1982 , Volume , Issue Feb-1982

1982 , Volume , Issue Feb-1982
O 000 d
3 O
3 3 3 3
3 3 0
0 0 0 0
© Copr. 1949-1998 Hewlett-Packard Co.
Technical Information from the Laboratories of Hewlett-Packard Company
FEBRUARY 1982 Volume 33 • Number 2
A Broadband, Fully Programmable Microwave Sweep Oscillator, by Rolf Dalichow and
Douglas E. Fullmer The latest word in sweepers offers controllability, usability, wide sweeps,
high frequency accuracy, calibrated output power, and plug-in flexibility.
A New W. of Programmable Sweep Oscillator Plug-ins, by Gary W. Holmlund, Glenn
E. Elmore, and Duaine C. Wood One plug-in sweeps frojn 10 MHz to 26.5 GHz. There are
many others.
Portable C. for Cardiac Resuscitation, by Paul I. Bennett and Victor C.
Jones When a person's heart stops, this instrument can be set up quickly, measure its own
effectiveness, and provide permanent records of the procedure.
In this Issue:
Frequency response is an important characteristic of receivers, filters, amplifiers, stereo
systems, and nearly every other kind of electronic device. With a voltage of a given frequency
applied, what is the device's output voltage? How much current does it draw from the source?
I If you know designing a device, designing with a device, or testing a device, you need to know
.sqÉB « example, things. Typically, the answers vary with the applied frequency; for example, an amplifier
will have a different output for different input frequencies even if the input voltage doesn't vary.
A device a response is the same for all frequencies is said to be flat, meaning that it has a
flat or nonvarying frequency response.
People the to measure frequency responses point by point: connect an oscillator to the device, tune the
oscillator to some frequency, measure the device's response, write it down, tune the oscillator to another
frequency, and repeat as necessary. Sweep oscillators, or sweepers, freed us from this drudgery. These signal
sources tune themselves rapidly and automatically across a band of frequencies, making it possible to see a
device's able response on an oscilloscope or plot it using an X-Y recorder. A good sweeper has to be able
to sweep be frequency bands, depending on the application. Its own output voltage should be flat with
frequency and accurately known, and its output frequency should be accurate, too. Over the years, sweeper
performance in these areas has steadily improved as new technology has made higher performance possible.
Hewlett-Packard's latest microwave sweep oscillator, Model 8350A, is our cover subject this month. It sets
new standards for frequency coverage and output accuracy. It's highly versatile, first because it's
microprocessor-controlled, and second because it accepts a wide variety of plug-in modules for different
frequency ranges and power levels. A major contribution is its complete programmability, which makes it
suitable design use in automatic computer-controlled test systems. The article on page 3 describes the design of the
8350A mainframe and the article on page 1 1 discusses the 83500 Series plug-ins, a new group of highperformance plug-ins designed for the 8350A. Our cover photo shows the production test system for some
of the provides used in 83500 plug-ins; an 8350A Sweep Oscillator provides stimulus signals for the
When ineffective person's heart stops beating normally and begins the ineffective quiver known as fibrillation, quick
action usually necessary to save the victim's life. The treatment of choice is usually the application of a high-voltage
pulse heart. known victim's chest, causing a momentary current to pass through the heart. This is known as defibrillation, defibrillator. most device that delivers the high-voltage pulse is known as a defibrillator. A problem with most
defibrillators is that they don't tell the operator whether they've done their job, that is, how much energy they've
actually delivered to the heart. HP's Model 78660A Defibrillator/Monitor not only displays and records actual
delivered energy, but also records the date and time, the selected energy, the peak current during discharge,
and patient impedance. An alarm light signals the operator if the paddles aren't making good contact, and an
intelligent electrocardiograph records the patient's EGG waveform while rejecting noise and false beats. The
78660A medical all of this information automatically, providing valuable medical and legal documentation of the
procedure. Many of these features aren't found in the most sophisticated hospital defibrillators. Yet the 78660A
is a compact, lightweight portable. The design story begins on page 22.
-R. P. Do/an
Editor Photographer. P Dolan . Associate Editor. Kenneth A Shaw • Art Director. Photographer. Arvid A Damelson • Illustrator. Nancy S Vanderbloom
Administrative Services. Typography. Anne S LoPresti. Susan E Wright • European Production Supervisor. Henk Van Lammeren
©Hewlett-Packard Company 1982 Printed in U.S.A.
© Copr. 1949-1998 Hewlett-Packard Co.
A Broadband, Fully Programmable
Microwave Sweep Oscillator
Nearly thirty RF and microwave plug-in modules are
available to tailor this high-performance swept signal
source to a wide range of applications in the frequency
range from 10 MHz to 26.5 GHz.
by Rolf Dalichow and Douglas E. Fullmer
IT IS SOMETIMES DIFFICULT to remember that there
was a time when the microwave designer didn't have the
help of a sweep oscillator and had to make measure
ments point-by-point. Once frequency tuning on a swept
basis became available with the first mechanically driven
sweepers, improvements came in a steady flow: electronic
tuning of backward-wave oscillators (BWOs), solid-state
sweepers using YIG-tuned oscillators, and broadband
coverage by means of multiple oscillators or frequency mul
tipliers and YIG-tuned filters. Now a microprocessorcontrolled, fully programmable sweeper, the HP Model
8350A Sweep Oscillator, Fig. 1, represents the state of the
art in this field.
In addition to controllability and usability, the 8350A
makes other substantial contributions through innovative
RF and microwave design. These contributions are in areas
considered important by microwave designers, such as fre
quency coverage — 0.01 to 26.5 GHz in one plug-in
(83595A), frequency accuracy — 10 MHz at 20 GHz, and
calibrated output power. Internal leveling holds output
power constant within ±0.25 dB with a 2-GHz plug-in and
±1 dB with a 26.5-GHz plug-in (Fig. 2).
In the design of the 8350A, strong emphasis was placed
on product continuity and the use of well proven design
concepts, with advances made where significant improve
ments could be achieved. For example, the plug-in concept
of earlier HP sweep oscillators has been retained (Fig. 3).
*YIG 1 yttrium-iron-garnet. a lerrite material. See reference 1 lor an explanation of YIG tuning.
The 8350A mainframe accepts high-performance 83500
Series plug-ins designed especially for it (see article, page
11). It also accepts older 86200 Series plug-ins with the
11869A Adapter, and gives the user nearly complete programmability of these plug-ins. As advances in RF, mi
crowave, and millimetre-wave technology become avail
able, it will be easy to upgrade an 8350A measurement
system with a new plug-in. The basic control, program
ming, and power supply circuits reside in the mainframe
while the circuits with the highest potential for improve
ment reside in the plug-in.
Great care was taken to make the mainframe/plug-in in
terface reliable and supportive of the performance of the
instrument. Important power supplies use remote sensing,
and the tuning and modulation voltages are connected via
shielded cables. The firmware controlling the plug-in func
tions resides in the plug-in, so the system firmware is au
tomatically updated when a plug-in with additional
capabilities and sophistication is inserted.
An Approachable Instrument
For ease of use, the front panel is organized into blocks of
associated functions (Fig. 1). Major parameters, such as
start and stop frequencies, sweep time, and power level,
have their own displays and knobs. The sensitivity of each
knob adapts to its speed of rotation and to the range of
values being modified. Parameter values can also be in
cremented or decremented with the step keys or entered on
a anno
a aaa«
a nnn n
FI nnn n
Fig.1. Model 8350 A Sweep Oscil
lator is a fully programmable in
strument that can accept various
RF and microwave plug-in mod
ules to provide broadband fre
quency sweeps (0.01 to 26.5 GHz
with the 83595A RF Plug-in
shown), calibrated output power,
and high frequency accuracy
(±10 MHz at 20 GHz).
© Copr. 1949-1998 Hewlett-Packard Co.
26.5 GHz can be determined within ±100 kHz.
Up to nine front-panel settings can be saved and recalled,
either singly or in sequence. The memory is nonvolatile, so
stored front-panel settings are retained even with the in
strument turned off. A footswitch may be connected to a
rear-panel connector to sequence settings in a hands-free
production setup.
Device response in two independent frequency ranges
may be observed "simultaneously" by using the alternate
sweep feature (Fig. 6). At the end of each sweep, the 8350A
switches between internal settings corresponding to two
different front-panel settings. This allows, for example, ad
justment of the passband of a filter while concurrently ob
serving the effect on the filter's rejection band.
Second-Generation HP-IB Compatibility
Fig. 2. Internal leveling keeps 8350A/83595A output power
constant within ±1 dB to 26.5 GHz.
the numerical keyboard.
Functions are designed for ease of use. For example
START/STOP and CF/AF aren't separate sets of sweep fre
quencies, but just two ways of displaying and entering the
same sweep. Pressing CF or AF after the start and stop
frequencies are entered displays the equivalent center
frequency and span and allows modification of these
Five independent markers can be set to frequencies of
interest and used to simplify microwave measurements. For
example, the marker difference mode aids bandwidth mea
surements by displaying the difference in frequency be
tween any two markers. The trace between the two markers
is intensified (Fig. 4).
A second function, marker sweep, sets the endpoints of
the sweep to marker 1 and marker 2 (Fig. 5). The resulting
display is a more detailed view of some area of interest, such
as the passband of a filter. A related function, marker to
center frequency, positions the sweep about some point of
interest identified by the marker.
Lastly, the sweep can be stopped momentarily at a marker
while the frequency is measured by the HP 5 343 A Mi
crowave Counter. Thus even while sweeping, a marker at
All 8350A front-panel functions are programmable via
the HP-IB, a standard feature. The programming language
uses easy-to-remember mnemonics and a straightforward
numerical format (e.g., CW 7.55 GZ).
The HP-IB interface not only lets the user program every
front-panel function, but also provides a number of
special-purpose functions to aid remote operation. For
example, a learn mode transfers the complete front-panel
setting into the memory of a controller, where it can be
stored on tape and later used to program the 8350A. This
allows multiple manually entered test setups to be saved
and recalled by means of a generalized program in the
To give the user the ability to interact with the instrument
in an automated test system or to document the present
value of a parameter, the 8350A allows interrogation of any
numeric function. For example, OPCF (output center fre
quency) outputs the center frequency to the controller in a
human-readable form.
An HP-IB service request (SRQ) can alert the controller to
a number of conditions in the 8350A. SRQs can be issued
when a sweep ends, the RF output becomes unleveled, the
airflow ceases, a self-test fails, an HP-IB syntax error is
detected, or a numerical entry is out of bounds.
Mainframe Design
As a completely programmable mainframe for two
«The 488-1978. Interface Bus is HP's implementation of IEEE Standard 488-1978.
Fig. 3. Two types of plug-ins can
be used in the 8350/4 Sweep Oscil
lator mainframe. The 83500 Series
(right) is a group of highperformance plug-ins designed
specifically for the 8350A. Existing
86200 Series plug-ins (left) can be
used with the 77869/4 Plug-in
Adapter (rear portion of unit at left).
