The prescaler
channel was the responsibility of Hans Jekat, who also
designed the IF amplifiers and provided many solutions to
systems problems. The mechanical designer of the 5355A
was Dick Goo, and of the 5356A/B/C was Keith Leslie. Tool
design was by Jerry Curran. Martin Neil provided the initial
10 MHz-r-
1 MHz--
Typical Measurements
1 kHz--
100 Hz--
-•-10 IJLS PRP-»--«0.10 Duty Factoria
—»••?•—1s*•« — 10 ms Gate Time — »--«-100 ms —
Gate Time
Gate Time
10 Hz
20 ns 100 ns 1 jxs 10 ¿is 100 /¿s 1ms
10 ms
External Gate Width
Fig. 11. Typical measurements on an 18-GHz pulsed source.
Peak pulse power is -10 dBm. Each point is the average of
100 measurements.
© Copr. 1949-1998 Hewlett-Packard Co.
marketing introduction, and Larry Johnson completed that
assignment. Randy Goodner was the service engineer, and
Larry Koepke built the microwave counter test system and
wrote the software. Quality assurance was under the sur
veillance of Joe Bourdet. Alex Campista and Ron Hartter
were the pilot run technicians. Ian Band was the lab en
gineering manager and Roger Smith the microwave counter
section head. Many thanks to all of the people above and to
all of the others that contributed to the production of these
1. J.L. Sorden, "A New Generation in Frequency and Time Mea
surement," Hewlett-Packard Journal, June 1974.
2. A. Har and V.A. Barber, "Microprocessor-Controlled Har
monic Heterodyne Microwave Counter also Measures
Amplitudes," Hewlett-Packard Journal, May 1978.
3. S.I. Gass, "Linear Programming, Methods and Applications,"
McGraw-Hill, 1964.
4. L. Peregrino, "A Technique that Is Insensitive to FM for Deter
mining Harmonic Number and Sideband," Hewlett-Packard Jour
nal, May 1978.
5. D.C. Chu, "Time Interval Averaging; Theory, Problems, and
Solutions," Hewlett-Packard Journal, June 1974.
Robert W. Offermann
Bob Offermann received his BS degree
in electrical engineering from California
Institute of Technology in 1 971 , and for
the next two years combined circuit de
sign work at the U.S. Naval Undersea
R&D Center with graduate studies at
Caltech. He received his MS degree in
1973 and joined HP shortly thereafter.
Bob has contributed to the design of the
5363A Time Interval Probes, done in
vestigations on time interval measure
ments, designed the IF and gating
circuits for the 5355A Frequency Con
verter, and served as the first 5355A
production engineer. A native of Stock
ton, California, he now lives in Saratoga, California. He's married and
enjoys swimming, sailing, ballroom dancing, and theater.
Richard F. Schneider
Dick Schneider is project manager for
the 5355A Frequency Converter and
the 5356A/B Frequency Converter
Heads. With HP since 1964, he's con
tributed to the design of the 5240A,
5260A, and 5257A counter products,
developed several microwave counter
and phase-lock systems, and served as
project manager for the 5340A Counter.
A native of Cleveland, Ohio, he
graduated from Case Institute of
Technology with a BSEE degree in 1952
a and spent several years designing mism sue and satellite test equipment, mi¿< crowave amplifiers, and telemetering,
radar, and receiving systems before joining HP. He also served in the
U.S. and Guard as a Loran specialist. Dick is a member of IEEE and
holds Jose, MSEE degree from California State University at San Jose,
received in 1968. He's married, has two sons, and relaxes with tennis
and woodworking.
Ronald E. Felsenstein
| With HP since 1969, Ron Felsenstein
i designed the processor for the 5345A
I Counter, served for a year as a laser
and logic production engineer, and was
responsible for the 6800 firmware and
the digital interface design for the
5355A Frequency Converter. Born in
Montevideo, Uruguay, he received his
SB degree in electrical engineering
from Massachusetts Institute of
-'*< Technology in 1969. Now a resident of
3*<E^ Santa Clara, California, Ron and his .
family (he's married and has two chil^^ dren) enjoy winter camping in their re.' cently acquired motorhome. Ron col
lects U.S. stamps and coins and is a dedicated do-it-yourselfer when
it comes to car and home repairs.
40-GHz Frequency Converter Heads
by Mohamed M. Sayed
THERE IS AN UPPER LIMIT to the frequencies at
which automatic microwave frequency measure
ments may be made simply by connecting a coaxial
cable between the source and the counter. This is because
the cable's losses are generally greater for higher frequen
cies, thus demanding more sensitivity from the counter,
while microwave counters become less and less sensitive at
higher frequencies. Thus a frequency is reached where not
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 1. Model 5356A/B/C Fre
quency Converter Heads offer a
choice of input connectors and fre
quency ranges for microwave fre
quency counting up to 40 GHz.
enough signal reaches the counter to trigger it properly.
These conditions dictate using a waveguide instead of a
coaxial cable. However, waveguides are suitable only for
certain frequency bands, e.g., K-band (18-26.5 GHz) or
R-band (26.5-40 GHz). Moreover, they are expensive and
lack the mechanical flexibility of the coaxial cable.
Frequency Converter Heads Models 5356A/B/C, Fig. 1,
combine the convenience of coaxial cables with the broad
frequency band of a microwave counter (1.5-40 GHz), and
can be used for either CW or pulse measurements. These
heads convert microwave frequencies to intermediate fre
quencies (IF) using the sampling technique. The sampling
frequency input to the head and the IF output from it are
connected by a 1.68-metre cable to the 5355A Automatic
Frequency Converter (see article, page 3). The microwave
input frequency to the head is calculated from measure
ments of the IF and the answer is displayed on the 5345A
Counter (the 5355A is a plug-in for the 5345A). The down-
conversion is performed completely in the head and only
the IF is .connected to the 5355A.
To cover the frequency band up to 40 GHz, four different
connectors are available: N, SMA, APC-3.5 and waveguide.
