1975 , Volume , Issue March-1975
MAttUtt 19 /D
© Copr. 1949-1998 Hewlett-Packard Co.
A High-Performance 2-to-18-GHz Sweeper
This precise, compact microwave sweep oscillator is a
significant contribution to its field. The latest microelectronic
and YIG technology makes it possible.
by Paul R. Hernday and Carl J. Enlow
microwave sources is well established. Produc
tion testing of multi-octave devices requires a broad
band sweeper that is simple to operate. Automated
test systems need a source that is programmable and
can be phase locked. Broadband receiver testing often
requires a compact unit for on-site swept tests. The
broad area of microwave design calls for a generalpurpose sweeper that offers precision performance
and is simple to interface with related equipment
and convenient to operate.
The truly broadband source, a single oscillator
capable of sweeping a frequency range such as 2 to
18 GHz in one uninterrupted sweep, is not presently
available, but there are alternatives. The most direct
method is to sequence several narrower-band
sweepers by means of an external controller so the
resulting sweep covers the desired range. Another
approach, using one sweeper and several plug-in
RF heads installed in a "head-holder", reduces cir
cuit redundancy and simplifies operation by giving
frequency control to a single sweeper, but the equip
ment is still physically large.
A more compact scheme is to put the required oscil
lators in a single sweeper and sequence them inter
nally. More compact yet is a single oscillator followed
by a harmonic multiplier to produce the higher
A New Sweeper
This last scheme is used in the new Model 86290A
2-to-18-GHz RF Plug-In for the 8620A Sweep Oscilla
tor mainframe, making the combination the most
compact broadband sweeper currently available (see
Fig. 1). Designed to meet the requirements of all ma
jor sweeper applications, the 8620A/86290A pro
vides more than +5 dBm of power, leveled to within
±0.9 dB, in fast, continuous sweeps over the 2-to-18GHz frequency range or any portion of this range
(Fig. 2).
Besides the 2-to-18-GHz band, the 86290A pro
vides three narrower bands: 2-6.2, 6-12.4, and 12-18
GHz. All bands are selectable by means of a frontpanel dial drum switch on the 8620A mainframe. In
CW mode in any of the narrower bands, the output
frequency is accurate within ±20 MHz and resolution
is 100 kHz, specifications approaching those of the
traditionally more precise signal generator. Sweep
ing over the widest band the frequency is accurate
within ±80 MHz. Undesired harmonics are at least
25 dB and typically 30 dB below the desired output
The new sweeper weighs just 15 kg (33 Ib), so it can
Cover: In the background
is the new Model 86290A
2-to-18-GHz RF Plug-in
installed in an 8620A Sweep
Oscillator mainframe. In the
foreground are the three
components whose develop
ment made the broadband
RF plug-in possible: a Y/Gtuned oscillator (top), a power amplifier (left), and
a YIG-tuned multiplier (right).
In this Issue:
A High-Performance 2-to-18-GHz
Sweeper, by Paul R. Hernday and Carl
J .
E n l o w
p a g e
Broadband Swept Network Measure
ments, by John J. Dupre and Cyril J.
Yansouni ................. pa§e 15
The Dual Function Generator: A
Source of a Wide Variety of Test
Signals, by Ronald J. Riedel and Dan
D. Danielson . .
page 18
e Hewlett-Packard Company, 1975
Printed in USA
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 1 . Model 86290 A is a new R F
plug-in for Model 8620A Sweep
Oscillator. Four frequency bands
are selectable by front-panel con
trols: 2-6.2, 6-12.4, 72-78, and
2- 18 GHz. The combination is the
most compact broadband sweeper
currently available.
easily be carried to field sites. Its versatility makes it
useful in a wide array of applications, some of which
are described in the article beginning on page 15.
RF Design
Fig. 3 is the RF block diagram of the 86290A. The
basic elements are a single wide-band YIG-tuned os
cillator (YTO), a power amplifier, and a 2-to-18-GHz
YlU-tuned multiplier (YTM). Consisting of a steprecovery-diode and a YIG-tuned filter, the multiplier
converts the amplified YTO signal to useful RF power
levels at twice and three times the YTO frequency.
The YTO's tuning range is 2-6.2 GHz. The YTM
passes this signal directly to provide the first of the
three narrower bands mentioned earlier. The other
two bands, 6-12.4 and 12-18 GHz, are derived by
tuning the YTM to twice and three times the YTO
frequency as the YTO is tuned over 3-6.2 and 4-6 GHz
respectively. Here the YTM is specially biased to
provide maximum harmonic level. These three single
bands will be referred to as bands 1, 2, and 3, cor
responding to their respective harmonic numbers
and 8620A dial drum positions. Band 4, 2-18 GHz, is
obtained by automatically sweeping, in sequence, the
required portions of these three bands.
Power leveling in the 86290A is achieved by sens
ing output power with a broadband directional detec-
5 dBm
tor, comparing this signal to a level control voltage,
and applying the error signal to a PIN modulator.
The modulator precedes the amplifier so the modula
tor's insertion loss is cancelled by the excess smallsignal gain of the amplifier (see box, page 11). The
modulator also contains a directional coupler to
sample the YTO signal and route it to the rear panel for
counting or phase-lock applications.
In general, the conversion efficiency of a harmonic
multiplier drops as the harmonic number n in
creases. Therefore, for a given drive power, the high
est output power is achieved by keeping n as small
as possible. In a sweeper, small n means fewer band
switch points and correspondingly simpler drive cir
cuitry and cleaner display. However, small n implies
greater oscillator range, since the YTO must tune, in
this case, from 2 GHz to an upper limit set by 18/n
GHz. The choice of n = 3 as the maximum for the
86290A was determined by the state of the art in YIG
oscillator technology.
While the multiplier approach is compact and
does not require the development of multi-octave os
cillators, it does pose several tough design con
straints. The YTM and YTO must have excellent tun
ing characteristics to permit good tracking of the de
sired harmonic by the narrow passband of the YIG
filter. Multiplier conversion loss must be 11 dB or
better and the power amplifier must supply at least
100 mWat 2-6. 2 GHz to guarantee the + 5 dBm power
output specification. In addition, the YIG control
circuitry must be very stable with temperature and
time and must compensate for the tuning errors in
herent in magnetically tuned devices.