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 4. Five independent, continuously variable markers iden
tify significant frequencies. The marker difference function
simplifies frequency bandwidth measurements as shown
here. The frequency difference between any two of the five
markers is displayed.
families of RF and microwave plug-ins, the 8350A is de
signed to meet a number of potentially conflicting objec
• Compatibility with existing and future plug-ins. The
8350A can be used with both the high-performance
83500 plug-ins and, through an adapter, the 86200 plugins. Generous power supply capacity is provided for fu
ture plug-ins.
• Controllability and usability via powerful but easy-to-use
front-panel and HP-IB functions.
• Maximum performance with a minimum number of
components. For example, one microprocessor controls
both mainframe and plug-in.
• High reliability and serviceability. Generous airflow and
extensive self-test routines are key contributions in this
Internally, the 8350A mainframe is organized into three
sections (see Fig. 7): the instrument control section, the
sweep, tuning and marker section, and the power supply
The microcomputer controls the sweep, tuning and
marker section and interfaces with the front panel, the
HP-IB, and the plug-in. To produce the tuning voltage for
the plug-in, a fixed-amplitude ramp from the sweep
generator is scaled and offset. Markers are generated by
comparing the output of a DAC (digital-to-analog convert
er) to the sweep ramp. Voltages from regulated supplies in
the mainframe are also used in the plug-in.
Instrument Control
One microprocessor controls both mainframe and plugin. An HP-IB preprocessor integrated circuit simplifies the
HP-IB interface. To reduce board count and increase relia
bility, the microprocessor, 2K bytes of CMOS RAM (com
plementary metal-oxide semiconductor random-access
Fig. 5. Marker sweep is useful for examining narrow re
sponses in more detail. The 8350A sweeps only the span
between markers 1 and 2.
memory), and 32K bytes of read-only memory organized as
four 64K-byte ROMs, are all on one board. A nickelcadmium battery and an on-board charging circuit ensure
nonvolatile storage of the instrument state in the CMOS
Instrument self-tests pinpoint failed components via
either the numerical display or, if the front panel is inopera
tive, light-emitting diodes on the microcomputer board (see
box, page 10).
Clock signals for both mainframe and plug-in are gener
ated by a state machine. The state machine design insures
that the address and data lines stay valid for 93 ns after the
interface clock changes state, thus allowing ample time for
address decoding.
Plug-in Interface
The 83500 Series plug-ins interface to the 8350A via a
dedicated bus and an extensive software protocol. Pro
grams in the plug-in's ROM are executed by the micro
processor in the mainframe. An address space of 10K bytes
is provided for the plug-in: 8K bytes for firmware and 2K
bytes for I/O.
© Copr. 1949-1998 Hewlett-Packard Co.
time processing is initiated is less than 16 ms.
The count is proportional to the rate of rotation, so it can
be used to index a table of rate-dependent knob sen
sitivities. The values in the table are chosen such that rapid
rotation causes the entire range of the parameter to be cov
ered in a few turns, while slow rotation results in fine
adjustment. RPG and display resolutions are linked to
hardware resolution, and display digits always increase or
decrease monotonically when an RPG is turned.
Sweep Generation
A current-output DAG, a retrace current source, and an
integrator are key components in the generation of the
sweep waveform (Fig. 8). When the output of the integrator,
a linear ramp, reaches +10.8 volts the retrace current source
is switched on, forcing the ramp to reverse direction (Fig.
9). When the ramp falls to -0.9 volts, the retrace current
source is turned off and the output of the integrator again
ramps upward, completing the cycle. The sweep may be
stopped by the plug-in (during bandcross, for example), or
by an external input.
Sweep time is set by both the integrator capacitor and the
DAG current. The 10,000:1 range of 10 ms to 100 seconds is
covered by switching between two capacitors, 100:1 in
value, and adjusting the DAG current over a 100:1 range.
When the sweep is stopped, the output of the operational
amplifier jumps in voltage as the current through the on
resistance of the analog switches is suddenly cut off. This
voltage jump is eliminated by taking the output of the
integrator directly from the capacitor via a high-impedence
Upper and lower ends of the sweep are clamped to preci
sion oc levels. YIG-tuned oscillator (YTO) settling oc
curs during the time the output of the integrator is clamped
to zero volts. The lower clamp eases YTO delay compensa
tion by starting the sweep with a maximally smooth func
Forward sweep starts when the output of the integrator is
—0.3 volts and ends when the upper clamp turns on at 10
volts. RF or Z-axis blanking is activated when not in for
ward sweep.
Fig. 6. The alternate sweep function causes the 8350A to
switch, on successive sweeps, between the current frontpanel settings and any of nine front-panel states stored in
memory. This makes it possible to observe a device's re
sponses to two different stimulus conditions on the same dis
In addition to data and address lines, the plug-in interface
bus contains flag, interrupt, and hardwired control lines.
Special attention was paid to minimizing bus transients
that could cause errors on the bus lines. Ringing and
crosstalk are reduced by reverse-terminating each bus
driver with a 100Ã1 resistor. To reduce crosstalk further,
each group of bus lines is separated from other groups by at
least one ground line. Schmitt-trigger gates are used to
sense the received signals.
Front-Panel Interfacing
Interrupts are generated by the front panel at a 1.6-kHz
rate. The microprocessor refreshes one of the 15 display
digits each time an interrupt is received, thus refreshing the
entire front panel at a rate of about 100 Hz. Other front-panel
tasks are performed at submúltiplos of the 1.6-kHz rate.
These tasks and others described below take only a small
percentage of the microprocessor's time.
The front-panel knobs function as rotary pulse generators
(RPGs). In addition to scanning and debouncing the
keyboard, the interrupt routine reads and resets the RPG
counters at a 60-Hz rate. Counter overflow isn't possible at
this rate, and the delay from the time a knob is turned to the
Sweep Scaling
To produce the tuning voltage, the sweep waveform is
Sweep Ramp
Instrument Control Section
Sweep. Tuning, and Marker Section
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 7. Block diagram of the
8350A Sweep Oscillator.
A System-Oriented Instrument
The 8350A was designed to be used in systems. Therefore, all
inputs and outputs necessary to synchronize and operate test
systems are provided on BNC and multipin connectors. Both
hardware and firmware were designed to make system applica
tions of the instrument possible.
A simple example of an 8350A system is a master/slave config
uration using two 8350As, making it possible to test mixers and
receivers very efficiently. The frequency offset feature of the
8350A simplifies the setting of the two instruments' frequencies.
There is a powerful synergism between the HP 8350A, the HP
5343A Microwave Counter, and the HP 5344A Source Syn
chronizer. A two-wire interface allows measurement of start, stop,
and marker frequencies with counter accuracy. The counter can
receive the RF energy from the auxiliary output of the broadband
plug-ins, thus eliminating any output power reduction caused by
power splitters or couplers. With the HP 5344A and the HP 5343A,
the sweeper can be part of a phase-locked system that allows CW
and swept frequencies to be set with counter accuracy. Sweeps
of up to 40 MHz are completely phase-locked, while wider sweeps
use the "lock and roll" principle. In this mode, the start frequency
is phase-locked and then, after achieving high accuracy, the
frequency is swept in an analog fashion. All of the frequency
controls are handled by the HP 5344A, which acts as an HP-IB
Another measurement system employing the 8350A is the
8755P Automatic Scalar Network Analyzer. This system is con
trolled by an HP-85 Computer and operates from 0.04 to 1 8 GHz.
Its amplitude resolution at 10dB per division is 0.1 dB over the full
range of the screen and 0.02 dB over a 40-dB range at the center
of the screen.
The most economical way to perform phase and magnitude
measurements at microwave frequencies is to use the HP 8408A
Automatic Network Analyzer, Fig. 1, which also employs an HP
8350A as a source. This analyzer has a range from 0.05 to 1 8 GHz
scaled by the AF DAC and the AF attenuators, then summed
with the CW and vernier DACs (Fig. 10). Attenuating the
output of the 10-bit AF DAC expands its 1,000:1 range to
64,000:1. The eight-bit vernier DAC covers ±2 leastsignificant bits of the 12-bit CW DAC, thus resolving
262,000 points per plug-in band.
A precision voltage reference, used in sweep generation
and scaling, reduces noise by averaging the voltage of three
Zener diodes. Since the noise contributions from the indi
vidual diodes are uncorrelated, the reference is only 58% as
noisy as a single Zener diode.
Marker Generation
Markers are generated by comparing the sweep ramp to
the output of a DAC (Fig. 11). Each time the comparator
fires, a counter advances the address of the marker RAM,
setting the DAC to the next marker-on or marker-off value.
When the marker shown in the front-panel display is
reached, a digital comparator turns on the active marker
line. The active marker has enhanced brightness and the
frequency at which it occurs can be measured with the
5343A Frequency Counter.
Since the marker is toggled each time the comparator
fires, the RAM holds pairs of marker-on and marker-off
positions. Marker positions are calculated from the start,
Fig. 1 . 8408 A Automatic Network Analyzer.
and an effective directivity greater than 40 dB.
Software available for the HP-85 Computer allows vector
error-corrected measurements that eliminate all degradation of
system performance because of imperfect adapters and cables.
These nonideal but necessary pieces of hardware used to adapt
the device under test to the measurement system normally drasti
cally impair the performance of any otherwise well designed
network analyzer.
Finally, the 8350A is used in the HP 8409C Automatic Network
Analyzer. Besides the new sweeper, this extension of the wellknown network analyzer family incorporates an HP 9845B Option
250 Desktop Computer and new accuracy enhancement
software. The interface between the 8409C system and the
sweeper is simplified and the software needed to program the
source is greatly streamlined.
stop, and marker frequencies, sorted, and then loaded into
the marker RAM before the next sweep starts. The markers
are normally one bit wide, 0.4% of the display screen. In
marker difference mode, the entire trace between two mark
ers is intensified by eliminating the first marker's off entry
and the second marker's on entry.
Counter Interface
During the sweep, start, stop, or marker frequencies can
be measured on the 5343A Frequency Counter. A userselected pulse at the start of the sweep, the end of the sweep,
or the active marker triggers the counter. The counter re
sponds by stopping the sweep and counting the frequency,
then releasing the sweep when the measurement is com
plete. While the frequency is being counted, the RF display
is blanked and the internal 27-kHz modulation used by the
8755C Scalar Network Analyzer is disabled.
Other Interfaces
When an 8755C Scalar Network Analyzer is connected to
an 8350A in alternate sweep mode, interface lines force the
8755C's channel 1 to display the normal sweep and channel
2 to display the background sweep. Separate sensitivities
can be used for the two channels so that, for example, a
broadband sweep at 10 dB per division can be alternated
© Copr. 1949-1998 Hewlett-Packard Co.