Three models and two options are available to suit various
applications. Table 1 shows the frequency ranges of the five
The heads have male connectors because sources gener
ally have female connectors. The 5356A has an N-type
connector and is useful up to 18 GHz. The 5356B has an
SMA connector, and can be used up to 26.5 GHz. To
strengthen the SMA male connector, a special collar was
designed to protect it. The collar also makes it easier for the
customer to connect it to the source. Some customers prefer
using a K-band waveguide from 18 to 26.5 GHz; the 5356B
Option 001 has a WR-42 connector to meet this need. The
Input f,
1.5-40 GHz
Table 1
Frequency Ranges and Connectors of
Model 5356A/B/C Frequency Converter Heads
IF Output
1-528 MHz
to 5355A Automatic
Frequency Converter
VVCO Input
885.2-1056 MHz
from 5355A
Fig. 2. Simplified block diagram of the 5356A/B/C Frequency
Converter Heads.
© Copr. 1949-1998 Hewlett-Packard Co.
40-GHz Synthesizer Tests Frequency
Converter Heads
by Mohamed M. Sayed
APC-3.5 connector, which is mode-free to 34 GHz,1 is used
up to 40 GHz in the 5356C. The effect of this connector's
moding between 34 and 40 GHz is taken into account by
reducing the specified sensitivity by 5 dB in this region. A
special collar was designed for the APC-3.5 connector to
strengthen it and to protect it from damage. Waveguide
connector WR-28 is also offered, for customers using only
R-band (26.5 to 40 GHz).
Frequency Converter Head Design
To test the accuracy of the 5356/5355 system in CW and pulse
modes, a synthesized source is needed, especially to check for
±1-count accuracy with a one-second gate time. Both synthesizers
and pulse modulators are commerically available up to 18 GHz, and
these are used in the 5356/5355 test system. To cover the 18-to26.5-GHz band, an amplifier and a K-band doubler are used. The
input frequencies to the doubler are 9 to 13.25 GHz and the output
frequencies are 18 to 26.5 GHz as shown in Fig. 1.
To 5356B/C
18-26.5 GHz
9-13.25 GHz
Fig. 2 shows the block diagram of the 5356A/B/C. The
input voltage-controlled oscillator (VCO) frequency to the
5356A/B/C varies from 885.2 MHz to 1056 MHz and the
output IF to the 5355A varies from 1 MHz to 528 MHz. The
IF output is proportional to the RF input within the
5356A/B/C's dynamic range. All of the components in the
head are built in thin-film microcircuit configurations.
The coaxial assembly shown in Fig. 2 is replaced by a
high-pass filter for the 5356A Option 001. Since the 5356/
5355 is a pulse counter, the input pulse may contain a video
signal. This signal may be so large that it overloads the
counter, especially since the IF gain is about 80 dB. To
attenuate such video signals, Model 5356A option 001 has a
high-pass filter between the input connector and the sam
pler. The filter's maximum insertion loss from 1.5 to 18 GHz
Fig. 1. Generation of an 18-to-26.5-GHz signal to test the
5356B/C Heads.
In the 26.5-to-40-GHz range an R-band doubler can be used, with
the primary synthesizer operating from 13.25 to 20 GHz. This means
that the synthesizer, the pulse modulator, and the amplifier must also
operate up to 20 GHz. Since most of the instruments operate only up
to 18 GHz, major modifications would have been needed.
The simpler method that is actually used to generate a synthesized
signal to 40 GHz is to use an amplifier and two doublers in cascade.
The first stage is an amplifier-doubter, and the second stage is
another doubler. The primary synthesizer, the pulse modulator, and
the amplifier operate in the frequency range 6.625 to 10 GHz. Fig. 2
shows a block diagram of the 26.5-to-40-GHz synthesizer.
6.625-10 GHz
6.625-10 GHz 13.25-20 GHz
885.2 to
1056 MHz
Fig. 2. Generation of a 26.5-to-40-GHz signal to test the
Special thanks are due Roger Stancliff of the HP Santa Rosa
Division for his help in designing the amplifier doubler.
1 to 40 GHz
Fig. 3. Photo and schematic of the 5356A/B/C sampler driver.
L1 is drive inductance and C, is tuning capacitance. L2, C2,
L3, and C3 form a matching network to match the impedance
seen at C-, to 50(1.
© Copr. 1949-1998 Hewlett-Packard Co.
S. -5-I -10Ã-15-i
» -25 +
-35-12 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Frequency (GHz)
is 1 dB, and its minimum insertion loss below 100 MHz is
more than 35 dB.
The power amplifier is housed in a TO-8 package using a
thin-film alumina substrate. It consists of two stages, a gain
stage and a power stage. The amplifier is driven to satura
tion so that its output is insensitive to input variations. The
IF amplifier is housed in a TO-12 package using a thin-film
sapphire substrate. It also consists of two stages.
The head casting consists of an upper half and a lower
half. The sampler and the two printed circuit boards are
mounted on the lower half. Heat sink materials are attached
to the upper half to dissipate the heat from the power
amplifier and the IF amplifier. There is also a heat sink on
the lower half for the power amplifier. As a result, the
temperature rise is less than 9°C. The heat sink materials
also serve as shock absorbers for mechanical vibrations. The
casting is designed to accept the different input connectors:
N, SMA, APC-3.5, waveguide WR-42, and waveguide
WR-28. The casting is also designed for improved EMI
(electromagnetic interference) performance. All of the parts
inside the head can be disassembled easily, using a screw
driver and an SMA wrench. As a result, the 5356A/B/C is
easy to troubleshoot.
Fig. 4. Spectrum of the sampler
driver output shows comb lines to
40 GHz. The input frequency is
1 GHz.
IF Out
Sampler Driver
The heart of the frequency converter head is the sampler
driver. Fig. 3 shows a schematic diagram. The driver is built
in a coaxial package. The step-recovery diode has a very fast
rise time that generates a comb of harmonics of the VCO
frequency. One of the comb frequencies is heterodyned
with the input microwave frequency to produce an IF out
put in the proper range.
When time domain measurements were used to test the
sampler driver, the test parameters were very sensitive to
operator error and test equipment limitations. Therefore,
frequency domain measurements are used. Since a 1-to40-GHz spectrum analyzer (without external mixer) wasn't
commercially available, an in-house 40-GHz spectrum
analyzer was designed. Fig. 4 shows typical comb lines up
to 40 GHz for a 1-GHz VCO. A 40-GHz test fixture was also
designed to test the driver and to adjust the tuning element
before sealing it. The input VCO frequency is varied from
885 MHz to 1056 MHz and the 38 comb lines are adjusted
to meet the required counter sensitivity. The input to the
Fig. 5. 40-GHz sampler used in the 5356C Head is a thin-film
hybrid circuit, fl, and R2 are chosen to optimize bandwidth
and dynamic range.