Tracking YIG Components
Frequency (GHz)
Fig. 2. The 2-18 GHz plug-in provides more than
+5 dBm of power output, internally leveled to within ±09 dB.
External leveling is a/so possible.
Tracking the YTO harmonic with the single-pole
YIG-tuned filter proved to be a challenging design
task. The YTM 1-dB bandwidth, as low as 20 MHz at
some frequencies, dictates how closely the two must
track to avoid unacceptable variations in output
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 3. RF block diagram of the
2-18 GHz plug-in. The YIG-tuned
oscillator and YIG-tuned harmonic
multiplier track each other in fre
The sources of tracking error include magnetic
saturation, hysteresis, and delay, and thermal drift
and aging in both the YIG devices and the control cir
cuits. Some of the magnetic effects can be compen
sated in the control circuits. Others must be con
trolled by optimizing the design of the YIG magnets
Each YIG-device/driver block has its own set of fre
quency errors. One way to guarantee tracking would
be to modify the errors of one block to make it track
the other. Another way is to correct the errors of both
blocks independently, so when they are put together
only very minor adjustments are necessary. This is
the approach taken for the 86290A. The benefits are
many. It simplifies production and service, provides
excellent frequency accuracy, and results in a typical
frequency nonlinearity less than ±8 MHz in bands
1. 2. and 3 and ±30 MHz in the 2-18 GHz band.
Magnet Design
Hysteresis and saturation are familiar effects
whose magnitudes depend upon the magnet core
material. Saturation is also influenced strongly by the
magnet geometry, particularly the shape of the pole.
For the 86290A, the magnet gap and pole face diameter
were chosen based on YIG sphere size, and the opti
mum pole taper for these dimensions was derived.1
Hysteresis is also affected by geometry in that the
overall magnetic hysteresis is a function of the
length of core that is made from high-hysteresis
material and of the flux density in this portion of the
core. Magnet core materials must be chosen to mini
mize both hysteresis and saturation. Of the soft
magnetic materials generally used in these applica
tions, the 50% and 80% nickel alloys are the most
•Nonlmeanty is the difference between the actual output frequency and that indicated by the SWEEP
OUTPUT voltage in the MANUAL mode
The configuration of these materials in the magnet
is complicated by their thermal expansion character
istics. Unless applied correctly, these materials can
contribute significantly, through gap change, to the
overall temperature dependence of the YIG device.
In the 86290A magnets, these materials are used
where required by hysteresis and saturation consider
ations and their lengths are chosen to satisfy a zerogap-drift constraint. Applying this technique to the
design of the YTO and YTM magnet structures re
sulted in the following magnetic properties, mea
sured in bands 1, 2 and 3.
Maximum Hysteresis Error (MHz) ±1 ±2
Maximum Saturation Error (MHz) ±3 ±18
The effect of hysteresis errors is further reduced by
the fact that the two magnet structures are similar
and have similar sweep histories. Thus there is con
siderable cancellation. The remaining nonlinearity
caused by saturation can be compensated by the YIG
control circuits, to be described later.
The third magnetic effect, sweep delay, is caused
by the bucking effect of eddy currents induced in
core and package materials by a change in applied
magnetic field. The lagging field at the YIG causes a
lag in the actual tuned frequency. A large portion of
this delay comes from the magnetic core itself and
cannot be eliminated without laminating the core,
which decreases rigidity and makes hermetic sealing
difficult. However, eddy currents in the circuit sup
port structure can be reduced by avoiding closed
loops and by using thin, high-resistance metals. The
remaining influence is compensated very well by the
control circuits, as explained later.
Magnet gap change is not the only source of ther-
© Copr. 1949-1998 Hewlett-Packard Co.
mal drift in the YIG components. The YIG spheres
themselves and the oscillator transistor in the YTO
can contribute drifts more than an order of magni
tude greater than even an uncompensated magnet.
Careful alignment of the YIG sphere's anisotropic
field and the use of heater/regulators greatly reduce
these effects in the new sweeper (see boxes, pages
6 and 8). Aging, mentioned earlier as a source of
tracking error, results mainly from gradual change
of the gap itself. At 6 GHz, a 0.0025-mm variation in
a 2.5-mm gap causes a frequency shift of 6 MHz. This
sort of variation can be expected over a few weeks time
unless great care is taken in the attachment of the
magnets. YIG magnets for the 86290A are held in
place by strong metal clamps so there is negligible
gap change with time.
YIG Driver Design
Ideally, spherical YIG resonators tune linearly
with applied magnetic field according to the relation
f = yH, where H is the applied field in gauss and y is
approximately 2.8 MHz/gauss (see box, page 13).
Magnetic field is proportional to current through the
magnets, so a linear frequency sweep requires a linear
curreni lamp.
The YIG driver circuits have three main sections:
the circuits that change the tuning current ranges as
the YIG components are swept across the bands, the
driving current sources, and finally, the compensa
tion for the effects of magnetic saturation and delay.
A representative driver schematic is shown in Fig. 4.
A frequency control signal of zero to ten volts is ap
plied for each of bands 1, 2, and 3. Depending on the
band, zero volts represents 2, 6, or 12 GHz, and ten
volts represents 6.2, 12.4, or 18 GHz. Ul and the resis
tive networks around it constitute an inverting ampli
fier with three possible combinations of gain and dc
offset. Each resistive network corresponds to one of
the three bands and is switched into effect by the
band control inputs. For an increasing input voltage,
V falls and the feedback loop comprised of U2,
Ql, Q2, Q3, and Rref assures that Vref=V. The
current through Rref increases proportionally. Ql's
base current is negligible and U2's input im
pedance is high; therefore, the entire current is de
livered to the magnet.
Each YIG device has a driver of this type. Every
range-switching network has two adjustments, and
must be aligned for proper YTO frequency excursion
and optimum YTM harmonic tracking.
If the YIG components were i Heal, this would be
Band 1 Control (T)
Band 2 Control (?)
Band 3 Control (T)
• 20V Frequency Reference ( -20VFR)
Band Switching
Fig. tune resonators. driver circuits provide the linear current ramps that tune the YIG resonators. The
drivers also compensate for magnetic saturation and delay in the YIG magnets.
© Copr. 1949-1998 Hewlett-Packard Co.