+ 5V
Stop Sweep
Lower Limit
Fig. 8. 8350x1 sweep generation circuits.
with a narrow sweep at 1 dB per division.
External sweep capability, along with the offset function,
allows two 8350As to test mixers and receiver front ends.
One sweeper acts as a master and externally sweeps the
slave, which is offset in frequency by the desired inter
mediate frequency. The 8350A can also be used as an upconverter simulator for HP Microwave Link Analyzers such
as the HP 3712A.
Rear-panel interface lines allow the 8350A to lift the pen
or mute the response of an X-Y recorder, trigger an HP
8410C Microwave Network Analyzer, or indicate that the
sweep is retracing. Other lines allow external instruments
to trigger the sweep, create a marker, or blank the RF. Digital
ground and a +5V output with 100 mA capacity are pro
vided for custom interface circuitry.
Power Supplies
To handle the 86200 Series plug-ins, the 8350A provides
+ 20V, -10V, and -40V, all remotely sensed and therefore
unaffected by the voltage drop across wiring and connector
resistance. The 8350A and the new 83500 Series plug-ins
share a ±15V supply, but have separate high-current +5V
supplies. Microcircuits in the new plug-ins also use +10V
and unregulated +5V. All of the supplies are currentlimited. Crowbars protect both plug-in and mainframe from
expensive damage caused by power supply failure. If a
monitor circuit detects that a supply has failed, an error
message is written to the front-panel display.
Reliability of the 8350A power supplies is improved by a
+ 10. 8V
+ 10V
Sweep ov ^
(Stop Sweep)
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 9. Sweep waveforms generated by the circuit of Fig. 8.
+ 10V.
A F Attenuator
Fig. 1 0. Sweep scaling and offset
circuits condition the sweep ramp
before it is sent to the plug-in.
large heat sink. If the temperature of the heat sink exceeds a
safe limit, a thermal switch removes line power.
A generous flow of air minimizes temperature rise and
further improves the reliability of both mainframe and
plug-in. The front panel displays an error number if the
airflow is impaired. Decreased airflow because of a clogged
filter or blocked ventilation allows a heated transistor to
increase in temperature, but has no affect on an unheated
one. An airflow-monitoring comparator turns on, indicat
ing insufficient airflow, when the forward voltages and
hence the temperatures of a heated transistor and an un
heated transistor differ by more than a preset limit (Fig. 12).
We would like to acknowledge the efforts and contribu
tions of the many people who so effectively worked to-
gether to create the 8350A. The original concept was pro
vided by Jack Du pré and received its initial support from Rit
Keiter. Arlen Dethlefsen and Irv Hawley gave us the en
couragement and support needed to successfully complete
the project. Special thanks go to R.C. Ho for designing the
power supply, the sweep generator and the HP-IB interface.
The product design was well executed by Bill Misson and
Bob F. Schaefer. Bill McDonald designed the self-test capa
bility. Diane Heggie and Roy Church provided the indus
trial design effort. Gary Dooley and Jay Gregerson helped
during the final design phase of the mainframe and the
adapter and arranged a smooth transfer into production.
1. P.R. Hernday and C. Enlow, "A High-Performance 2-to-18-GHz
Sweeper," Hewlett-Packard Journal, March 1975.
Fig. 11. Markers are generated
by comparing the sweep ramp to
the output ofaDAC that has been
loaded with the next marker fre
quency from a 16-byte RAM.
© Copr. 1949-1998 Hewlett-Packard Co.
8350A Self-Test Capabilities
Microprocessor-controlled instruments present both difficulties
and advantages in the areas of repair and maintenance. On one
hand, a single defective component may render an instrument
totally inoperative. On the other hand, self-tests can detect faulty
components or sections of circuitry and operator-initiated tests
can pinpoint the specific problem. Built-in test capabilities thus
increase the usefulness of the instrument by building confidence
and shortening the repair cycle.
When the 8350A is switched on or when the PRESET key is
pressed, routines testthe microprocessor, RAM, ROM, I/O buses,
power supply, DACs and plug-in. If either the microcomputer
board or the front panel fails a test, four LEDs on the
microcomputer board indicate, in binary code, which component
is defective. If RAM or ROM tests fail, the test program goes into
an endless loop and does not proceed to routines that rely on the
defective components.
If the microcomputer and front-panel tests pass, normal instru
ment operation resumes and error numbers are displayed on the
front panel as well as on the error-indicator LEDs.
Each byte in RAM is tested without altering its contents. The
byte is read, complemented and written back, then read again
and EXCLUSiVE-ORed with the original. If the result is all ones, the
byte is not defective and its original contents are restored. ROMs
are tested by comparing a calculated checksum to a reference
checksum in the self-test ROM.
The front-panel and instrument buses are tested by writing
patterns into a latch, then reading and verifying the contents of the
latch. Address lines on the instrument bus are tested by writing
out a pair of complementary address patterns. A counter on the
bus is incremented and decremented by the decoded patterns,
after which the count is read and verified.
A checksum is performed on the plug-in's ROM. If the test isn't
passed, the 8350A indicates a plug-in failure and defaults to an
emulation of a O-to-10-GHz plug-in.
Fig. 1 2. Airflow sensor detects decreased airflow by sensing
a temperature difference between a heated transistor and an
unheated one.
Douglas E. Fullmer
Doug Fullmer graduated from the Uni
versity of Washington with a BSEE de
gree and joined HP in 1975. He has
contributed to the design of several
86200 Series sweeper plug-ins, had
overall responsibility for 8350A
firmware, and designed the 8350A mi
crocomputer and front-panel boards. In
June he will receive his MSEE degree
from Stanford University. Born in Seat| tie, Washington, Doug is married and
£ lives in Santa Rosa, California. His leiy sure activities include gardening, hik'.•{ ing, swimming, home projects, and
/•'. learning about history, anthropology.
£*•' and animal behavior.
Analog Tests
If the power supply voltages are found to be within limits, the
DACs are checked by comparing their outputs. An analog switch
allows the marker DAC to be compared to both the CW and
sweep-time DACs.
During normal instrument operation, the power supply and the
airflow are tested at ten-minute intervals.
User-Initiated Tests
Special tests can be run to isolate faulty components. If the
self-test of the microprocessor fails or if faulty address decoding
is suspected, the microprocessor can be forced to free-run. If the
DAC tests fail, a rotating-1 exercise, specific to each DAC, allows
a quick check using an oscilloscope.
For general testing, the user may read from or write to any RAM,
ROM, be I/O location. For example, a rotating-1 pattern may be
written to some location, such as a latch. An oscilloscope can then
be used to trace the pattern from the microcomputer board
through the bus and data buffer to the selected latch and any
circuitry it drives.
-Bill McDonald
Rolf Dalichow
Rolf Dalichow was born in
Braunschweig, Germany and received
his Ing. Grad. degree from the Staatliche Ingenieurschule Giessen. He
joined HP in 1 973 with 1 2 years' experi
ence in hybrid RF circuits and com
munications transmitters in Berlin and
the U.S.A. He has served as project
manager for the 8350A Sweep Oscil
lator and as a project leader for the
8505A Network Analyzer. He is named
as a co-inventor on a frequency count
er patent. Rolf is married, has two
children, and lives in Santa Rosa,
California. He enjoys backpacking and
is in the startup phase of building a new house for himself and
his family.
© Copr. 1949-1998 Hewlett-Packard Co.
A New Series of Programmable Sweep
Oscillator Plug-ins
by Gary W. Holmiund. Glenn E. Elmore, and Duaine C. Wood
AS TEST SYSTEMS have become more automated
to increase their efficiency, a need has arisen for a
new generation of swept signal sources that are
completely programmable. The 83500 Series plug-ins for
the 8350A Sweep Oscillator are designed to meet that
In the design of a new series of plug-in swept signal
sources, several objectives needed to be met:
Programmability of all plug-in functions
Wide frequency coverage
Improved frequency accuracy
• Calibrated output power.
Table I lists the plug-ins in the 83500 Series along with
their frequency ranges and power output levels. Fig. 1
shows representative plug-ins in the series and Fig. 2 shows
the general block diagram of an 83500 Series plug-in. The
front panel of each plug-in contains controls that affect RF
output power and the power leveling mode. All front-panel
settings can be saved and recalled in any of the 8350A's
nine save/recall registers. Any two settings can be alter
nated in the alternate sweep mode. Each plug-in has a
keyboard/display processor that refreshes the front-panel
power level display and enters keys as they are pushed.
Table I
On most plug-ins, the calibrated power level is adjustable
over a 15-dB range. It can be set by keyboard entry, rotary
pulse generator, step up and down keys, and the HP-IB.
With the addition of an optional 70-dB attenuator (55 dB on
the 83594A and 83595A) the power level can be adjusted
over an 85-dB range.
A slope adjustment allows the user to compensate for
attenuation in cables, connectors or other hardware that
increases with frequency. The adjustment causes the plugin to increase power output as frequency increases. This
feature can also be used with test sets in network analyzers
so that the device under test has nearly constant incident
power in spite of frequency-dependent losses between it
and the sweeper. This is particularly important when
measuring active devices at higher power levels.
The power sweep mode is useful for measuring the com
pression points of amplifiers. It is used to sweep the power
level at a single frequency to measure the gain of an
amplifier as a function of input power level. When used
with the 8755C Scalar Network Analyzer, the horizonal
scale is calibrated in input power level to the device under
test. of vertical scale can be adjusted to display the gain of
the device. The 1-dB compression point can then be read
directly from the display. The step keys may be used to step
the measurement frequency up or down so that compres
sion level information can be measured as a function of
The user also has the ability to alternate the power be
tween two different levels as the oscillator sweeps. This
feature can also be used to measure an amplifier's perfor
mance. For example, one sweep can be set to measure the
small signal gain of the amplifier while the power level in
the second sweep is adjusted to a given compression point.
This allows simultaneous observation of the small and
large-signal performance.
•HP-IB is Hewlett-Packard's implementation of IEEE Standard 488-1978.
3 H
Fig. the . from Series plug-ins for the 8350 A Sweep Oscillator span the frequency range from 0.01
to 26.5 GHz.
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 2. General block diagram of
an 83500 Series plug-in.
The automatic level control (ALC) has three modes of
operation. Internal leveling uses the internal coupler and
detector to measure the power level at the output of the
instrument and hold it at a specified power level. The ex
ternal leveling mode is similar but uses an external coupler
and detector to measure the output power. The external CAL
adjustment on the front panel adjusts the power level out of
the coupler to agree with the power displayed by the plugin. The power meter leveling mode uses an HP 432A/B/C
Power Meter to measure and control the output power.
RF Designs
Fig. 3 is the RF block diagram of the 83525A plug-in. This
plug-in is a combination of the 83522A (0.01-2.4 GHz) and
the 83540A (2-8.4 GHz). The high band of the 83525A is
formed by the YIG-tuned oscillator, the modulator/
amplifier, two PIN diode switches, and a coupler/detector.