© Copr. 1949-1998 Hewlett-Packard Co.
sampler driver is almost constant since the power amplifier
is working in the saturation region. Therefore, the comb
lines are insensitive to power amplifier variations and
6 "
Fig. 5 shows a schematic diagram of the sampler. It con
sists of a thin-film hybrid mounted in an aluminum pack
age. Two versions of the sampler are used. The 5356A/B
sampler is the same as the one in the 5343A Counter with
out the thin-film buffer amplifier.2'3-4 The 5356C 40-GHz
sampler is the same basic design with slightly different
component values. For the 5356C, Rt and R2 are chosen to
maximize the sampler's dynamic range up to 40 GHz.
In both versions of the sampler, two beam-lead
Schottky-barrier diodes are placed on the hybrid across the
slotted line. This type of diode provides a low, easily con
trolled inductance and is easy to mount on the thin-film
substrate by the thermocompression bonding technique.
To work up to 40 GHz, the diodes are chosen for
minimum series resistance, junction capacitance, and stray
capacitance. The diode capacitance is incorporated in a
low-pass filter that has a cutoff frequency of 41 GHz. The
circuit is optimized for low SWR up to 40 GHz using an
in-house OPSNAP computer program. Fig. 6 shows the
relative sampler conversion efficiency up to 40 GHz, and
Fig. 7 shows the return loss up to 40 GHz. Four experimen
tal setups were used for these measurements: 1.5-2.4 GHz,
2-18 GHz, 18-26.5 GHz and 26.5-40 GHz.
The sampler's IF output is designed to be insensitive to
sampler driver variations. The minimum output from the
sampler driver is enough to drive the sampler into satura
tion. Therefore the IF output is also insensitive to tempera
ture variations.
5 9-o
ir 10--
5355A Compatibility
The design goal was to make any 5356A/B/C work with
any 5355A. The interface between these two instruments is
analog, since the VCO signal comes from the 5355A to drive
the power amplifier in the 5356A/B/C, and the IF comes
from the 5356A/B/C to drive the IF amplifier in the 5355A.
To guarantee complete compatibility the following condi
tions were established.
• The IF output from the 5356A/B/C is insensitive to the
level of the VCO input from the 5355A. The lowest VCO
input power level at any frequency is sufficient to drive
8 - -
Frequency (GHz)
Fig. 7. Return loss of the 40 GHz sampler.
the power amplifier to saturation. Thus the output of the
sampler driver is insensitive to the 5355A VCO level.
The IF output from the 5356A/B/C to the 5355A is suffi
cient to guarantee the minimum sensitivity of the com
bined system. The counter sensitivity is defined as the
5356A/B/C conversion efficiency (RF to IF) plus the
5355A IF sensitivity. The 5355A IF sensitivity is adjusted
to meet the required specifications and the 5356A/B/C
conversion efficiency is tested from 1.5 to 40 GHz to
assure that it meets the necessary levels.
The input of the 5355A is unconditionally stable so that it
will not oscillate with any 5356A/B/C. Also, the output of
the 5356A/B/C is unconditionally stable with any 5355A.
This is especially important since the IF gain of the com
bined system can exceed 80 dB.
Sensitivity, Flatness, and Distortion
Since the 5356C is so broadband, there were trade
offs to be made among sensitivity, frequency response flat
ness, and distortion caused by sampler overload. The bias
resistor R2 between the sampler IF output terminals (Fig. 5)
was chosen to maximize the dynamic range of the com
bined 5356C and 5355A up to 40 GHz.
Dynamic range is a function of frequency. For the 5356C
it ranges from -25 dBm to +5 dBm below 12.4 GHz and
from -20 dBm to +15 dBm above 12.4 GHz for full accuracy
(±1 count). However, the harmonic number is correctly
Frequency (GHz)
Frequency (GHz)
Fig. 6. Relative conversion efficiency of the 40 GHz sampler.
The IF is 300 MHz and the VCO frequency from the 5355A is
1 GHz.
Fig. to Relative conversion efficiency of the 5356A/B/C up to
40 GHz. The IF is 100 MHz and the VCO frequency is 1055
MHz. Note the higher sensitivity of the 5356C.
© Copr. 1949-1998 Hewlett-Packard Co.
the variation of the conversion efficiency with temperature.
A 5355A that has the lowest IF sensitivity within the
system specifications is used to test each 5356A/B/C. Fig. 10
shows the combined sensitivity for CW and pulses.
1 --
S -2
2 0
4 0
Temperature (°C)
Fig. 9. Relative change of 5356A/B/C conversion efficiency
with temperature. The IF is 100 MHz and the VCO frequency is
1055 MHz.
determined for a wider range of input signal levels: —30 to
+ 8 dBm below 12.4 GHz and -25 to +18 dBm above
12.4 GHz.
The dynamic range of the 5356A can be shifted by using
one of the HP 84938-Series Attenuators to replace the coax
ial assembly (see Fig. 2). For example, the 8493B Option 010
will make the dynamic range —10 to +15 dBm instead of
-20 to +5 dBm. The damage level will change from +25
dBm to +33 dBm CW and +35 dBm pulse.
Fig. 8 shows the relative conversion efficiency of the
5356A up to 18 GHz, the 5356B up to 26.5 GHz, and the
5356C up to 40 GHz. These curves are for 25°C. Fig. 9 shows
Frequency (GHz)
Fig. 10. (a) Sensitivity of the 5356C/5355A system, (b) Sen
sitivity of the 5356A/5355A and 5356B/5355A systems. In all
cases the 5355A is a worst-case unit.
The author would like to thank all members of the hybrid
department of the HP Santa Clara Division, especially
Kathy Luiz, who assembled the original thin-film circuits.
The 5356 product design was accomplished very effec
tively by Keith Leslie. Special thanks are due Jeff Wolfington and Al Barber for their constructive criticism dur
ing the course of this project. Many individuals from the HP
Santa Rosa, Stanford Park, and Microwave Semiconductor
Divisions deserve credit for their help and constructive
discussion, especially Young Dae Kim of Stanford Park for
his help in designing the high-pass filter for the 5356A
Option 001. The product introduction of the 5356A/B was
handled by Martin Neil, and of the 5356C by Larry Johnson
and Doug Nichols. Service engineers were Randy Goodner
for the 5356A/B and Joe Dore for the 5356C. Production
engineers were Bob Offermann for the 5356A/B and Art
Bloedorn for the 5356C. Special thanks are due to Luiz
Peregrino for his continuing encouragement. The author
would like to express his appreciation to Roger Smith,
microwave section manager and Ian Band, engineering lab
manager, for their support and interest in this project.