A 2-18-GHz YIG-Tuned Multiplier
The YIG-tuned multiplier (YTM) is the key to the 86290A's
broad frequency range. While multipliers have an established
place this microwave design, the important contributions of this
device are its extremely wide input and output frequency
ranges and its efficient generation of harmonic power. Given a
high-level, 2-6.2-GHz drive signal, the YTM provides second
and third harmonic power to 18 GHz with conversion loss
under 11 dB. It can also pass the fundamental signal with ap
proximately 6 dB loss, completing the frequency range without
the need to bypass the YTM. These features, combined with
tuning linearity better than ±18 MHz in any of the harmonic
ranges, make the YTM a very powerful microwave component.
The YTM is shown schematically in Fig. 1. An HP step-reco
very-diode (3RD) is employed as the nonlinear, harmonicgenerating element. It is followed by a single-sphere YIG-tuned
filter, which tracks the desired harmonic and rejects the others.
To minimize conversion loss, the SRD's self-biased condi
tion is modified with an external bias signal. This signal is
shaped from the 86290A's frequency control voltage and is set
in the factory for optimum YTM performance. The 3RD is for
ward biased for 2.0-6.2 GHz output.
Coupling to and from the YIG resonator is accomplished via
conducting loops that pass over opposite sides of the sphere
to ground. These loops are situated at right angles to the ap
plied magnetic tuning field, H, and are at right angles to one
another to avoid direct coupling from input to output. The output
power is proportional to RF current (see box, page 13), so
the 3RD must be driven from a low-impedance source. The
input matching circuit provides this low impedance, trans
forming the 50ÃÃ generator impedance to less than 10Ã1 over a
broad frequency range. It also prevents the flow of harmonic
energy back into the driving source.
The input circuit, the 3RD, and the YIG filter's input coupling
loop are on a sapphire substrate, and the output loop is located
on a second substrate. Beneath each loop is a hole slightly
larger than the YIG sphere. The substrates are placed back to
back with the sphere between the loops.
Temperature compensation of the YIG filter is accomplished
enough. In reality, saturation and delay make addi
tional circuitry necessary. Linearity is improved by
shunting Rref with selected resistors as the YIG fre
quency is increased. The circuit that does this is in
dicated in Fig. 4. As Vref drops with increasing mag
net current, it crosses the level set by the divider, R4
and R5, and the diode conducts. The ratio of R4 and R5
determines the switch point, and their magnitudes
govern the amount of correction applied. Several
such networks are used as needed to reduce nonlinearity of a given YIG device to a few megahertz
across any band.
Delay compensation can be implemented at the
reference resistor or in the range switching circuitry.
The delay mechanism itself can be represented quite
well by a simple RC low-pass circuit (Fig. 5). The cir
cuit's response to a linear ramp is given by
eout(t) = at - aRC(l-e-tRC).
where ein(t) = at. Thus a signal of the form 1 -e~l ic is
2-6.2 GHz
S t e p Y | G
Recovery «. h
Fig. 1.
using two established techniques. First, the sphere is mounted
near a known crystalline axis to reduce temperature depen
dence. Then, when the circuit is in operation, the entire sub
strate assembly is heated to a constant 75°C. A tiny beadstyle thermistor monitors the temperature near the sphere. An
external circuit, mounted on an attached printed circuit board,
supplies a control signal to a resistive heater alongside the
Roben Jo/y
Alejandro Chu
1 J Moll and S Hamilton. "Physical Modeling of the 3RD for Pulse & Harmonic
Generator Circuits . Proceedings of the IEEE. Vol 57. No 7. July 1969
2 "Harmonic Generator Using 3RD & 3RD Modules HP Application Note 920
3 Matthaei. Young, and Jones. Handbook ol Microwave Filter Design", McGrawHill Book Company, 1964
required to compensate the magnet for delay. This is
present in the current flowing through the low-pass
circuit, given by
i(t) = aC[l-e-^c].
Magnetic delay rm is approximately 150 micro
seconds for unlaminated structures of good design.
RC is analogous to rm, so compensation can be real
ized with components of convenient size.
The use of this circuit in an actual driver is illus
trated in Fig. 4. Point A is a virtual ground when con
nected to Ul by the FET switch. Current into this
node causes an increase in frequency. Elements R¡
and C¡ (i = 1 ,2 ,3) are selected to duplicate the magnet
time constant and to provide, in conjunction with the
feedback resistor Rfi, a signal level consistent with the
actual delay effect. A second RC leg may be paralleled
with the first to tailor the compensation more closely
for a particular device.
The selection and loading of the proper R and C val-
© Copr. 1949-1998 Hewlett-Packard Co.
eln = at
Fig. 5. Magnetic delay is repre
sented well by an PC network.
Thus a simple network can pro
vide a compensating signal.
Magnet |
ues for each range could be a tedious affair. In the
86290A, a fixed capacitor, an operational amplifier,
and variable resistors provide a compensating wave
form that is continuously variable in both time con
stant and magnitude, eliminating the nefiH for selec
ted components.
Minimizing Drift
Drift of the frequency control voltage causes a fre
quency error, but since this signal is common to both
drivers, no tracking error occurs. However, great care
must be taken to minimize sources of drift within the
drivers. All operational amplifier offsets are nulled
and ±2 ppm/°C metal film resistors are used in gain
and offset functions. The driver reference resistors are
matched to ±1 ppm/°C and are derated greatly to re
duce self-heating. In the band-switching circuits, the
FETs are operated directly into the high-impedance
operational amplifier to eliminate the effects of FET
on-resistance drift.
Another source of drift is found in the 8620A main
frame's + 20V frequency reference ( + 20VFR), which
is the reference voltage for both current sources. It is
related to drive current by the equation,
+ 20VFR -Vref
Variations in +20VFR may be as high as ±0.4 mV/°C.
To avoid the corresponding frequency and tracking
drifts, the supply is compared to a stable reference in
the 86290A and a correction signal is applied to the
drivers. This technique also makes the instrument
immune to variations in + 20VFR among mainframes.
Frequency Control
Two types of signals control the 86290A output fre
quency: logic signals that select the range of the cur
rent source in the YIG driver, and the 0-10V fre
quency control voltage that drives the current source
across the selected range. The control scheme for
generating these inputs is illustrated in Fig. 6.