If the two PIN diode switches are removed, the remaining
components form the RF block for the 83540A. The YIGtuned oscillator, PIN diode switch, modulator-mixer,
amplifier, coupler/detector, and PIN diode switch form the
low-band block. If the PIN diode switches are removed the
83522A RF block diagram remains. When sweeping the full
frequency range, the 83525A first sweeps the low band from
0.01 to 2.05 GHz. Then the PIN diode switches change to the
high band, which sweeps from 2.05 to 8.4 GHz.
The block diagram of the 83570A (18-26.5 GHz) plug-in is
shown in Fig. 4. The YIG-tuned oscillator operates over a
frequency range of 9.0 to 13.25 GHz and drives the fre
quency doubler (see page 17), which delivers 10 mW out
put power over the 18-to-26.5-GHz range. The output port is
a WR 42 waveguide which provides a reliable connection in
this frequency range and assists in the reduction of funda
mental feedthrough. An auxiliary output allows the use of
equipment in the 9-to-13.25-GHz range, thus eliminating,
for example, the need of an 18-to-26.5-GHz frequency
counter. The output power is internally leveled and is ad
justable and HP-IB programmable over a 10-dB range.
The 83545A plug-in covers the 5.9-to-12.4-GHz range. It
consists of a FET YIG-tuned oscillator and a power
amplifier capable of delivering 50 mW of output power. An
optional step attenuator together with the ALC circuitry
provides a calibrated output power range of 84 dB.
Fig. 5 is the RF block diagram of the 83595A (0.01-26.5
Fig. This combination diagram of the 83525A 0.01-to-8.4-GHz plug-in. This plug-in is a combination of the
83522A 0.01-to-2.4-GHz plug-in and the 83540A 2-to-8.4-GHz plug-in.
© Copr. 1949-1998 Hewlett-Packard Co.
Microprocessor Control of Plug-ins
18-26.5 GHz
+ 10 dBm
9-13.25 GHz
Fig. 4. Block diagram of the 83570/4 1 8-to-26.5-GHz plug-in.
GHz). The top of the diagram shows the low band (0.01-2.4
GHz). It is very similar to the low band of the 83525A except
that the modulator-splitter replaces one PIN diode switch
and the switched YIG-tuned multiplier (SYTM) replaces
the other PIN diode switch. In the high bands the RF from
the YIG-tuned oscillator goes through the modulator/
splitter to the power amplifier. It is amplified to 500 mil
liwatts and goes to the high-band port of the SYTM. The
SYTM generates harmonics of the signal and tunes out all
but the desired harmonic (see page 15). The 83594A (2-26.5
GHz) is made by deleting the low-band components and
changing the modulator-splitter to a simple modulator. The
83592 A (0.01-20 GHz) and 83590A (2-20 GHz) are similar to
the 83595A and 83594A, respectively, but with the RF
components operating only to 20 GHz.
To the microprocessor, which is in the 8350A mainframe,
the plug-in looks like auxiliary read-only memory and
input'output space. Address, data, and control lines are
brought from the mainframe to the plug-in. Addresses are
decoded in the mainframe so that 8K bytes of ROM and 2 K
bytes of I/O are available in the plug-in. Fig. 6 is a block
diagram of the digital control section of the plug-ins.
The sweep oscillator interface consists of buffers that
receive the address, data, and control signals from the main
frame. These signals are sent to ROM, interrupt control, or
the plug-in interface depending on the address. If the sig
nals are sent to the plug-in interface they go to other control
boards of the plug-in (i.e., front panel, YIG control, fre
quency modulation, or automatic power level control
The interrupt control section can generate interrupts to
get the processor's attention. The interrupts can come from
other boards of the plug-in or from timers within the inter
rupt control block. Interrupts that come from other boards
generally occur at the beginning or end of a sweep or at the
bandswitch of a multiple-band plug-in. At bandswitch, fre
quency gains and offsets are changed to cause the plug-in to
sweep the next band. The timers are used only on the
multiple-band plug-ins. After the band number is switched
during a sequential sweep, the timers are used to control the
time spent waiting for the frequency to settle. The 200-kHz
clock drives the timers and the keyboard/display processor
on the front panel.
Oscillator f>\,)
3.8 GHz
0.01-2.4 GHz
ALC Circuits
Band 0
ALC Circuits
PIN Switch Drive
ALC Circuits
Band 1-4 Amplifier
2.3-7 GHz
Directional Coupler/Detector
0.01-26.5 GHz
2.3-7 GHz
0.01 -26.5 GHz
ALC Circuits
Fig. The 2-to-26.5-GHz diagram of the 83595A 0.01-to-26.5-GHz plug-in. The 83594 A 2-to-26.5-GHz
plug-in is made by deleting the low-frequency section. The 83592A (0.01-20 GHz) and the
83590A GHz. GHz) are similar, but use RF components designed to operate to only 20 GHz.
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 6. Digital control section of
the 83500 Series plug-ins.
The configuration switch is a set of eight switches that
can be used to control the state of the plug-in at power-up.
Among other things, the switches tell the firmware which
model plug-in is present, whether the attentuator option is
present, and the FM sensitivity.
The plug-in firmware contains the routines to control the
plug-in from front-panel inputs and HP-IB commands. An
HP-IB controller can read power level, slope, and power
(continued on page 16)
Fig. 7. 83500 YIG driver converts
the tuning voltage received from
the mainframe to the correct mag
net current for the plug-in' s YIGtuned oscillator or multiplier.
Offset Adjust
© Copr. 1949-1998 Hewlett-Packard Co.
A Switched YIG-Tuned Multiplier
Covering 0.01 to 26.5 GHz
A new frequency multiplier capable of producing output
frequencies from 0.01 to 26.5 GHz was developed for the
wideband 83590-Series plug-ins for the 8350A Sweep Oscillator.
The new multiplier is tuned by means of a YIG (yttrium-iron-garnet)
sphere and is called a switched YIG-tuned multiplier, or SYTM. It
has two inputs: 0.01 to 2.4 GHz, which is switched to the output
bypassing the YIG filter, and 2.4-7.0 GHz, which is multiplied by 1 ,
2, 3 and 4, filtered, and summed into the same output path.
Because of the wide tuning range, the electrical, mechanical, and
magnetic designs had to overcome significant challenges to
maintain low cost and high production yields, yet achieve
state-of-the-art performance.
The SYTM is constructed on a single 0.25-mm sapphire
substrate containing bias circuitry, impedance matching circuits,
a step (see diode, the YIG filter, and a PIN diode switch (see
Figs. a and 2). Substrate temperature is maintained at 85°C with a
heater control loop consisting of two load resistors and a
thermistor sensor on the substrate. Band 0 input is applied to port
1 , which is connected to the output coax by means of the output
coupling loop of the YIG filter. During Band 0 operation, the PIN
diode is reverse biased, thus minimizing losses. Typical insertion
loss in this band is less than 1 dB.
In the frequency range 2.0 to 26.5 GHz, a maximum input of
+25.5dBm is applied to port 2 of the SYTM. In band 1, this signal
passes through the impulse generator including the step recovery
diode with little attenuation, since the diode is heavily forward
biased. As the input sweeps from 2.4 to 7 GHz, the YIG filter is
tuned to the same frequency, thereby coupling the input signal to
the output structure. The PIN diode switch is also turned on,
providing the ground return for the output coupling loop of the YIG
For bands 2, 3 and 4 the step recovery diode is actively biased
so that during one half of the input RF cycle the diode depletion
layer stores charge. As the input polarity reverses, the diode
0.01-2.6S GHz
Port 3
Diode Multiplier
Port 2
Fig. 2. Equivalent circuit of the SYTM.
current reverses, sweeping out the stored energy. By adjusting
the bias so that the remainder of the energy is depleted at the
maximum of the input current, an impulse rich in harmonics is
produced. The YIG filter is tuned to the desired harmonic and
couples that frequency to the output.
Tuning the filter to 26.5 GHz required a new magnetic design,
capable of producing a 10,000-gauss magnetic field in the
1 ,27-mm air gap containing the YIG sphere. The magnet pole tips
are made of two materials, one of high permeability and the other
capable of handling high flux densities without saturation. By
adjusting the ratio of the materials, the temperature coefficients
are minimized while reducing the hysteresis to approximately 15
MHz for a ±12-GHz tuning change. A linearity error of less than
3% between bands is achieved.
The YIG filter is a single 0.66-mm sphere mounted near the
temperature-compensated axis, between the (100) and (110)
axes.1 This position was found empirically to give the best
temperature stability with satisfactory insertion loss. Typical filter
1-dB bandwidth is 25 MHz from 2 to 20 GHz and 35 MHz from 20
to 26.5 GHz. Passband modes are less than 0.5 dB when tuned
over the complete frequency range. Typical suppression of
unwanted harmonics is 35 dBc.
The majority of the electrical design was done by Ganesh
Basawapatna. Bob Jewett invented the PIN switch concept and
contributed to the electrical design. Lee Olmstead produced the
mechanical design and with Steve Sparks developed the magnet.
-Lynn Rhymes
Fig. 1. Switched YIG-tuned multiplier (SYTM) produces out
put frequencies from 0.01 to 26.5 GHz.
1 G L Matthaei, L Young, and E.M.T. Jones, "Microwave Filters. Impedance Matching
Networks, and Coupling Structures," Artech House Books, 1980 (Reprint of
McGraw-Hill Book Co. Edition. 1964). Sec. 17.05.
© Copr. 1949-1998 Hewlett-Packard Co.
sweep settings. Plug-in front-panel conditions are saved
and recalled by the plug-in firmware, which also contains
self-test routines that check the state of the plug-in at
power-up and routines to check the front panel for invalid
operation. All digital-to-analog converters (DACs) in the
plug-in can be checked for valid operation with a series of
operator-initiated tests.
YIG Coil Drivers
The technique for converting the tuning voltage received
from the mainframe to the correct YTO or SYTM magnet
current is the same in all 83500 Series plug-ins. Since the
tuning voltage is defined to be 0 to 10V for any complete
frequency band regardless of its length, the tuning sensitiv
ity (Hz/volt) can vary depending on the particular frequency
band. One of the primary functions, therefore, of the YTO
and SYTM coil drivers is to scale the tuning voltage in each
band to create a constant-sensitivity ramp. In addition, the
driver must add to that ramp an offset that tunes the YTO or
SYTM magnet to the start frequency of each band when the
input tuning voltage is 0V. Finally, it must convert the
resulting voltage to the required magnet current, thus tun
ing the YTO or SYTM across the desired frequency range.
Fig. 7 is a block diagram of the YIG driver.
The first two driver functions are accomplished using
multiplying DACs. One DAC scales the tuning voltage and
the other scales a precision -10.000V reference voltage.