1. G.R. Kirkpatrick, R.E. Pratt, and D.R. Chambers, "Coaxial Com
ponents and Accessories for Broadband Operation to 26.5 GHz,"
Hewlett-Packard Journal, June 1977.
2. J. Mi "A dc-to-20-GHz Thin-Film Signal Sampler for Mi
crowave Instrumentation," Hewlett-Packard Journal, April 1973.
3. A. Bologlu and V.A. Barber, "Microprocessor-Controlled
Harmonic Heterodyne Microwave Counter also Measures
Amplitudes," Hewlett-Packard Journal, May 1978.
4. A. Bologlu, this issue, p. 20.
Mohamed M. Sayed
Mohamed Sayed joined HP's Micro
wave Technology Center in 1 973. After
working on microwave silicon and
GaAs FET transistors for two years, he
joined HP's Santa Clara Division, and
since that time has worked on micro
wave counters. Born in Egypt, he re
ceived his BSEE and MSEE degrees
from Cairo University and his PhD from
Johns Hopkins University in Baltimore,
Maryland. He has taught at Cairo,
Johns Hopkins, and Howard Univer
sities, and is currently teaching at San
Jose State University. Before joining
HP, he spent a year doing research on
solar in at the University of Delaware. He has published papers in
the field of microwave measurements, microwave transistors, and
solar energy. He's a member of IEEE and is active in the National
Alumni Schools Committee of Johns Hopkins University. Mohamed is
married, has a daughter, and lives in Cupertino, California. He is
currently attending Santa Clara University to obtain his Master's
degree read engineering management. In his spare time he likes to read
and travel.
© Copr. 1949-1998 Hewlett-Packard Co.
A 26.5-GHz Automatic Frequency Counter
with Enhanced Dynamic Range
A new sampler provides higher frequency coverage and
10 dB greater sensitivity than previous designs.
by All Bologlu
restricted by the speed of today's logic circuitry
to maximum frequencies of 500 MHz or so. Con
sequently, automatic microwave counters must employ
some method of down-conversion to extend counting into
the gigahertz range. Traditional techniques, such as the
transfer oscillator and heterodyne techniques, were
supplemented by the harmonic heterodyne technique with
the introduction of the HP 5342A in the spring of 1978. l The
advent of the microprocessor made this technique possible
along with a significant reduction in instrument cost.
A new microwave frequency counter, Model 5343A,
makes its own contribution by extending the frequency
range to 26.5 GHz and improving sensitivity and dynamic
range by about 1 0 dB across the band. Furthermore, features
have been added, making the instrument more systemoriented.
to 300 MHz in 100-kHz steps. The offset oscillator fre
quency f2 is maintained at fj — 500 kHz by a phase-locked
loop. When the IF detector indicates the presence of an
IF signal in the range of 50 MHz to 100 MHz the synthesizer
stops its sweep and the counter starts its determination of
the harmonic number N. The pseudorandom sequence out
put switches between the main oscillator and the offset
oscillator and between counters A and B. Counter A ac
cumulates fIF1 and counter B accumulates fIF2. The
pseudorandom sequence is then disabled, the main os
cillator is selected and the frequency fj is measured by
counter A to the selected resolution. The pseudorandom
sequence prevents any coherence between the switching
rate of the multiplexer and the modulation rate of the FM
that might be present on the input signal. Such coherence
might produce an incorrect computation of N. Finally, the
harmonic number and the sign of the IF are computed and
the input frequency fx is computed as follows:
System Architecture
The block diagram of the 5343A is very similar to that of
the 5342A (see reference 1). Besides software, the major
changes are in the sampler area, which will be dealt with in
detail later in this article. The operating algorithm is as
follows. The multiplexer selects the main oscillator output
and the main oscillator frequency fa is swept from 350 MHz
fx = Nf, - fIFl
IF2 fIFl
fx = Nfa + fIF1
ÃIF2 >
N =
It has been shown2 that the length of the pseudsorandom
-20 r
-25 E
-30 -
-35 -
Fig. 1 . Model 5343A Microwave Frequency Counter provides
high in and automatic amplitude discrimination in CW
frequency measurements to 26.5 GHz. Offsets and scale fac
tors can be entered via the front panel.
© Copr. 1949-1998 Hewlett-Packard Co.
Frequency (GHz)
Fig. 2. 5343A sensitivity.
sequence required to tolerate frequency modulation on the
input signal is given by the expression
Thus to allow the counter to tolerate 10 MHz peak FM on
Pulse Input
Thin Film
the input signal with Af = 500 kHz, the P value should
exceed 25,600. Since P = 2m - 1 where m is the number of
shift register stages in the pseudorandom sequence
generator, a 15-stage shift register would be needed to gen
erate this sequence. It is obvious from the expression for P
that the more FM one wants to tolerate the longer the se
quence has to be. which in turn affects the counter's mea
surement time.
The 5343A has three different sequence lengths of 22 ms,
360 ms, and 2.2 s. The corresponding FM deviations the
instrument tolerates in the automatic mode are 6 MHz, 20
MHz, and 50 MHz peak to peak, respectively. These limits
in fact only apply when the modulation rate is synchronous
with the pseudorandom sequence. Since the modulation in
microwave communication systems is usually either data
or voice, the probability of synchronization is very remote.
Consequently, although the deviation may be large, the
signals may often be measured by the 5343A using the short
sequence length, thereby making acquisition times faster.
New Sampler Improves Sensitivity
RF Input
Diode, Beam
L e a d M i c r o s t r i p
+ 5V
1 kil
IF Out
100 pF^=
22 kfl
Front-Panel Inputs and Controls
RF Input
1 8 pF
100! Ã-
Fig. 3. The principal design
crowave Counter is this new
previous designs. A thin-film
impedance match between the
5343A sensitivity is shown in Fig. 2. The main con
tributor to this improved sensitivity is a new microwave
sampler, which is the only microwave component in the
instrument. Operation of the sampler is similar to the sam
pler used in the 5 342 A, ' the main difference being the use of
a thin-film buffer amplifier to provide a better impedance
match between the output of the sampler and the first IF.