For the narrow ranges (2-6.2 GHz, 6-12.4 GHz, and
12-18 GHz), the 0-10V tuning voltage from the 8620A
mainframe is essentially a direct input to the YIG
drivers. In all three uf these ranges, Ql is switched
on, routing the tuning voltage through U4, a voltage
follower, to the drivers. Similarly, the range logic is
determined by direct mainframe inputs: the band 1,
2, and 3 lines are routed through dual-input OR gates
to an 1C that shifts the TTL level to 0-1 0V for driving
the FET switches in the YIG drivers.
When the 2-to-18-GHz range is selected, the 0-10V
tuning voltage from the mainframe is no longer suit
able as a direct input to the YIG drivers. To cover this
range, the drivers must sweep across the narrow
ranges sequentially. This requires that for each
0-10V excursion by the mainframe tuning voltage,
the YIG driver frequency control voltage must
make three 0-10V excursions. Waveforms 4 and 5 in
Fig. 6 show this relationship.
Selection of band 4, 2-18 GHz, causes Ql to turn
off, Q2 to turn on, and Kl to close. The mainframe
tuning voltage is then directed through Kl to the in
puts of two comparators, Ul and U2, which are refer
enced at +2.625 volts and +6.500 volts, respectively.
The combined outputs of these comparators, wave
forms 1,2, and 3 in Fig. 6, indicate the desired YIG
driver range as a function of mainframe tuning vol
tage: 0-2. 625V corresponds to 2-6.2 GHz, 2.6256.500V corresponds to 6-12.4 GHz, and 6.50010.000V corresponds to 12-18 GHz. In addition to
controlling the YIG driver range, these signals switch
offset and gain around U3 to convert the tuning vol
tage into the required frequency control voltage,
waveform 5 in Fig. 6.
From this discussion, it might be concluded that
© Copr. 1949-1998 Hewlett-Packard Co.
A 2.0-6.2-GHz YIG-Tuned Oscillator
The fundamental signal source in the 86290A is a
2.0-6.2-GHz, YIG-tuned transistor oscillator (see Fig. 1). It fea
tures a 20-mW output level, ±3-MHz tuning accuracy, hystere
sis under 2 MHz, and frequency drift of less than 150 kHz/°C.
Fig. 1.
The YTO circuit is built in microstrip on a sapphire substrate.
All transistors are HP 35820. The circuit is shown schematically
in Fig. 2. The oscillator stage consists of Q1, feedback
inductance L1, and the YIG sphere with its coupling loop.
Q1 presents a negative resistance to the resonator over the
desired frequency range. Oscillation amplitude grows until
the average negative resistance seen looking into the emit
ter of Q1 equals the resistance of the YIG resonator.1 The re
sonator detunes very slightly to present a reactance equal and
opposite to that seen at the emitter of Q1 . Q2 and Q3. along with
their matching structures, form a broadband buffer amplifier.2
The five bias resistors are located externally and allow the selec
tion each a unique operating point for each transistor; hence each
stage can be optimized for maximum output power or adjusted
for desired harmonic level.
The choice of YIG sphere parameters is crucial to oscillator
performance. RF coupling increases with sphere size, given a
constant coupling loop diameter. As sphere and loop dia
meters become comparable, however, the uniformity of the RF
coupling field degrades. This enhances spurious YIG reson
ances which show up as sharp frequency discontinuities as the
oscillator is swept. Sphere and loop diameters in the YTO are
0.063 mm and 0.127 mm respectively.
The density of magnetic dipoles within the sphere is deter
mined by its saturation magnetization. This parameter is
bounded on the high side by the intrusion of spurious YIG reson
ances in the lower portion of the band, and on the low side by a
decrease in RF coupling. For the 2-6.2 GHz range, a 600-gauss
sphere is the best compromise.
The greater portion of the YTO's uncompensated frequency
drift is caused by the temperature dependence of the sphere's
anisotropic field. Such drift can be reduced greatly by orienting
the YIG for zero anisotropy. With proper initial orientation, as
many as eight temperature compensated (TC) points may be
found in a single 360° rotation of the sphere about an axis per
pendicular to the applied dc magnetic field. In this oscillator,
the YIG sphere is mounted on a beryllium oxide rod in such an
orientation. It is then rotated until a TC point is found. Finally, a
75°C of is attached to the rod to further reduce the effect of
ambient temperature changes. The whole assembly is sup
ported in a quartz tube to minimize heat losses (see Fig. 1).
The design of the tuning magnet is discussed in the main
text. Magnets are held in place with pairs of metal clamps, and
are hermetically sealed with a stable, moisture-resistant epOxy.
The oscillator is housed in a mu-metal shield can. Heat is trans
ferred from the oscillator circuit via a copper bar that sup
ports the circuit and passes through the wall of the center body
and shielding can to make thermal contact with the 86290A's
aluminum heat sink casting.
Roger Stanc/iff
Paul Hernday
1 P. M Ollivier. "Microwave YIG-Tuned Transistor Oscillator-Amplifier Design
Application to C-Band . IEEE Journal of Solid-State Circuits. Volume SC-7, pp 54-60.
February 1972
2 J Dupre. "1 8 to 4 2 GHz YIG-Tuned Transistor Oscillator with a Wideband Buffer
Amplifier1 , IEEE GMTT Symposium Digest. 1969. pp 432-438
Fig. 2.
© Copr. 1949-1998 Hewlett-Packard Co.
Band 1
2-6.2 GHz Range
6-12.4 GHz Range
12-18 GHz Range
2-18 GHz Sequential
Logic Generator
Control Lines
for YTO and YTM
Magnet Drivers
Frequency Control
Fig. band-select from control circuits take the tuning voltage and band-select signals from the
8620 A mainframe and provide 0-10V ramps that drive the YIG-driver current sources across the
proper to In band 4, 2-18 GHz, the sweep stops momentarily at band-switch points to
avoid frequency and power gaps.
there is considerable overlap — 200 MHz and 400
MHz — at the range switch points. This would be the
case if on each excursion of the frequency control vol
tage the limits were exactly 0 and +10 volts. Actual
ly, the limits are adjusted at U3 so the overlap is less
than 20 MHz at each switch point.