The outputs of the two multiplying DACs are summed and
sent to the output current driver which performs the third
function. The digital scale factors for each band are stored
in the plug-in ROM and the DACs are updated whenever the
plug-in changes bands. This results in greatly increased
flexibility over previous plug-in YIG driver designs which
required custom precision resistors for each separate fre
quency band. By making the drivers programmable, it was
possible to use identical driver boards for an entire family
of plug-ins, thus reducing inventories and simplifying
Programmability also realizes improvements in another
key area: instrument adjustment and calibration. Because
YIG magnet sensitivities can vary over a significant range,
previous plug-in YIG drivers required gain and offset
potentiometers for each band and factory-selected resistors
to adjust for these variations. For a multiband plug-in such
as the 83595A, which has five frequency bands and two
driver boards (one for the YTO and one for the SYTM), a
total of nineteen pots, together with their associated
custom-value precision resistors, would be required to
make the necessary offset and slope adjustments. (The
SYTM is not swept in the low band, so only an offset
adjustment is needed in that band.) Since potentiometers
degrade both temperature performance and reliability, a
reduction in their number was considered very desirable.
Thirteen potentiometers were eliminated by using the slope
and offset DACs as digital potentiometers. By modifying the
slope and offset DAC scale factors in firmware, adjustments
for variations in YIG magnet sensitivities can be made over
a relatively wide range with no sacrifice in performance or
reliability. Furthermore, since the tuning is linear, knowing
the required offset and slope corrections in any one band
allows computation of the required corrections in every
other band. Thus, by calibrating the beginning and end
frequencies of one band in the 83595A, for example, all five
bands can be calibrated simultaneously. For convenience, a
calibrate mode can be entered by using the SHIFT key on the
mainframe which allows adjustment of the offset and slope
{continued on page 18)
Voltage R
«-^A-w^yjK. — 0
Data Bus
Fig. several frequency capable of very wide sweeps actually sweep several individual frequency bands
in sequence. The additional circuitry required to do this is shown here.
© Copr. 1949-1998 Hewlett-Packard Co.
A Frequency Doubler with High
Output Power from 18 to 26.5 GHz
One of the plug-ins for the 8350A Sweep Oscillator, Model
83570A Swept Source, uses a frequency doubier to deliver
+ 1 0 dBm output power from 18 to 26.5 GHz. Doubling is achieved
by operating dual-gate field-effect transistors in a highly nonlinear
mode.1 Two dual-gate FETs are used in a full-wave rectifier
configuration which provides effective suppression of
fundamental (input) frequency feedthrough as well as all
odd-harmonic components. Fig. 1 is a photograph of the doubier
and Fig. 2 is a schematic diagram.
Fig. 1. Frequency doubier delivers +10 dBm output power
from 18 to 26.5 GHz.
Input power to the doubier is split and directed to two sections
of coplanar transmission lines. The conductors in one section are
reversed with respect to those in the other. This accomplishes a
broadband 180° phase shift allowing the first gates of the
doubling FETs to be driven out-of-phase. The drains (outputs) of
the FETs are directly combined through short lengths of
transmission line. A virtual ground is presented at this point to the
fundamental frequency and to all odd harmonics since
equal-amplitude, opposite-phase components are being
combined. The desired second harmonics generated in the two
devices are combined in-phase.
Optimum device terminations were found using a harmonic
load pull network analyzer system.2 This allowed separate
determination of optimum fundamental and second-harmonic
frequency load terminations. As expected, the fundamental load
yielding best doubling efficiency is a pure reactance, shifted
slightly from an open circuit because of device output shunt
capacitance. The optimum load for the desired second-harmonic
output was found to be a large-signal conjugate match. Recall
that the fundamental frequency sees a virtual ground through
some length of transmission line. Thus the optimum fundamental
load termination is easily realized by varying the length of
transmission line between the output of each device and the
combining point. The optimum second-harmonic load is near 50
ohms and is also not difficult to achieve.
Dual-gate FETs have been found empirically to be significantly
more This as frequency doublers than single-gate FETs. This
is believed to result from their superior gain during the linear
amplification portion of the cycle and their greater output
nonlinearity. Input nonlinearity is not believed to be a significant
contributor to FET harmonic generation. The dual-gate FETs' first
gates are conjugately matched at the input frequency and the
second gates are grounded through a short wire bond.
Use of the dual-gate FET output nonlinearity for harmonic
generation implies higher second-harmonic signal level for larger
output signal swings. Thus it was desirable to provide one stage of
FET preamplification to drive the doubier. This was accomplished
by using another of the dual-gate FETs with conventional
conjugate input and output matching. Again the dual-gate device
offers advantages over the single-gate device by virtue of its
superior gain through the input frequency range.
Two stages of postamplification are used to boost the output
signal and compensate for the insertion loss of the PIN diode AM
modulator which is placed between the postamplifier stages.
Each of these amplifiers uses a 0.5-/xm-gate-length FET (gate
width is 350 /urn) which is conjugately matched for its mediumpower output operating conditions.
Finally, a microstrip coupler at the output delivers power to a
Schottky diode detector. The dc signal from this detector is then
fed to in instrument's automatic leveling control loop which in
turn drives the PIN modulator. A coupler at the component's input
provides a signal to the rear panel of the instrument. A customer
might wish to use this 9-to-13.25-GHz signal to drive a counter or
other instrument limited to 18-GHz operation.
Fig. 2. Schematic diagram of the
frequency doubier.
© Copr. 1949-1998 Hewlett-Packard Co.
The doubler component is driven with + 1 3 dBm from 1 8 to 26.5
GHz. The doubler stage itself delivers a minimum of +10 dBm to
the postamplifiers and modulator. The resulting +13 dBm output
level +10 sufficient to exceed the instrument specification of +10
dBm output even with the inclusion of an 1 8-to-26.5-GHz isolator
which directly follows the doubler component.
We would like to thank Roger Stancliff and Rolf Dalichow for
their guidance and technical leadership. We also wish to recog
nize Washington Gonzalez for his infinite patience as the first test
technician on the product, Ron Blanc for his early production
corrections using the front-panel power knob. The plug-in
light-emitting diode display in this mode displays a
hexadecimal code indicating the switch settings of two
switches (one for offset, the other for slope) which are
mounted on the driver boards. When the instrument is
powered-up, or when INST PRESET is pushed, these
switches are read by the microprocessor, which then calcu
lates the required corrections for each band, adds them to
the nominal offset and slope DAC scale factors stored in
ROM, and saves them in RAM for use whenever a band
change is necessary.
To enhance the overall frequency accuracy of the plugins, the YIG drivers compensate for errors introduced by
YIG magnet nonlinearity, power supply variations, and
magnet eddy current delay caused by sweeping. Fig. 7
shows how magnet nonlinearity compensation is ac
complished. The output current driver uses feedback to
force Vref equal to V, the drive voltage. Since Ql's base
current is negligible (the actual circuit uses three transis
tors) and the input impedance of the operational amplifier
is high, all of the current through Rref is required to flow
through the coil. As Vref drops with increasing magnet
current, it crosses the level set by selected resistors Rl and
R2 and the diode conducts, thus shunting Rref with the
parallel value of Rl and R2. The ratio of Rl and R2 deter
mines the switch point and their magnitudes govern the
amount of correction applied. Several such networks are
required to reduce the nonlinearity of a given YIG device to
a few megahertz across any band. To correct for variations
in the +20V current driver supply, its voltage is added to
the offset and slope DAC voltages in such a way that Vref
tracks any changes in +20V, thus resulting in no net voltage
change across Rref. This technique also assures good im
munity to variations in +20V among mainframes.
When the current in a YIG magnet is changed rapidly,
eddy currents are produced in the core, causing the net
magnetic field at the YIG sphere to lag the drive current. A
good approximation of the resulting frequency error caused
by this delay is given by the equation,
terror = a (df/dt)(bF(f) + c)
where df/dt is the sweep rate, a, b, and c are all constants,
and F(f) is a function that is directly proportional to fre
quency, but whose value is zero at the start of any sweep.
The delay compensation circuitry uses a differentiator and
an analog multiplier to implement the above equation, and
the resultant correction voltage is added to the main drive
engineering support, and Larry Studebacker and Bill Loofbourrow for their process engineering efforts. Many others have
worked hard to build this complex product, and their efforts are
greatly appreciated.
R Stancliff
i . n.
oiai luii 1f, 1,
Daiai p^cu
re Frequency
n tjqutjriuy Doublers."
uuuuitÃ-rs, IEEE
iccc (VI
I I Syrr
posium Digest,
it, 1981,
1981, pp. 143-145
2. R. Symposium and D. Poulin, "Harmonic Load Pull." IEEE MTT Symposium Digest. 1 979,
pp. 185-187.
-Val Peterson
-Jerry Orr
voltage. This technique represents an improvement over
previous designs, since it results in better swept frequency
accuracy while reducing the number and complexity of the
adjustments required.
Care was taken throughout the design of the YIG drivers,
as well as all other sensitive circuitry, to see that residual
FM at the output was kept to a minimum. Precision, lownoise components are used throughout the tuning circuitry,
and particular attention was given to the routing of signal
and ground lines to avoid unwanted noise due to crosstalk
or ground loops. Low-noise Zener diodes were used in the
offset DAC reference circuitry, and in the case of plug-ins
using harmonic multiplication (8359x), the uncorrelated
noise of three Zener diodes is averaged to further reduce
their total noise. As a result of these techniques, the
maximum residual FM of the 83595A, for example, is only
12 kHz peak at 26.5 GHz.
Multiband Sweep Circuitry
Plug-ins capable of wide, multiband sweeps, such as the
83590A, 83592A, 83594A, and 83595A, do so by sweeping
several individual frequency bands in sequence. When
doing this, the plug-in stops the sweep and requests a band
change whenever it reaches the end of a band. Once the
band change has been completed, the sweep is released and
the plug-in proceeds to sweep the next band. The additional
circuitry required to accommodate multiband sweeps is
shown in Fig. 8. It performs two main functions. First, it
monitors the tuning voltage TV and indicates when a pre
determined switch point is reached. Second, it converts the
0-to-10V tuning voltage ramp into multiple 0-to-— 10V
ramps, one for each of the separate bands in the multiband
sweep. These functions are performed only during a mul
tiband sweep.
FM Driver
The 83500 Series of plug-ins offers as a standard feature
wideband FM capabilities previously obtainable only with
a few specific plug-in models. The 10-MHz external FM
bandwidth of all 83500 plug-ins is made possible by an FM
driver design that is common to the entire series. A
simplified schematic of this circuit is shown in Fig. 9.
The FM input signal is first separated into low (<700 Hz)
and high (>700 Hz) frequency channels. The lowfrequency portion of the signal is buffered, scaled, and sent
to the YIG driver board(s) where it is summed with the main
YIG coil driver voltage(s). The high-frequency component
is also scaled, but drives instead a small FM coil internal to
© Copr. 1949-1998 Hewlett-Packard Co.