In this sampler structure (Fig. 3), the sampling pulse
couples to the slotted line through a microstrip balun that
generates two opposite-polarity pulses to drive the sam
pling diodes. The down-converted signal is taken from two
isolated resistors to the second substrate in the structure,
which is the buffer amplifier. Resistors across the slot are
used to absorb secondary reflections introduced by the
sampling pulse. The input structure forms the essence of a
low-pass filter with an effective cut-off greater than 26.5
GHz. This structure provides an input return loss as shown
in Fig. 4.
contribution in the 5343A Mi
sampler, a refined version of
buffer amplifier improves the
sampler output and the first IF
The 5343A has two inputs, one going from 10 Hz to 520
MHz and the other from 500 MHz to 26.5 GHz. The righthand side of the front panel deals with input signal channel
selection and sample rate control of the measurement. The
left-hand side of the front panel enables the user to do data
manipulation by keyboard control of the processor. Instruc
tions for doing this are on a label that is affixed to the
instrument top.
The panel layout is in algebraic notation, thereby making
panel operation closely resemble remote programming via
the HP Interface Bus (HP-IB *). When the instrument powers
up it is in the auto mode with 1-Hz resolution. As the user
selects other resolutions, insignificant zeros are truncated.
Display digits are in groups of three to facilitate reading.
In case the user wants to bypass the acquisition cycle of
the algorithm, a manual mode of operation is available. In
this mode the user should know the unknown frequency
within 50 MHz and enter it via the keyboard. The counter
then acts like a receiver making frequency measurements.
Offsets can be specified from the front panel. Any fre'Compatible with ANSI/IEEE 488-1978
© Copr. 1949-1998 Hewlett-Packard Co.
o rAli Bologlu
Ali Bologlu has been with HP for fifteen
years and has been project manager
for microwave counters since 1970.
He's contributed to the design of many
HP frequency synthesizers and micro
wave counters, most recently the
5343A. Ali received BS and MS de
grees in electrical engineering in 1962
and 1963 from Michigan State Univer
sity and the degree of Electrical En
gineer f rom Stanford University in 1965.
Born in Istanbul, Turkey, he's married,
has three children, and now lives in
Mountain View, California. He plays
tennis, enjoys water sports, and
coaches a youth soccer team.
5, -10
-20 -
where m is the multiplying integer, x the measured fre
quency and b the offset.
Frequency (MHz)
Fig. 4. Return loss of the 5343A sampler.
quency offset can either be subtracted from or added to the
measured frequency. In the auto offset mode of operation
the counter holds the initial measurement and then dis
plays all succeeding measurements as deviations about the
initial reading. Frequency readings may also be multiplied
by integers and offsets then added to the product, in effect
solving the equation
y = mx + b
I would like to acknowledge the efforts of John Shing, Jeff
Wolfington, and Keith Leslie for their important and timely
contributions to the 5343A.
1. A. Har and V.A. Barber, "Microprocessor-Controlled Har
monic Heterodyne Microwave Counter also Measures Ampli
tudes," Hewlett-Packard Journal, May 1978.
2. L. Peregrino, "A Technique that Is Insensitive to FM for Deter
mining Harmonic Number and Sideband," Hewlett-Packard Jour
nal, May 1978.
HP Model 5343A Microwave Frequency Counter
Input 1
Input 2
ACCURACY: ~5 mV ±0.3 mV/"C (from 25°C).
FREQUENCY RANGE: 10 Hz to 520 MHz Direct Count.
CONVERSION SPEED: <50 ^s to ±0.01% of full scale readmi
SENSITIVITY: 50Ã1 10 Hz to 520 MHz 25 mV rms; 1 Mil 10 Hz to 25 MHz 50 mV rrns.
5 0 0
M H z - 1 2 . 4
G H z
- 3 3
d B m
IMPEDANCE: Selectable: 1 Mil, <50 pF or 50Ã1 nominal.
OUTPUT: 5 mA. Impedance <1.0 Ohm.
12.4 GHz-18.0 GHz -28 dBm
1 8 . O G H z - 2 6 . 5 G H z
CONNECTOR: Type BNC female on rear panel.
CONNECTOR: Type BNC female.
- 2 3
d B m
5.0 Vrms
ACCURACY: ±1 count -time base error
RESOLUTPON: Front-panel pushbutton select 1 Hz to 1 MHz.
12.4 GHz-18.0 GHz 35 dB
18.0 GHz-26.5 GHz 30 dB
DAMAGE LEVEL: -25 dBm, peak.
IMPEDANCE: 50 ohms, nominal.
CONNECTOR: APC-3.5 male with collar (SMA compatible).
5OO MHz-10 GHz <2:1 typical
MAXIMUM INPUT: 501Ã 3.5 Vrms ( + 24 dBm) 5 Vdc fuse protected; 1 Mil 200 Vdc H
18 GHz-26.5 GHz <.3:1 typical
COUPLING: dc to load, ac to instrument.
FM TOLERANCE: switch selectable on rear panel. For modulation rates from dc to 1 0 MHz.
WIDE: 50 MHz p-p worst case.
NORMAL: 20 MHz p-p worst case.
Time Base
SELF-CHECK: Selected from front-panel pushbuttons. Measures 75 MHz for
SHORT TERM: <1x-10~9for 1 s avg. time.
TEMPERATURE: <±1x10~^ over the range 0°C to 50°C.
OUTPUT FREQUENCY: 10 MHz. s2.4V square wa
wnpatible), 1.5V peak-
to-peak into 50Ã! available from rear panel BNC.
EXTERNAL TIME BASE: Requires 10 MHz. 2.0V peak-to-peak sine wave or square
wave either 1 kil via rear panel BNC connector. Switch selects either internal or extern;
time base.
NARROW: 6 MHz p-p worst case.
MANUAL: Center frequency entered to wrthm ±50 MHz of true value
Option separate provides an oven-controlled crystal oscillator time base, 10544A (see separate
data sheet), thai results in better accuracy and longer periods between calibration.
AGING RATE: <5x10~'°/day after 24-hour warm-up.
AUTOMATIC MODE: Narrow FM: 200 ms worst case.
Normal FM: 530 ms worst case.
Wide FM: 2.4 s worst case.
MANUAL MODE- 80 ms after frequency entered.
SHORT TERM: <ix1o~10for 1 s avg. time.