Mainframe/Plug-ln Interface (Band 4)
The 86290A plug-in carries on a two-way com
munication with the 8620A mainframe. To avoid fre
quency and power gaps as it sweeps across three sepa
rate ranges sequentially, the 86290A must have a means
of stopping the sweep generator in the mainframe dur
ing range switching intervals. (This is what causes the
two flat levels in waveform 4 of Fig. 6.) Also, to
achieve fast sweep tracking between YTO and YTM
in sequential operation, the maximum rate of fre
quency change must be no greater than in singlerange operation. Since the 2-to-18-GHz range is
about three times wider than the other ranges, the
plug-in needs a way to reduce the maximum sweep
rate by a factor of three when operating sequentially.
The mechanism for accomplishing both of these
things is shown in Fig. 7.
© Copr. 1949-1998 Hewlett-Packard Co.
8620A Mainframe
Bipolar Current
R e f e r e n c e
86290A Plug-in
Sweep Ramp
_jrVi R
Reduces Forward Sweep Rate
Fig. signals the interface circuits generate stop-sweep signals and limit the maximum
sweep rate in band 4.
To follow the sequence of events, assume that
band 4, 2-18 GHz, has just been selected. The band 4
logic line causes Kl to close and the tuning voltage is
directed through Kl-1 to the switch point compara
tors, Ul and U2 (these are the same comparators
shown in Fig. 6 as a part of the logic generator). The
outputs of Ul and U2 are monitored by an EXCLUSIVE
OR gate, U3, which triggers a monostable circuit
whenever Ul or U2 changes output states. The out
put of this monostable is a six-millisecond pulse,
which is returned to the mainframe. This pulse is in
verted in the mainframe and used to gate open the
path between the current source and the ramp inte
grator. This pulse also goes to the Z-axis output,
where it is available for blanking an oscilloscope dis
play during the stop-sweep interval, thus avoiding
bright spots and spikes that might otherwise ap
pear on the display at the two range switch points.
Continuing around the circuit path, the output of the
sweep ramp integrator is clamped at its two extremes
and amplified to produce a zero-to-ten-volt sweep
output voltage. This ramp voltage is then offset and
attenuated as a function of front-panel control set
tings to generate the desired tuning voltage and com
plete the stop-sweep loop.
The maximum sweep rate in band 4 is fixed by
switching a single resistor in the 86290A, Rl, to
ground. Rl loads the current source reference cir
cuitry to reduce the forward sweep rate reference
voltage, V2, by a factor of three.
The mainframe stop-sweep inverter, U4, has one
other important input path. It comes from the pro
gramming connector on the rear panel. This input is
used in test systems with the 8410B network ana
lyzer (see article, page 15). It allows the 8410B to
stop the sweep oscillator in the same manner that the
86290A does. In a system where both the 86290A and
the 8410B are used, this input is a wired-OR line,
which allows either instrument to stop the sweep.
This is a particularly important feature. It allows the
8410B to make continuous phase and amplitude mea
surements from 2 to 18 GHz when used with the
Leveling Loop
As Fig. 8 shows, 86290A power leveling is accom
plished by controlling the operating point of the PIN
modulator as a function of Vs
and V,pt. V
a voltage derived by detecting and amplifying a small
portion, -16 dB, of the 2-18 GHz output signal. Vset
© Copr. 1949-1998 Hewlett-Packard Co.
(2-6.2 GHz)
RF Output
(2-18 GHz)
Coupler Detector
Band Switching and
Retrace Blanking
27.8 kHz
Fig. operating PIN power :c internally leveled by controlling the operating point of the PIN modulator.
The ALC the is designed to have a high enough slew rate to respond to the ¡>iyu are-wave modu
lation from the 8755/4 Network Analyzer.
is a shaped reference voltage; its magnitude is deter
mined by the front-panel power level control.
In the loop, Vsense and Vset are summed, and their
sum is compared to zero volts at the main amplifier.
A 2.0-6.2-GHz Power Amplifier
The power amplifier used in the 86290A had to meet one ma
jor requirement: it had to provide 20 dBm (100 mW) of power
output to the YIG-tuned multiplier, given an input power of
approximately 10 dBm. The term "power amplifier" might be
surprising, because much higher powers have been achieved
at these frequencies. However, to obtain the power and the
bandwidth at the same time is a real design challenge.
The amplifier design is based on the HP 35820 transistor. The
amplifier has two stages of preamplification that account for 4.5
dB gain, and three stages of power amplification that account
for an additional 5.5 dB. The preamplifier has been des
cribed in detail elsewhere.1
Power amplifier design could have followed any of three ave
nues. A large number of preamplifier stages could have been
paralleled, using power dividers and combiners, or two-stage
power amplifiers with 18 dBm power output could have been
built and paralleled to get the desired output. A third method
was to parallel two transistors within a stage and obtain the 20
dBm output with three stages. Although it required more circuit
optimization, this option had the advantages of lower manufac
turing cost, smaller size, and lower power dissipation, and was
therefore used in the amplifier. The HP 35820 transistor pair was
modeled as a single device to obtain its small-signal S-parameters. The large-signal parameters were obtained experimen
tally using external tuners at high RF power levels. The input, in
terstage and output matching networks were then designed
and the designs verified using a computer.
All amplifier stages are built on sapphire substrates. Fig. 1
shows the amplifier with its lid removed.
Fig. 1.
The amplifier has a rather large small-signal gain between 2
and 3 GHz. Since the amplifier follows the PIN modulator, ampli
fier gain compression detracts from system on/off ratio, an
important parameter when the 86290A is being modulated by
the 8755A Network Analyzer (see text, page 13). To shape the
gain and saturating characteristics, a frequency selective
attenuator consisting of three PIN diodes imbedded in a microstrip is is used at the input of the amplifier. Its attenuation is
set with an external bias voltage and ranges from 2 to 8 dB at
2 GHz. Loss drops to 1 dB above 4 GHz, where compression
is slight.
Ganesh Basawapatna
1 P. Chen, "Design & Applications of 2-6.5 GHz Transistor Amplifiers", IEEE Journal
of Solid-State Circuits. August 1974. Vol SC-9, No. 4
© Copr. 1949-1998 Hewlett-Packard Co.