To SYTM Driver
(8359X Series Only)
Fig. 10-MHz bandwidth. FM driver used in all 83500 plug-ins provides a wide 10-MHz external FM bandwidth.
This is a simplified schematic of the FM driver.
the YIG oscillator. The microprocessor selects the gain of
both channels depending on the FM sensitivity selected
(-6 or -20 MHz/V), whether that sensitivity is relative to
the plug-in's output frequency or to an auxiliary, funda
mental output, and whether the plug-in is operating in a
fundamental or harmonic frequency band. Of course, the
last two of these factors apply only to those plug-ins using
the SYTM; the others require only a sensitivity select, and
the unneeded gain control components are deleted.
The high-frequency channel consists of a video amplifier
(U4), and a broadband, bilateral output current driver con
sisting of a high-speed operational amplifier (U5) and cur
rent buffer (U6).
Automatic Power Level Control
In addition to controlling and modulating the RF output
frequency of a sweeper, it is very desirable to control the
amplitude of the signal being produced. Knowledge of a
source's absolute amplitude and the ability to vary it are
needed in the measurement of many nonlinear devices. To
achieve this the automatic level control (ALC) circuitry in
the 83500 Series performs several functions:
• Maintenance of calibrated power at the output connector
as frequency changes
• Provision for operation with the 8755C Scalar Network
• Compensation for increasing attenuation versus fre
quency in external hardware between the sweeper and
the device under test (slope)
• Continuously increasing power as a function of sweep
voltage (power sweep)
• Generation of amplitude markers
^ Blanking of RF during reverse sweeps
• Provision for use of an external crystal detector or an HP
432A/B/C Power Meter for power control
• Provision for analog and square-wave amplitude mod
All of these functions are under the control of the micro
processor in the 83 50 A mainframe.
To provide for the potentially large range of output power
that could be desired from a source, the decision was made
to maintain logarithmic rather than linear control of output
power. That is, the output is controlled in dBm (decibels
compared to one milliwatt) rather than in milliwatts or
watts. This decision also reduced the extremes over which
the control loop needed to operate. Fig. 10 shows a block
diagram of the ALC circuits.
In internal leveling mode a directional coupler produces
an output that is a small sample of the forward power
appearing at the output connector. This sample is then
converted by a diode detector to a dc voltage proportional to
© Copr. 1949-1998 Hewlett-Packard Co.
A Broadband 2-to-7-GHz
Power Amplifier
The broadband GaAs MESFET Amplifier for the 8359x sweep
oscillator plug-ins provides 500 mW of RF power from 2 to 7 GHz
to the is multiplier. The signal input to the amplifier is
approximately 10 mW from the preceding oscillator and mod
ulator. The YIG-tuned multiplier generates harmonics and filters
unwanted signals to create a 2-to-26.5-GHz sweeping output (see
page 15).
The new amplifier design is based on an existing 2-to-6.2-GHz,
300-mW amplifier.1 The requirements are an operating frequency
of 2-7.0 GHz, a power output of 500 mW, and power consumption
less than the existing design's 1 8 watts. A logical block diagram,
simple design, uniform performance, and ease of manufacturing
are as important as the electrical specifications.
The basic block diagram is shown in Fig. 1, and a photograph in
Fig. The The numbers indicate the gate widths of the devices. The
final stage consists of two 1500-/nm gate-width FETs in a bal
anced configuration. These FETs use a VDS of 7 volts and are
100% 14 tested for drain breakdown voltage greater than 14
volts. The optimum power match for this device was measured
using single load pull technique.2 The results revealed that a single
quarter-wavelength transformer would produce adequate output
Lange Couplers
Fig. 2. 2-to-7-GHz Amplifier.
matching across the 2-to-7-GHz frequency range.
The interdigitated coupler3 has four fingers with 0.03-mm spac
ing between conductors. An advanced thin-film technology pro
duces mm with this spacing to an accuracy of 0.0025 mm
on a 0.64-mm-thick sapphire substrate.
Typically, the small-signal gain is flat within ±2.5 dB, largesignal 2.0 is flat within ±1 .0 dB, and output v'SWR is less than 2.0
from 2 than 7.0 GHz. Total power consumption is typically less than
10 watts including the bias board.
The original amplifier was designed by Derry Hornbuckle, who
also gave guidance to this amplifier design.
1. D. IEEE "A 2-6.2 GHz, 300 mW GaAs MESFET Amplifier," IEEE MTT Sym
posium Digest, IEEE Catalog No 78CH 1355-7
2. D. Microwaves "Load pull measurements help you meet your match," Microwaves
November 1980, p. 61.
3. J. IEEE "Interdigitated Stnpline Quadrature Hybrids," IEEE Transactions,
Vol. MTT-17, December 1969, p. 1150-1.
Fig. 1. 2-ÃO-7-G/-/Z amplifier block diagram. Numbers are
gate widths of the FET amplifier stages.
output power. After being processed by appropriate
switches and buffering, the signal is applied to a
logarithmic amplifier which produces an output propor
tional to the logarithm of the output power. This voltage is
scaled and compared at the summing node to a voltage
proportional to the desired output power which is gener
ated in the reference circuits.
The reference voltage is dependent on a number of in
puts: the requested power input from the rotary pulse
generator, keyboard, or HP-IB, modification of that power
due to power sweep or slope requirements, any signal
applied at the external AM input for amplitude modulating
the RF, and any correction applied to compensate for
coupler/detector variations. The power sweep and slope
functions are produced by processing a voltage propor
tional to the sweep (0-10V, start to stop) in a multiplying
DAC. In slope mode, the microprocessor calculates the
power modification required based on the frequency range
-Michio Furukawa
being swept and the slope correction requested. Because of
the logarithmic amplifier in the detector path and the def
inition of the signal from the reference path, any voltage
difference at the summing node is proportional to the error
between the desired output and the obtained output, ex
pressed in dB.
The error voltage is applied to the main loop integrating
amplifier. The output of this amplifier is used to control an
exponential current source which drives the appropriate
PIN diode modulator. The exponential current source en
ables the RF output to respond linearly in decibels to a
voltage change at its input. This is necessary since the loop
variable is dBm. The main loop amplifier also drives a
comparator to indicate an unleveled power condition. Unleveled power may occur when output power greater than
the specified instrument power is requested.
External leveling operates in a manner very similar to
internal leveling. In the case of external crystal detector
© Copr. 1949-1998 Hewlett-Packard Co.
Power Sweep Slope
Detector Coupler Compensation
Data Bus
Power Level
Information from:
Power Meter
Amplitude Marker
Square Wave Modulation
Pulse Modulation
Fig. maintain 0. Block diagram of the internal leveling circuits that maintain constant output power and
provide for amplitude modulation.
control, the external detector input is applied to the buffer
and logarithmic amplifier. HP 432A/B/C Power Meter oper
ation is similar, with the exception of additional loop com
pensation included ahead of the logarithmic amplifier and
in the main loop integrator. This compensation is necessary
to stabilize the loop with the slower response time of the
power meter.
To provide compatibility with the 8755C Scalar Network
Analyzer, the RF signal is modulated with a 27.8-kHz
square wave. To provide this, a track-and-hold circuit is
used. This allows use of the internal ALC modulator and
avoids any additional losses that an external modulator
would produce in the RF path. The track-and-hold circuit
allows the RF power to be removed without requiring that
the loop follow. The amplifiers in the loop hold the levels
that existed just before power was removed and then begin
tracking again when power is returned.
Amplitude markers operate in a similar manner, with
power removed and the track-and-hold circuit holding dur
ing an amplitude marker.
many people made to the design of these plug-ins. We
would like to thank Irv Hawley and Arlen Dethlefsen for
their managerial support. Paul Hernday and Roger Stancliff
were project managers during the early phases of the de
sign. The product design was done by Daniel Swan, David
Copley, and George Baker. Lynn Rhymes, Sue Conway,
John Regazzi, and Jerry Orr worked on the printed circuit
board design.
Gary W. Holmlund
A native of Bucklin, Missouri, Gary
Holmlund attended the University of
Missouri at Columbia, receiving his
BSEE and MSEE degrees in 1972. He
joined HP in 1973, helped design the
8620C/86290A Sweep Oscillator, and
has served as project manager for the
86290B and several 83500 Series RF
plug-ins. He is a co-inventor on a patent
on a method of providing crystalaccuracy markers while sweeping.
Gary is married, has two sons, lives in
Santa Rosa, California, and has a
strong interest in microcomputers.
We would like to acknowledge the contributions that
Duaine C. Wood
Duaine Wood received his BSEE and
MSEE degrees from Brigham Young
University in 1976 and 1977. With HP
since 1977, he had primary design re
sponsibility for the 835xx YTO drivers,
the 8359x YTM driver, and the 835xx FM
driver. He also helped design the 8359x
multiband sweep circuitry. Duaine was
born in Provo, Utah, grew up in the San
Francisco Bay Area, and now lives in
Santa Rosa, California. He's married
and has three children. In addition to
running, gardening, woodworking, and
electronics, Duaine is involved in
church work and enjoys family ac
Glenn E. Elmore
Glenn Elmore joined HP's Santa Rosa
Division in 1972 after two years of run
ning his own communications system
business. He spent four years in instru
ment test, then joined the R&D lab and
helped design several sweeper prod
ucts, most recently serving as an as
sociate engineer on the new 83500
Series plug-ins. Glenn was born in
Sebastopol, California and now lives in
nearby Santa Rosa. He is married and
his interests include bicycling, amateur
radio, and signal propagation via manmade and natural satellites.
© Copr. 1949-1998 Hewlett-Packard Co.
Portable Defibrillator-Monitor for Cardiac
This new portable defibrillator monitors the patient,
measures its effectiveness in delivering a high-voltage
pulse to the patient, and provides a permanent record of
the resuscitation procedure.
by Paul I. Bennett and Victor C. Jones
A VERY TENSE DRAMA begins when a human heart
ceases to beat. Death from irreversible damage of
the nervous system is three to five minutes away
but can be avoided if proper action is taken to restore blood
circulation promptly. Since the probability of a successful
resuscitation procedure is highly dependent upon the
amount of time the heart is not beating, little time is avail
able for setting up equipment and looking for things; atten
tion must be directed to the patient. Afterwards, the
procedure must be documented for legal purposes, and
for the review committee who will analyze the entire
event for possible improvement at the next emergency
procedure. The new HP Model 78660A Defibrillator-Moni
tor (Fig. 1) is designed for this type of medical emergency.
The heart is a mechanical pump that is bioelectrically
controlled. During a heart attack the muscle action of the
heart deteriorates from a coordinated periodic contraction
to a convulsive quiver known as fibrillation. This is often
brought about when the myocardial muscle is deprived of
blood by a restricted artery (myocardial infarction) leading
to a condition where many areas of the heart rather than one
attempt to electrically initiate the pumping contraction.