6 dB (lypical) above any signal within 500 MHz: 20 dB above any signal,
500 MHz-26.5 GHz.
multiplied by any integer up to 99. Then offset can be added c
jred fi
iubtractedlory = mx +b
TOTALIZE: Input 2 can totalize at rate up to 520 MHz. Readout on the fly is controlled by
typical at
SWEEP MODE: Selected from front-panel pushbutton. Allows interface
SAMPLE RATE: Variable from less than 20
D appropriate
which holds display indefinitely.
IF OUT: Rear-panel BNC connector provides 25 MHz
125 MHz output of down-
V rms.
- .- 5%.
- 10V
46- 66
TEMPERATURE: <7x10~9 over the range 0°C to 50°C.
ACCESSORIES FURNISHED: Power cord. 229 cm 7Vfc ft.
LINE VARIATION: <1x10~'°for 10% change from nominal.
SIZE: 133 mm H « 213 mm W * 498mm D (5V< • SH • 19% in).
WARM-UP: <-5x10~9 of final vaiue 20 minutes after turn-on, at 25=C.
Automatically measures the largest oi all signals present, providing that signal is:
MULTIPLY ROUTINE: Selected from front-panel pushbutton
EXTERNAL TRIGGER: TTL type low-level c
Optional Time Base
Option 001
sensitivity range.
panel pushbuttons. Displayed frequency is off-
set by entered value to 1 Hz resc
LINE VARIATION: <±1 x 10"7 for 10% change from nominal.
the sensitivity specification.
AUTOMATIC: Counter automatically acquires and displays highest level signal within
external higher-stability timebase <4x10~11 rms typical.
DISPLAY: 11 -digit LED display, sectionalized to read GHz, MHz. kHz. and Hz.
AGING RATE: <1 x10~7 per month.
AM TOLERANCE: Any modulation index provided the minimum signal level is not less than
WEIGHT: Net 9.1 kg (20 Ib). Shipping 12.7 kg (28 Ib).
PRICES High U.S.A.: 5343A Microwave Frequency Counter. $5,200. Option 001 High
Digital-to-Analog Converter
Option 004
into an full voltage output. A display ot 000 produces 0V output. 999 produces 9 99V full
Stability Time Base, $500. Option 004 Digital-to-Analog Converter, $250. Option
011 Digital Input'Outpul (HP-IB). $350.
5301 Stevens Creek Boulevard
Santa Clara. California 95050 U.S.A.
© Copr. 1949-1998 Hewlett-Packard Co.
Microwave Counter Measurements
by Richard F. Schneider
parameters that must be measured are the average
burst frequency, the pulse repetition frequency
(PRF). the pulse repetition period (PRP), the pulse repeti
tion interval (PRI), and the pulse width. These measure
ments are automatically made by connecting the equipment
as shown in Fig. 1. To measure the average burst frequency,
the user need only assure that the peak pulse power is
within the counter system specifications. Usually a test port
is available in the form of a directional coupler built into the
radar system, or a test horn can be connected to the 5356A/
B/C Head for measurements after the radar system has been
"buttoned up", such as on the flight line. Since the gating
signal for the 5345A Counter is generated by the 5355A
Converter, no auxiliary equipment is required. This gate
signal, as described in the article on page 3, is about 30 ns
shorter than the RF burst to avoid turn-on and turn-off
transients. The measurement is made by selecting the
plug-in and external gate functions of the 5345A Counter
and the pulse mode of the 5355A Converter. The average
burst frequency is then displayed on the 5345A Counter.
5345A Automatic
Counter Frequency
Frequency Converter Head
1.5 to 40 GHz
Fig. 1 . 7"esi sefup for conventional radar measurements (av
erage burst frequency, pulse repetition frequency, pulse
repetition period, pulse repetition interval, pulse width).
Since the IF is detected in the 5355A Converter plug-in
and is available on the rear panel PULSE OUT connector, this
signal can be used to make the other measurements. PRF is
measured by setting the 5345A Counter function switch to
FREQ A after adjusting Channel A to the proper levels. PRP
is measured by simply setting the 5345A Counter function
switch to PERIOD A. Pulse width is measured by setting the
function switch to TIME INT A TO B, the input to COMMON,
channel A slope to -, and channel B slope to +. PRI is
measured by reversing the A and B channel slope polarities.
Pulse Out
5345A Automatic
Counter Frequency
5359A Time
As Narrow
as 20 ns
Fig. 2. 7esf setup for frequency profiling of an RF burst.
Since the IF is limited before detection, pulse width mea
surement accuracy is dependent on the rise time and the
pulse width. For slow rise times, the pulse out (and the
measured width) will probably be longer than the width
defined by the time interval between actual 50% points,
while for fast rise times it will be within 3% of the actual
pulse width. This occurs since the pulse out signal starts
and stops when the input RF level exceeds the system
sensitivity. Jitter can also occur in the pulse width mea
surement if the pulse modulation is not coherent with the
carrier. This can cause the IF envelope to vary by one period
of the IF, orin the worst case, as much as ±10 ns. However,
this jitter is automatically averaged out if time interval
averaging is used.
Frequency Profiling and CW Measurements
Frequency profiling of an RF burst is done using a time
synthesizer such as the HP 5359A, as shown in Fig. 2. Here
the 5359A Time Synthesizer is triggered by the pulse out
put signal from the 5355A Converter plug-in. The 5359A's
delayed output pulse, a -1.0-volt signal, is used to enable
the 5345A Counter's gate control input. The width of this
external gate signal determines the gate time of the 5345 A.
The delay is incremented after each measurement so that
measurements are made at successively later times within
the RF burst. Fig. 3 shows the frequency profile and spec
trum of a 250-ns wide chirp pulse. The chirp has a
bandwidth of 200 MHz; it was generated by ramping a
voltage-controlled oscillator (VCO) and synchronously
pulse modulating its output. For chirp radar applications,
the linearity of the ramp is paramount, while Doppler
radars require minimal FM on the burst. These measure
ments can be made to 1 00-Hz resolution with external gates
as narrow as 20 ns.
For CW measurements, such as on a CW radar, a STALO
(stabilized local oscillator), ora COHO (coherent oscillator),
© Copr. 1949-1998 Hewlett-Packard Co.
vco .,
"! c
2 2. ,--
250 ns»r (To)
I, (From)
50 MHz
10 dB
t, t,
10.55 MHz
Fig. 3. Frequency profile and spectrum of a 250-ns chirp
pulse. Profile can be measured using a 20-ns gate.
frequencies are measured to 1-Hz resolution in one second.