Product Design of the 86290A
Along with the RF performance goals that defined the
86290A in the early development stage were some equally de
manding mechanical objectives. The plug-in had to fit into what
seemed like an impossibly small space for a 2-18-GHz source
(13 x 15 x 28 cm). It had to be quick and easy to assemble and
every section had to be readily accessible for easy servicing. It
had to to approximately 50 watts; yet, it would have to
run with no more than a 1 5-20°C temperature rise throughout to
ensure good reliability.
The 86290A is a fully modular design consisting of four main
assemblies (Fig. 1): an RF section that houses the microcircuits,
a control section, and front and rear panel assemblies. Com
munication between the control and RF sections takes place via
flexible shielded cables for YTO FM coil drive and ALC detector
signals, and a single ribbon cable from which each microcircuit takes its supply voltages and control signals.
venient replacement of a failed component. As a further service
aid, a printed-circuit extender board is mounted on the under
side of the control section. With the 8620A top cover removed,
this extender can be used to elevate any of the six boards for
convenient testing.
The RF section (Fig. 2) is easily removed and opened. The ex
tra-length interconnect cables allow servicing of the RF com
ponents. A special RF test cable clipped to the side of the RF
section casting aids in servicing.
Product design of the RF section and rear panel focused on
cooling the microcircuits and YIG-dnver components. Since microcircuit reliability was the most important objective, a special
heat sink casting design was undertaken. The result was a
structure that makes maximum use of the air flow from the fan in
the 8620A. The air passes through this casting first, then over
the circuit boards and through the rear panel, where a finned
Fig. 2.
Fig. 1.
Printed-circuit boards are dedicated to separate electrical
functions for optimum servicing and minimum repair cost. They
are interconnected by a single, multilayer printed-circuit mother
board. Edge connectors at the ends of this board form the sole
interface with the front and rear panel assemblies.
This design greatly streamlines production of the instrument.
The RF section is preassembled, connected to a special test
set, and fully RF tested before being installed in the instrument.
Front and rear panel assemblies are processed similarly.
Printed-circuit boards are dc pretested by computer and then
functionally pretested in a simulated instrument before being
plugged into the final instrument. These techniques allow
troubleshooting at the lowest, simplest levels, and result in little
if any troubleshooting at final turn-on.
The same modular features result in excellent serviceability.
The front and rear panels come off with the removal of four
screws. They can be tilted outward and tested with the instru
ment operating normally, and will then readily unplug for con
casting provides heat sinking for the YIG-dnver reference resis
tors and driver transistors.
The main text describes the care taken to provide excellent
frequency accuracy and ensure tracking between the YTM
filter and the desired YTO harmonic. Absolute delay compensa
tion is built into the YTO-driver pair. As a result, the YTO and its
driver are treated as an assembly. If a YTO fails, a new YTO and
dedicated driver board are installed. Adjustments can be done
with basic test equipment.
Mounted atop the RF section are three labels. One explains a
simple service technique for aligning YTO and YTM for opti
mum frequency accuracy and tracking. The others call out re
sistor values selected at the factory for YTO and YTM linearity
compensation— a reference, should a part be misplaced
during servicing.
Any deviation from zero is amplified and sent to a
modulator driver stage which shifts the modulator
operating point, changing Vsense so that its sum
with Vset is returned to zero.
In a swept, broadband leveling application such as
this, unflat power sensing elements can degrade the
maximum leveled power if their frequency responses
are not properly compensated. For the 86290A,
this compensation is provided by shaping the inter
nally generated 1-volt/GHz sweep reference voltage
and using that waveform as one input to the power
reference amplifier shown in Fig. 8. Shaping is ad
justed once in each instrument to compensate for
variations in directional detector sensitivity. The
corresponding frequency-related variations in Vset
force the loop to make the desired correction.
William Misson
Billy Knorpp
© Copr. 1949-1998 Hewlett-Packard Co.
How YIG Tuning Works
The resonators in the 86290A's wide-band oscillator and tun
able multiplier are tiny 0.5-0.6-mm spheres fabricated from
single crystals of the ferrite Yttrium-Iron-Garnet (YIG).
Placed in an RF coupling structure in a dc magnetic field,
highly polished YIG spheres exhibit a high-Q resonance at a fre
quency proportional to the dc field. When electromagnets are
used to produce this field, the resonant frequency is proportion
al to magnet current. While YIG of other shapes is useful in mi
crowave design, spheres are used here because of their linear
tuning characteristic.
To understand the phenomenon of ferri magnetic resonance,
consider diagrams (a) through (d). In the ferrite with no dc
magnetic field applied, there is a high density of randomly
oriented magnetic dipoles, each consisting of the minute cur
rent loop formed by a spinning electron. When a dc magnetic
field, H0, of sufficient magnitude is applied, the dipoles align
parallel to the applied field, producing a strong net magne
tization, M0, in the direction of H0. If an RF magnetic field is
applied at right angles to H0, the net magnetization vector will
(a) Randomly oriented
magnetic dipoles in
the unmagnetized
(b) The effect of a dc
magnetic field on mag
netic dipoles in the
(c) Precession of net
magnetization vector
due to RF magnetic
(d) YIG bandpass filter
showing RF coupling
loops, sphere, and
magnetic poles.
As with any high-performance ALC loop in a swept
source, the 86290A loop had to have sufficient band
width that power variations as a function of sweep
rate would be negligible. Also, a primary design goal
was to achieve loop slew rates great enough that the
instrument could be modulated directly when used
with the 8755A Network Analyzer (see article, page
15). This would eliminate the need for the external
modulator normally required in this application, and
as a consequence, deliver the full +5 dBm to the de
vice under test.
Satisfying this requirement meant that the loop
would have to respond to the 27.8-kHz square wave
precess, at the frequency of the RF field, about an axis coin
cident with H0. The processing magnetization vector may be
represented as the sum of M0 and two sinusoidally varying RF
magnetization components mA and my. The angle of pre
cession <)>, and therefore the magnitudes of mx and my, will be
small except at the natural precession frequency. This fre
quency, known as the ferrimagnetic resonant frequency, is a
linear function of the dc field H0.