The ensuing biological pandemonium is ineffective in
pumping blood through the body. Cardiac resuscitation
requires that fibrillation be terminated and the heart be left
in a condition allowing the body's natural electrical
pacemaker to stimulate proper pumping action.
Medical personnel must make many assessments before
emergency treatment is begun. The status of heart action is
indicated by its bioelectrical activity which can be observed
by placing electrodes on the chest and connecting them to
an EGG (electrocardiogram) monitor or recorder. Vital signs
are checked. Attention is given to ensure an open breathing
passage. Patient history is mentally reviewed but time con
straints dictate that action be quick and optimal.
Among the treatment options open to medical personnel
Fig. 1. The HP Model 78660 A
Defibrillator-Monitor is a con
venient portable instrument for
resuscitation, monitoring, and
documentation in cardiac-arrest
© Copr. 1949-1998 Hewlett-Packard Co.
Power Supply,
Peak I Circuit
Fig. 2. Block diagram of the 78660A.
are chemical, mechanical and electrical stimulation of the
body. Chemical or drug injections are more effective in
managing the situation after the heart is restarted than in
arresting fibrillation. Mechanical means such as the CPR
(cardiopulmonary resuscitation) technique of compressing
the heart between the anterior chest wall and the thoracic
spine restores only about 20% of normal blood flow and it
does not alter the basic problem of fibrillation. It may pro
long life, however, until additional help arrives. Precordial
thumping or beating on the chest to terminate fibrillation
has enjoyed only limited success. The chaotic bioelectrical
activity of the heart can be stopped if a momentary electrical
current is passed through it. This can be done by applying a
high-voltage pulse to the chest. This procedure is known as
defibrillation and is now a standard emergency treatment in
the hospital and for paramedic teams in many areas.
Hewlett-Packard has produced defibrillators since the
procedure became standard in the early 1960s. At that time
a defibrillator was simply an ac-power-line-operated box
that produced a high-voltage pulse. However, resuscitation
is an evolving science and so is the equipment. Researchers
are satisfied that a dc damped-sinusoid waveform having
an energy up to several hundred joules (watt-seconds) and a
duration of several milliseconds has a high probability of
success if applied promptly. However, much work remains
to optimize the required equipment. In the heat of battle to
save a patient's life the operator must not be faced with
distractions or confusion about operating the instrument
and should be provided with as much pertinent informa
tion as possible to aid in making correct decisions. Tradi
tionally, informational feedback from the defibrillator has
been minimal. An EGG monitor which displays the bioelec
trical activity of the heart is usually provided and a recorder
to document the waveforms has been available, but judg
ment about the effectiveness of the high-voltage discharge
is often based on an observation as to how high the patient's
body jumps (convulses) during the procedure.
The HP Model 78660A Defibrillator-Monitor measures
the current flowing through the patient during the highvoltage discharge and the patient's impedance is calculated
from the knowledge of the peak current. During discharge,
transthoracic impedance is usually within the range of 25 to
100 ohms so anything over 100 ohms is immediately
brought to the attention of the operator.
A block diagram of the basic components used in the
78660A is shown in Fig. 2. For various operator activities
the recorder runs automatically, minimizing the need for
operator attention. Time and heart rate are periodically
printed and procedure parameters such as selected energy,
discharge energy, peak discharge current and patient im
pedance are recorded automatically. The EGG monitor has
been improved with an automatic gain setting, baseline
(offset) restore, 60-Hz filter and a very effective heartbeat
detector. Should a drug injection or other special treatment
be given, the operator can press a button and the exact time
of that event is recorded. The instrument package is de
signed to solve some logistics problems by providing on
board storage for most desired accessories and a novel
mounting scheme for use on carts and storage between
Measurement of Defibrillation Parameters
Providing an effective defibrillation pulse during an
emergency procedure is critical to the success of the resus
citation effort. Many variables can change the output dose,
but without direct feedback the operator has little informa
tion to use in assessing the performance of the procedure. A
method of providing defibrillation pulse information inFrom High- Voltage
L 1 H i n t
36 mH 11.5ÃÃ
R«t, Ed
Fig. 3. Schematic diagram of the basic defibrillator discharge
© Copr. 1949-1998 Hewlett-Packard Co.
87A at 25Ã1
64A at 50ÃÃ
H i n t
1 1 . 5 0
L 3 6 m H
C 3 2 / n F
Energy Stored = 492 Joules
Energy Delivered Into 50 U
= 400 Joules
400 joules at 100 ohms and as much as 780 joules at 150
ohms, an energy variation greater than 10 to 1.
It is therefore clear that because patient impedance can
vary considerably it has a strong effect on defibrillation
effectiveness. Hence, knowledge of peak current, patient
impedance and actual delivered energy will greatly en
hance the ability of the operator to assess and improve
defibrillation effectiveness. Optimizing patient impedance
through appropriate techniques will enhance the probabil
ity of successful defibrillation at lower selected energies.
Patient impedance and delivered energy can be directly
determined from the peak discharge current Im if the stored
energy Es and the defibrillator circuit parameters (capaci
tance C, inductance L and internal resistance R¡nt) are
known.1 Peak current can be expressed in the most gen
eral form:
Im = f (ES,R,L,C)
13 14
Fig. 4. Discharge currents versus time for various patient
resistances given a fixed stored energy of 492 joules.
stantly is needed along with direct documentation of the
parameters for clinical use.
Fig. 3 shows a schematic diagram of a defibrillator cir
cuit. A selected amount of electrical energy Es is stored in a
capacitor. The discharge waveform is determined by the
defibrillator circuit parameters (capacitance C, inductance
L, and internal resistance R¡nt) and by the external resis
tance Rext presented to the defibrillator paddles. Unless
poor skin preparation produces a significant interface resis
tance, the external resistance should be very close to the
patient's transthoracic impedance. The following discus
sion assumes that the external resistance is essentially
equal to the patient's impedance.
It is now thought that, in addition to patient related
parameters, defibrillation effectiveness depends primarily
on the current density through the myocardium, and hence
correlates more directly with peak current than with deliv
ered energy.
Fig. 4 shows the discharge current going through the
chest for four values of external resistance Rext> given a
fixed amount of stored energy (that which would deliver
400 joules into 50 ohms). When Rext is of the order of 50
ohms the circuit is critically damped and the discharge
duration is minimal at approximately 5 milliseconds. These
conditions are considered optimum for effective defibrilla
tion and therefore the defibrillator circuit parameters are
usually chosen to yield a critically damped discharge for a
value of Rext close to the average observed value of patient
Fig. 5 shows the relationship between the peak current
and delivered energy for different values of patient imped
ance using the 78660A. If we accept the present theory that
current defibrillates and assume that, for instance, a certain
adult patient requires a peak current of 40 amperes, Fig. 6
shows that the required delivered energy is 150 joules at 50
ohms , but ranges from a low value of 60 joules at 20 ohms to
where R is the total circuit resistance, i.e., R = Rint + Rext.
The dependence of peak current Im on stored energy Es is
very simple, Im being proportional to the square root of the
stored energy. For instance, Fig. 5 shows that the peak
current at any external resistance is exactly twice as high at
400 joules as at 100 joules. Consequently, the general ex100
£ 60
~ 50
4 0
400 Joules Delivered into
50 Ohms (Stored 492 Joules)
100 Joules Delivered into
50 Ohms (Stored 123 Joules)
5 0
1 0 0
1 5 0
2 0 0
External Resistance Re*t (Ohms)
Fig. 5. Peak discharge current versus external load resis
tance for two stored energies.
© Copr. 1949-1998 Hewlett-Packard Co.
25 Ohms
J 25
20 30 40 50 70 100
200 300 400 500
Delivered Energy Ea
Fig. 6. 78660/4 peak discharge current lm versus delivered
energy Ed for various values of patient resistance.
pression for peak current can be restated:
X f[(Rint + Rext).L.C]
The defibrillator circuit parameters (R¡nt,L,C) are known
and fixed for a given defibrillator, thus the only unknown
variable is patient impedance Rext. Hence, substituting
known values for the defibrillator circuit parameters, the
normalized peak discharge current im becomes a function
of Rext only:
Annotating Strip
Chart Recorder
Energy Display
Fig. 7. Simplified block diagram of the 78660A's discharge
control and recording circuitry.
= f (Rext)
The function of Rext is most conveniently derived from
general circuit equations, and can also be measured directly
through a series of discharges with different external resis
tors. The function is single-valued and monotonic, hence
there is a one-to-one correspondence between Im and Rext.
Fig. 7 shows a simplified diagram of the discharge con
trol and recording circuit. Before the discharge the opera
tor selects the desired energy E¿ to be delivered into a
50-ohm load. The microprocessor determines the corres
ponding value of stored energy Es by using the expression
Es = E¿ (Rext + R¡nt)/Rext- The corresponding storage capa
citor voltage V0 (Es = Q.5CV20) is sensed and regulated by
the microprocessor. The discharge current passes through
a current-sensing transformer placed in the wiring. Use of
a sensing transformer provides ground isolation for the
patient circuit, yielding a safe, simple method for mea
suring the discharge current. The transformer provides a
voltage signal that is peak-detected and recorded by the
microprocessor. Since the entire waveform is not needed, it
is much more practical to sample the peak current in a
highly filtered noise-immune analog circuit than to digi
tize the entire waveform. The microprocessor takes the
measured peak discharge current and uses this value along
with the stored energy Es to determine patient impedance
Rext and delivered energy E¿ as discussed above. The mi
croprocessor then drives a three-digit display and annotates
peak current, patient impedance and delivered energy on a
strip-chart recorder. A POOR PADDLE CONTACT warning
indicator is also activated when the patient impedance ex
ceeds 100 ohms. The 100-ohm caution level was selected
because, with good paddle preparation and location, the
patient impedance should almost always be well below 100
ohms. The time of day is added to the strip-chart to com
plete documentation of the defibrillation episode and pro
duce a good permanent record.
QRS Detection
Both the heart-rate calculation and a procedure called
synchronized cardioversion require that the R-wave por
tion of the EGG signal be detected. To terminate atrial fibril
lation, the defibrillator can be used in a synchronized mode
where the high-voltage pulse to the heart is initiated at the
moment of the patient's R-wave. To be effective, the QRS
detector must detect the QRS complex (see Fig. 8) within 20
milliseconds of the peak of the R wave and also minimize
chances of triggering the discharge during other portions
of the EGG signal, such as the T wave, which could
change atrial fibrillation into the more serious ventricular
Also, calculation of the heart rate, which is displayed on
the instrument, is highly dependent upon the reliability of
the signal from the QRS detector. Should a high-level signal
occur, a simple peak-following threshold would miss the
next several R waves. By using a floating threshold that is an
average of several previous peak values, the number of
undetected beats is reduced.