Also, the average frequency of a fully loaded microwave
carrier with traffic can be measured, since the counter sys
tem's specified FM tolerance is 15 MHz p-p (80 MHz p-p in
the special FM mode).
VCO Measurements
Transient measurements needed to evaluate a VCO are
settling time and post-tuning drift, as shown in Fig. 4. The
settling time is the time (tst) required for the output fre
quency to enter and stay within a specified error band
(±fst) centered around a reference frequency (f) after appli
cation of a set input voltage. Post-tuning drift is the maxi
mum change in frequency (±fptd) during the time interval
ta to t2, where tj is a specified time after t0, the start of the
step input voltage. These measurements may be made with
the equipment shown in Fig. 5.
In this measurement, the tuning voltage pulse generator
is set to the necessary step voltage levels to drive the VCO.
This generator also provides an external trigger to the
5359A Time Synthesizer, which synchronizes the 5345A
Counter by generating a gate control input signal. The time
(tr) that it takes the VCO to reach the reference frequency is
then set into the time synthesizer as a delay, and the width
of the synthesizer output pulse is set to an appropriate gate
width for the measurement. Then selecting the automatic or
manual pulse mode of the 5355A Converter plug-in causes
the counter to display the VCO reference frequency.
The start of the reference frequency step may also be
Fig. 4. Transient measurements needed to evaluate a
voltage-controlled oscillator (VCO) are settling time and posttuning drift.
observed by stepping the time synthesizer delay by one-half
the step generator period. In this manner, the absolute
levels of the VCO step generator may be adjusted to set the
initial and final reference frequencies. Next, the delay is
decreased until the displayed counter frequency exceeds
the specified error band (±fst). This delay defines the set
tling time. In a similar manner, the post-tuning drift may be
measured by observing the change in frequency from delay
time ta to time t2. In both measurements, changes of fre
quency may be easily observed by setting the counter to
Input X
5359A Time
Ext Trigger
Fig. 5>. fesÃ- setup for VCO measurements shown in Fig. 4.
© Copr. 1949-1998 Hewlett-Packard Co.
-I— f- â € ”
Time (ns)
Fig. 6. Settling time measurement on a VCO.
display deviations from the reference frequency. This is
done by subtracting the reference frequency by entering the
last measurement into the 53 55 A Converter as a frequency
For large frequency steps, the 5355A Converter will have
to reacquire the signal when changing from the initial to the
final frequencies. This can be prevented if the initial and
final frequencies fall within the 5355A IF bandwidth for
two harmonics of the 5355A synthesizer frequency that is
used to down-convert the microwave input signal (see arti
cle, page 3). For instance, if the VCO is stepped from 8 to 12
GHz, a synthesizer frequency may be found to satisfy both of
these frequencies. By using the 5355A's diagnostic mode 9,
the synthesizer frequency may be set to 1025.2 MHz. The
8th harmonic is 8201.6 MHz, and the IF frequency is 201.6
MHz. The 12th harmonic is 12.3024 GHz and the IF fre
quency is 302.4 MHz, which is still within the limits of the
IF bandwidth. The actual IF can be observed in diagnostic
mode 10.
By adjusting the synthesizer frequency and the initial or
final VCO frequency, it is possible to nearly center both IFs
so maximum deviations can be measured. It is possible then
to observe frequency transients whose excursions are less
than one-half the IF bandwidth. The size of the step from
initial to final VCO frequencies is ultimately limited by the
bandwidth of the 5356A/B/C Frequency Converter Head
being used.
A settling time measurement of a VCO is shown in Fig. 6.
Fully automatic measurements can be configured using the
HP- IB to control the time synthesizer (delay generator),
plotter, tuning voltage pulser, and digital voltmeter. Other
VCO measurements, such as frequency accuracy, frequency
range, frequency linearity, pushing and pulling factors,
modulation sensitivity, hysteresis, and warm-up frequency
drift may be easily made in the CW mode.
Here issue: the caption that was missing from page 31 of our March issue:
Fig. 7. Calibration oscillator. The LC tank circuit alternately
turns off each transistor so the output power is restricted to I02RI4.
I0 and Ft were chosen to make this equal to -10 dBm.
Laboratory Notebook
A Flexible Software Development Technique
A common problem in the development of microprocessorbased instruments is that there comes a time when the software has
to be committed to firmware in the form of read-onJy memories
(ROMs). Generally the software development engineer delays
committing to masked ROMs as long as possible to avoid the costly
mask changes that would be required should "bugs" or necessary
modifications appear in the future. The software development
technique used for the 5355A Automatic Frequency Converter (see
article, page 3) is flexible enough to allow an early commitment to
ROMs without being penalized for changes later on. This technique
can be used in any instrument where the ROMs are not coded tofull
The technique also offers advantages once the instrument is in
production. Generally, it is possible to implement a relatively
complex software modification by changing one correction ROM
instead of all ROMs. If the correction ROM can be replaced with an
EROM (erasable ROM), then the production change can be im
plemented without incurring a long lead time for a new masked
To explain the technique, the 5355A can be used as a case
history. Approximately one year before product release, it was
evident that it would take 9K to 10K bytes of microcode to complete
the 5355A. Our microprocessor board had two 24-pin sockets to
accommodate ROMs. With the ROMs available at the time, the
microcode could have been programmed into 8Kx8 and 2Kx8
ROMs, the filling up both ROMs. Fora slight incremental cost, the
2Kx8 ROM was replaced with a 4fCx8 ROM, making the total
available microcode space 12K bytes. The extra 2K bytes of ROM
became valuable space to accommodate corrections for the 8K x8
ROM microcode.
Unlike the 4Kx8 ROM, the 8Kx8 ROM could not be simulated
with neces single pin-for-pin-compatible EROM. It was therefore neces
sary to order a masked ROM eight months before product release.
In anticipation of future microcode corrections, any software
routine larger than 250 bytes was partitioned into relocatable sec
tors with coding taking up 250 bytes or less.
To make a sector relocatable, one or more jump instructions were
required. The jump instructions have a fixed location within the
© Copr. 1949-1998 Hewlett-Packard Co.