Diagram (d) shows the basic elements of a YIG bandpass
filter.1 The filter consists of a YIG sphere at the center of two
loops, whose axes are perpendicular to each other and to the
dc field H0. One loop carries the RF input current, and the other
loop is connected to the load. When H0 is zero, the mutually
perpendicular orientation of the loops results in large input-tooutput isolation. With H0 applied, there is a net magnetization
vector in the direction of H0. The magnetic field hx produced by
the RF driving current in the input loop causes the net magne
tization vector to precess about the z-axis. The resulting RF
magnetization component, my, induces a voltage in the output
loop. fre frequencies away from the ferrimagnetic resonant fre
quency, my and the voltage it induces are small, so input-tooutput isolation is high. When the input current is at the ferri
magnetic resonant frequency, <S> and my are maximum. There
is a large transfer of power from input to output, and insertion
loss is low. Thus the filter center frequency is the ferrimagnetic
resonant frequency and can be tuned by varying H0 with
rr.agnet current
In oscillator designs, a single loop couples the YiG spheie ic
an active circuit having a negative resistance over the desired
frequency range. The resonator can be modeled as a parallel
RLC circuit. When the resonator resistance is greater than the
negative resistance of the active circuit, and the imaginary
parts of the two impedances are equal and opposite, osci llation
will occur. Oscillation amplitude will grow until the average
values of the resonant and negative resistances are equal.
Frequency stability in both filter and oscillator designs is
affected by the temperature dependence of the YIG sphere's
inherent anistropic field. The contribution of anisotropy to re
sonant frequency is dependent, in turn, on the orientation of the
sphere's crystalline axes with respect to H0, the dc magnetic
field. The sphere can be rotated such that this contribution
is zero.
'For 'Magnetically aetailed treatment see. for instance, PS Carter, Jr . 'Magnetically Tunable
Microwave Filters Using Single-Crystal Yttrium-lron-Garnet Resonators IRE
Transactions on Microwave Theory and Techniques. Vol MTT-9. No 3, May 1961
modulating signal produced by the 8755A. This sig
nal would drive the loop into and out of saturation for
a minimum on/off ratio of 20 dB. Also, the loop would
have to come out of saturation fast enough to main
tain no worse than 45/55 symmetry. To add a further
complication, equally good performance was ex
pected over the 10-dB range of the power level control.
In general, as modulator attenuation approaches
saturation in an ALC loop, the corresponding loss of
sensitivity in control current constitutes a loss in
loop gain, and hence a loss of bandwidth and slew
rate. However, the slewing capability of the modula
tor itself is not reduced. Consequently, a technique
© Copr. 1949-1998 Hewlett-Packard Co.
Knorpp designed the RF section heat sink casting.
Dick Bingham designed the FM driver board. Jack
Kuhlman provided the coupler/modulator design.
Phil Chen did the early YTO design. Industrial de
sign support came from Dave Eng and Roy Church.
We also wish to thank Callum Logan, Jeff Gomer and
Jim Yarnell for fabricated part process development.
The lab team had important assistance in micro
electronic technology from Weldon Jackson, Yeng
Wong, Jim Smith and Pete Planting. Phil Froess and
Harry Portwood of the transistor fabrication group
helped ensure a good supply of high-performance
Production Engineering support for the 86290A is
provided by John Turner at the instrument level and
Alejandro Chu and Val Peterson for the components.
Additional help has been given by Jeff Ho and Alan
Kafton, and by Bud Edgar for the 8620A modification.
The 86290A operation and service manual was
written by Doug Andrus and Don Jackson. Jim Arnold
set up the fine instrument support program.
We aré indebted to Cyril Yansouni and Jack Dupre
for their leadership and encouragement over the
several years of developmental work. 2
for achieving a uniformly fast slew rate, even while
approaching saturation, involves alternately switch
ing between closed and open-loop control of the mo
dulator drive current. Ideally, for a pulse input, the
loop would remain closed until it began to lose gain
as the modulator neared saturation. At this point the
loop would be switched open, and the modulator
would be driven on into saturation by bypassing the
main amplifier and driving the modulator directly.
In recovering from maximum attenuation, the re
verse sequence would occur. The 86290A ALC loop
uses this technique to satisfy the 8755A modulation
requirement. Level sensing circuitry establishes the
attenuation level at which the closed-to-open-loop
transition occurs.
The authors gratefully acknowledge the contri
butions of the many talented people who cooperated
very effectively to make the 86290A a reality. The de
tailed YTM concept was provided by Kit Keiter, and
Robert Joly did the early design and optimization. Earl
Heldt did much of the early magnet and circuit pack
age designs for the YIG tuned devices. Alejandro Chu
made several important improvements in YTM per
formance, did the original ALC design, and gave valu
able help in many other areas, including FET switch
ing and delay compensation in the YIG drivers.
Ganesh Basawapatna designed the 2-6.2 GHz power
amplifier and contributed greatly to several other RF
designs. Roger Stancliff made valuable improve
ments in the ALC design and did final design of the
YTO. Gary Holmlund did the exacting YTO and YTM
driver designs and helped to coordinate the many
changes that took place during the prototype phase.
Bill Misson did the excellent product design and Bill
1. Y. Ishikawa and S. Chikazumi, "Design of High Power
Electromagnets", Japanese Journal of Applied Physics,
Volume 1, No. 3, September 1962.
© Copr. 1949-1998 Hewlett-Packard Co.
Broadband Swept Network Measurements
Vector and scalar measurements can be made from 2 to
18 GHz in one sweep by pairing the new 2-18-GHz sweep
oscillator with other microwave instruments.
by John J. Dupre and Cyril J. Yansouni
BROADBAND COVERAGE, precision frequen
cy characteristics, and compactness make the
new 8620A/86290A Sweep Oscillator (see article,
page 2) suitable for a wide variety of network mea
surement applications. Its internally leveled output
power and low harmonic and spurious levels over
the 2-to-18-GHz frequency range are important in the
measurement of broadband microwave components.
Its frequency accuracy and stability are important for
testing narrow-band, higb-O devices.
frequency by the frequency reference voltage. The
VTO searches around that frequency until phaselock is achieved. The sweep oscillator then begins its
sweep with the analyzer maintaining phase-lock.
When the VTO reaches the end of its range, the ana
lyzer commands the sweeper to stop sweeping and
the search and lock procedure is repeated before the
sweep is resumed.
An example of this capability is shown in Fig. 2.