An integrating analog-to-digital (A-to-D) converter
supplies digitized EGG data to the microprocessor at 4-ms
intervals. It is then digitally bandpass filtered, full-wave
rectified and compared with the floating threshold (Fig. 9).
The equation for the bandpass filter of 7 to 35 Hz is:
Y(t) = (3/4) Y(t-l) + [X(t) - X(t-4)]
© Copr. 1949-1998 Hewlett-Packard Co.
Automatic Gain Control
Before the ECG waveform is digitized it passes through a
variable-gain stage which is simply an operational
amplifier with one of five inverting feedback resistors
selected by the microprocessor (Fig. 9). If the signal is
greater than a maximum threshold, the processor selects the
next lower gain setting. If the signal is below a minimum
threshold for two seconds, the processor selects maximum
gain and then decrements the gain until the maximum
threshold is no longer exceeded.
This gain algorithm works fine until the patient being
monitored moves and the ECG baseline wanders with the
patient's motions. By filtering the ECG signal with a highpass filter the baseline wander can be removed. Sampling at
4.1-ms intervals a high-pass pole below 2 Hz can be ac
complished and represented by an equation somewhat like
this one:
Y(t) = (31/32)Y(t-l) + [X(t) - X(t-l)].
Fig. 8. Typical ECG heartbeat waveform.
where X(t-4) represents the fourth previous input into the
filter and Y(t-l) is the previous output of the filter. This
filter is preferred over a simple differentiator because the
spectral density of a QRS complex centers about 10 to 20 Hz.
Using this filter delays the signal only about 16 ms.
A pulse is output whenever the filtered signal exceeds the
floating threshold level. This level varies with the time
elapsed since the last detected R wave. For the first 200 ms
after an R wave all detection is inhibited, establishing the
upper limit of 300 beats per minute. After this period, the
threshold is set at 75% of the average peak value detected
during the previous heartbeats. If no signal exceeds the
threshold after 75% of the time measured between the two
previous detections, the threshold is reduced again by 25%.
If no R wave is detected after two seconds (corresponding to
30 beats per minute) the threshold drops to a predetermined
minimum level.
Hardware Functions -4
But when an 8-bit signal is being divided by a number
greater than 16 the resulting quotient is 3 bits or less (in
teger arithmetic). However, if consecutive A-to-D samples
are added in pairs, the result is essentially a 9-bit A-to-D
conversion with an 8.2-ms sampling time. After this, the
same filter can be derived from
Y(t) =
[X(t) - X(t-l)]
Both methods yield a high-pass filter with a pole at 1.2 Hz
but the latter increases the resolution of the division by 2
bits — one from the extra ninth A-to-D bit and the other from
dividing by 16 instead of 32.
An advantage of software signal processing is the ma
nipulation of filters without component changes. With
crystal timing, poles and zeros are unaffected by tempera
ture changes. Also, optimization of the floating threshold
algorithm would be impractical in hardware.
Noise Susceptibility
Microprocessors and other high-density large-scale in
tegrated (LSI) circuits have small voltage differences be-
Software Functions
To Display Memory
R Wave
Fig. 9. Block diagram of
heartbeat detection circuitry.
© Copr. 1949-1998 Hewlett-Packard Co.
tween logic states, small transistor cell size, high-speed
performance, and numerous leads to the outside world. All
of these factors make them prime targets for noise, both
conducted and radiated. Malfunctions can take the form of
altered RAM (random-access memory) cell contents, output
latches that change state, and altered internal registers such
as the program counter. If the program counter is modified,
software execution continues, but operation becomes
erratic and highly unpredictable. Normal operation is
restored only when the operator switches the power off and
on again.
When the instrument in question is a critical medical
device, malfunctions cannot be tolerated. In a defibrillator
noise is a problem because the instrument must control
discharge of a capacitor charged to as much as 5000 volts
along current paths to both the patient and a safety bleeder
resistor through low-bounce relay contacts. Electrical noise
of frequencies up to 100 MHz is rampant during the brief
discharge interval. Cables and printed circuit board foils
become receiver antennas and capacitive couplers.
Power-supply bypassing must be well applied and aug
mented by bypass networks on the I/O (input/output) ports
and control signal lines. An additional mandatory
technique involves joining the software to the automatic
hardware reset of the microprocessor. Called a tickle cir
cuit, a a oscillator will reset the processor if a
software-generated signal, called the heartbeat, does not
continually inhibit the oscillator. If noise causes the proces
sor to run off to execute some other section of software and
fail to produce the heartbeat on schedule, the tickle reset
restarts the processor back at address location zero. The
processor then looks in the RAM for the presence of a
password, previously installed upon the successful com
pletion of the power-up sequence. The presence of this code
differentiates this warm start (was on, but got reset) from a
cold start (power was just turned on).
Each of the three processors in the 78660A uses this
technique in a slightly different manner. In the control
processor, this warm-start code identifies the particular
software module that was operating before the malfunction
and a vectored jump to that module after the reset im
mediately restores operation, completely transparent to the
operator. Since the control processor is aware of each im
pending discharge and its attendant noise, the processor
goes into a sleep function after energizing the discharge
relays and remains in a reset state during the noise period.
The recorder processor has a number of tasks, each in a
software module, that are sequentially performed every 16
ms. As each is performed it adds one bit to the code. At the
end of 16 ms, the code byte is examined and if proper, a
heartbeat pulse is output. If improper, signifying irregular
processor activity, a software halt is issued terminating
heartbeat output. The tickle circuit will then reset the pro
cessor and the presence of the proper password will warm
start the processor back into normal operation. Time to
rectify the abnormal operation is about 25 ms and is trans
parent to the operator. The ECG processor issues a heartbeat
to its tickle circuit every 4 ms and will reset in 30 ms if it
experiences abnormal software execution. In addition, its
software contains halt traps in blank ROM (read-only mem
ory) areas.
The attractive polycarbonate case allows the 12.7-kg in
strument to be easily carried by its integral handle. The
center of gravity is directly below the handle to avoid bang
ing against the carrier's body and the perimeter of the case is
designed to absorb energy when accidently hitting walls,
doors, etc. Integral paddle holders and cables that retract
into the instrument ensure that cables are not snagged when
rushing down the hall (Fig. 10). A pouch is provided to hold
the necessary ECG lead set and paddle electrode paste. The
unit is stable when placed on its bottom or back and is
rainproof. Small electrodes for pediatric patients are avail
able by simply unscrewing the adult electrodes from the
paddles. The nickel-cadmium battery pack can be replaced
in a few seconds. The standard power base mounts verti
cally, horizontally, or on a drug cart, and provides a socket
for automatic connection to ac power when the 78660A is
placed in the base. Various display bezel overlays snap in
on the front panel for different production models and
language options.
The noise studies and implementation of patient imped
ance were accomplished by Paul Long. ECG software and
hardware are to the credit of Peter Wai. Industrial designer
Kail Peterson evolved the highly functional yet attractive
package. Many others including Martin Rockwell, Mickey
Fig. 10. The 78660 A can be quickly detached from its 78668 A
Quick-Mount Power Base and easily carried to a cardiac
emergency. All of the necessary cables, electrodes, and re
cording paper are contained in the instrument.
© Copr. 1949-1998 Hewlett-Packard Co.
Victor C. Jones
Vic Jones joined HP in 1 973 after earn
ing the MS degree in electrical en
gineering at Rensselaer Polytechnic
Institute. He also received the BS de
gree in the same subject in 1971. Vic
started at HP's Waltham Division and
worked on bedside monitors, produc^^!MJ * ^^ (ion engineering, and defibrillator relia^^^^L ^SJL ^^^^^^ kjijjy 5tuc|jes He transferred to the
I ^^[ fl I McMinnville Division in 1977 and was
• > 1 I the project manager for the 78660A. Vic
• ^^ I ¡s now Dack at Waltham as a project
•k^k • I manager for bedside monitors. He has
I written four papers on the 78660A and
HH1HB ^I^^^HH its delivered-energy algorithm. Vic was
born with Buffalo, New York and now lives in Stow, Massachusetts with
his wife and their two adopted children. He is a member of the local
Kiwanis club and enjoys skiing, automotive sports, and furniture
HP Model 78660A DefibrillatorMonitor
McMinnville Division
1700 S. Baker Street
McMinnville, Oregon 97128 U.S.A.
TECHNICAL DATA: HP Publication Nos. 5952-6834,
6835, 6836.
PRICE IN U.S.A.: 78660A, $7,400; includes
78668A Quick Mount Power Base.
Paul I. Bennett
Paul Bennett received a BA degree in
^° — "~-» iÉt- ' Physics from Linfield College, Oregon
¡ÉL·irAJ á^H^tp^ ^j ¡n 1960. After experience working overseas in Italy and Brazil he went to work
for Field Emission Corporation which
became HP's McMinnville Division in
1973. Paul has worked on high-voltage
and control systems for x-ray equip
ment and was the division's interna
tional sales manager for several years.
More recently he has worked on mi
croprocessor control of a medical x-ray
system and the 78660A. Paul was born
in Chicago, Illinois and now lives in
McMinnville, Oregon. He is married,
has three children, and enjoys playing volleyball and the piano,
working with his home computer, and sailing catamarans.
Mathena, John Brewster, and Ty Hegna contributed in no
small part together with inputs from marketing and pro
duction engineering in a cooperative fashion which ex
emplifies the excellent communication possible in a
small division.
1 . V.C. Jones, et al, "A New Method of Determining Transthoracic
Impedance, Delivered Energy and Peak Current During Defibril
lator of Journal of the Association for the Advancement of
Medical Instrumentation, Vol. 15, no. 6, November-December
HP Model S350A Sweep Oscillator
Santa Rosa Division
1400 Fountain Grove Parkway
Santa Rosa, California 95404 U.S.A.
TECHNICAL DATA: HP Publication No. 5952-9321
PRICE IN U.S.A.: 8350A Mainframe, $4250.
83500 Series RF Plug-ins. $5100 to $27,000.
Bulk Rate
U.S. Postage
Hewlett-Packard Company, 3000 Hanover
Street, Palo Alto, California 94304
FEBRUARY 1982 Volume 33 • Number 2
Technical Information from the Laboratories of
Hewlett-Packard Company
Hewlett-Packard Company, 3000 Hanover Street
Palo Alto. California 94304 U.S.A.
Hewlett-Packard Central Mailing Department
Van Heuven Goedhartlaan 121
1181 KK Amstelveen. The Netherlands
Yokogawa-Hewlett-Packard Ltd.. Suginami-Ku Tokyo 168 Japan
Hewlett-Packard (Canada) Ltd
6877 Goreway Drive. Mississauga, Ontario L4V 1 MS Canada
To change your address ui u«ieie yuur name ..UM. wu< ,.,u M ..„, , ,
changes to Hewlett-Packard Journal, 3000 Hanover Street, Palo Afto, California 94304 U.S.A. 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