4Kx8 address space. Thus, whenever a correction and/or modifica
tion jump needed, the operand address code of the appropriate jump
instructions was changed from one pointing to 'the 8Kx8 HOM to
one pointing to some unused address space within the4K x8 EROM
(see example in Figs. 1 and 2 on Jine 15J. If necessary, an entire
sector be be redone. On the average, however, a correction will be
needed half-way into a sector. FoiJowing the corrected code, an
extra the instruction is added to get back to the usable part of the
sector in the original 8K x8 HOM (see Fig. 2 line 24 j. Assuming 2K
bytes of correction space and 250-byte relocatable sectors, one can
expect 16 corrections (125 bytes on the average) be/ore the memory
is filled, in the case of the 5355A, many routines were much
smaller than 250 bytes. Therefore, the 2K-byte correction space
was adequate for more than 16 corrections.
With this technique only one version of the 8K x8 masked ROM
was needed. When the time came to produce the instrument, the4K
bytes of software code could be implemented and shipped in either
a 4Kx8 EROM or a 4fCx8 ROM. The choice between the two was
determined by 1) how "final" the software appeared to be, and 2J
the cost tradeoff between the higher EROM unit price and the mask
charge of the ROM. For the 5355A, the break-even point between
EROMs and ROMs was one production run. It was therefore de
cided to build the first two production runs initially with EROMs.
Masked ROMs were ordered with sufficient lead time to allow a
last-minute replacement before shipping the instruments.
Routine Call
Correction ROM
Routine Call
ShouJd software bugs be discovered in the future, a new4fCx8
ROM would be ordered. In the event that the4K x8 ROM runs out of
correction space, then the complete set of ROMs will be required.
At that time, it will be beneficia] to clean out the correction space in
the 4K HOM by changing the jump operand addresses back to the
8KROM. Note that the example in Fig. 2 has been amply annotated
so that a correction can be spotted easily. .Removing the special
corrections months or years later is therefore a simple task.
Figs. 1 and 2 show an example of a correction in 6800 micro
processor code. The4Kx8 correction EROM is assigned addresses
$5000 to $5FFF and the 8K ROM is assigned addresses $6000 to
$7FFF. Note that the correction penalty here is 12 bytes in the
4Kx8 EROM plus about five microseconds f or the jump instruction
at the end of the correction. Redoing the entire routine would have
required 26 bytes.
-Ronald E. Felsenstem
C o r r e c t i o n
F i r m
ui o
Ot 1
0' 7
0' 4
08-12-79 FILE MAME : AREA
Firm ROM
X X f R A D I U S ,
X X f R i f D I U S
Y Y Ã - - R A D I U Ã ‡
X X < - R A D I U S * R A D I U S
Ú 034
33E 5 0
33F 15
340 39
Fig. 1 . (a) Relationship of the correction ROM to the firm ROM
for routines that have not been corrected, (b) An example in
6800 microcode.
Fig. 2. (a) Relationship of the correction ROM to the firm ROM
for a routine that has been corrected, (b) 6800 microcode
© Copr. 1949-1998 Hewlett-Packard Co.
HP Model 5355A Automatic Frequency Converter and
HP Models 5356A/B/C Frequency Converter Heads
Input Specifications
CW and Pulse Modes
5355A 0.4-1.6 GHz Input:
DAMAGE LEVEL: -24 dBm (Fuse in BNC Connector).
IMPEDANCE: 50!1 nominal.
SWR: <2.5: 1 typical.
INSERTION LOSS: <1 dB from 1.5 to 18 GHz.
IF OUT: OUT converted signal in range of 80-375 MHz available at 5355 rear panel IF OUT
connector. 0 dBm nominal level.
GATE OUT: 0 to - 1 volt detected IF signal used to drive 5345A external gate control input.
Width of gate out is approximately 30 ns less than rf burst width.
PULSE OUT: Detected IF signal: TTL levels; TTL low indicates signal present; +1 to 0V
typical into 50Ã1.
5355A: 3.75 kg (8 Ib, 4 oz) net.
5356A/B/C: 0.54 kg (1 Ib, 3 oz) net.
5356A/B/C DIMENSIONS: 27.4 mm x 138 mm x 56.5 mm (1.08 x 5.43 x 2.23 in). Cable
length: 1.68 metres (66 in).
PRICES Standard), U.S.A.: 5355A Automatic Frequency Converter Plug-In (HP-IB Standard),
5356A 18 GHz Frequency Converter Head, $1300.
Option 001 High Pass Filter, add $125.
5356B 26.5 GHz Frequency Converter Head, $1800.
Option 001 18-26.5 GHz Waveguide Input, add $600.
5356C 40 GHz Frequency Converter Head, $2400.
Option 001 26.5-40 GHz Waveguide Input, add $500.
5345A Electronic Counter, $4900.
Option 01 1 HP-IB, add $800.
Option 012 HP-IB (includes Programmable Trigger Level), add $1450.
5301 Stevens Creek Boulevard
Santa Clara, California 95050 U.S.A.
Operating Mode Specifications
CW Mode
Specifications warranted. the instrument's warranted performance. Typical or nominal performance characteristics provide useful application information but are not warranted.
© Copr. 1949-1998 Hewlett-Packard Co.
Pulse Mode
•rms jitter = (5345A Gate Time x Ext. Gate Width) ~'/2 +100 Hz
"Specifications apply only to external gating of 5345/5355.
Bulk Rate
U.S. Postage
Address Correction Requested
Hewlett-Packard Company, 1501 Page Mill
Road, Palo Alto, California 94304
APRIL 1980 Volume 31 • Number 4
Technical Information from the Laboratories of
Hewlett-Packard Company
Hewlett-Packard Company, 1501 Page Mill Road
Palo Alto, California 94304 U.S.A.
Hewlett-Packard Central Mailing Department
Van Heuven Goedhartlaan 121
1 1 80 AM Amstelveen The Netherlands
Yokogawa-Hewlett-Packard Ltd., Suginami-Ku
Tokyo 1 68 Japan
/"""» I I A K I /~N ^
\ ^ / n / \ I N \ J L
2 0 8 1 0
A f"~\ ["""v O I O O • To Chan9e y°ur address or delete your name from our mailing list please send us your old address label Send
/xl— Page I— 94304 Fi LOO . changes to Hewlett-Packard Journal, 1501 Page Mill Road, Palo Alto, "California 94304 USA Allow 60 days
© Copr. 1949-1998 Hewlett-Packard Co.
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