The input reflection coefficient of a ferrite junction
circulator from 2 to 18 GH/ is displayed in polar for
mat. Although designed primarily for the 5-to-10-GHz
range, its out-of-band performance including the
phase of the reflection coefficient is often important.
This display allows convenient adjustment of inband and out-of-band parameters.
Transmission gain and phase may also be mea
sured with the system of Fig. 1. Since the system is
frequency selective and therefore unaffected by
sweeper harmonic output, dynamic range is 60 dB.
Vector Network Measurements
Complete characterization of a network requires
measurement of the magnitude and phase of its driv
ing point and transfer characteristics. Phase informa
tion is vital for component design and often neces
sary for system, component, and antenna testing. At
microwave frequencies, scattering parameters1 are
widely used because they relate incident to reflected
and transmitted waves at the network terminals.
Measurement of network scattering parameters in
one continuous 2-to-18-GHz sweep is now possible
with a new Network Analyzer, the HP 8410B, and the
8620A/86290A Sweep Oscillator. The HP 8410B, like
its predecessor, the 8410A,2 measures complex ra
tios by translating the microwave signal to an IF by a
sampling process. A phase-lock loop locks a harmon
ic of the VTO that drives the sampler to the micro
wave signal. The contribution of the 8410B is new cir
cuitry that interfaces with the 8620A Sweep Oscilla
tor and allows phase-locking over broad frequency
sweeps instead of only over octave ranges.
Three interface lines are necessary for reliable
broadband phase-locking (Fig. 1). A frequency refer
ence line from the oscillator provides an analog vol
tage corresponding to the output frequency with an
accuracy of ±35 MHz. A blanking line provides a
pulse during retrace and at band-switch points. Final
ly, a stop-sweep line from the analyzer can command
the 86290A to stop sweeping momentarily. At the be
ginning of each sweep and at band-switch points, the
analyzer VTO is tuned approximately to the locking
Scalar Network Measurements
For a broad range of applications, amplitude-only
or scalar measurements are sufficient to characterize
841 4A Polar
Sweep Oscillator
841 OB
Analyzer «41 2A Phase/
8743A Option 18
Test Set
Under Test
Fig. 1. A 2-to-18-GHz scattering parameter measurement
© Copr. 1949-1998 Hewlett-Packard Co.
achieved by direct modulation of the sweep oscilla
tor by the 8755A Analyzer instead of an external mo
dulator. This external modulation capability of the
sweeper (see article, page 2) makes its full power
available at the test device and leads to greater mea
surement dynamic range.
A typical setup for simultaneous reflection and
transmission measurements is shown in Fig. 3. The
dual directional coupler covers the 2-to-18-GHz fre
quency range with a directivity better than 30 dB at
2 GHz and 26 dB at 18 GHz. The 8755A displays simul
taneously the return loss (A/R) and the transmission
characteristic (B/R) of the device under test (Fig. 4).
Absolute power measurement is also possible by dis
playing any one of the three channels (R, A, B).
For accurate transmission measurement, typically
low-loss measurements (cable loss, attenuation,
etc.), one would use the power splitter in an alterna
tive setup, also shown in Fig. 3 . The power splitter ex
hibits better than 0.25 dB tracking between the two
arms over the full dc-to-18-GHz frequency range.
Coupled with the ratio capability of the 8755A, it pro
vides a means of accurately measuring insertion loss
without having to measure and correct for system er
rors. The equivalent output SWR of this splitter
when used either for ratio measurement or as a
source leveling device is typically better than 1.25
over the same frequency range. Thus the power split
ter minimizes measurement uncertainty caused by
source mismatch.
Fig. 2. Reflection coefficient of a ferrite junction circulator
from 2 to 78 GHz. Full scale is unity reflection.
the component under test. This is the case for inser
tion loss or gain and return loss or reflection coeffi
cient (magnitude) measurements.3 The 8755A Swept
Amplitude Analyzer4 allows for simultaneous dis
play of return loss and transmission parameters. When
it is used with the 8620A/86290A Sweep Oscillator
and the 11692D Dual Directional Coupler, this simul
taneous measurement can be made continuously
over the entire 2-to-18-GHz range. For precision
transmission measurement (low insertion loss), the
new 11667A Power Splitter can be used. In both
cases, the required modulation of the signal can be
Sweep Oscillator
Reference Channel
Reflection Channel
11692D Dual
Under Test
Det 11664A Detector
11 667 A Power
Under Test
Fig. 3. System for simultaneously
measuring transmission and re
flection parameters.
© Copr. 1949-1998 Hewlett-Packard Co.
Paul R. Hernday
! i"W^^ Paul Hernday, project manager
- ->~:· for the 86290A RF Plug-in, re
ceived his BSEE degree from the
University of Wisconsin (Madison)
in 1968 and joined HP in 1969.
Besides the 86290A, he's worked
on analyzers and Gunn oscil
lators and. before joining HP.
on Van de Graff accelerators.
.JQ I He's a member of IEEE. A native
jl^ | of southern California, Paul is
^ I married, has two small children,
and lives in Santa Rosa, where he
serves on the board of directors
of the Family Service Agency of
Sonoma County. A major interest, shared by his wife, is the
evolution of life styles that emphasize human values and a res
pect for the natural environment. Paul also enjoys hiking and
camping with his family, woodworking, gardening, reading,
and keeping up with developments in many areas of science.
Fig. 4. Frequency response of a bandpass filter measured
by the system of Fig. 3.
We would like to acknowledge the contributions
of Doug Rytting, Bruce Donecker, Lewis Newton,
and Richard Barg for the 8410B Network Analyzer.
The 11667A Power Splitter was designed by Julius
Botka and Brent Palmer. Jim Davis, Russ Johnson, and
Jim Kaylor have contributed in the area of network
analyzer applications.
1. R.W. Anderson, "S-Parameter Techniques for Faster,
More Accurate Network Design," Hewlett-Packard
Journal, February 1967.
2. R.W. Anderson and O.T. Dennison, "An Advanced New
Network Analyzer for Sweep-Measuring Amplitude and
Phase from 0.1 to 12.4 GHz," Hewlett-Packard Journal,
February 1967.
3. "High-Frequency Swept Measurements," HewlettPackard Application Note 183.
4. H. Vifian, F. David, and W. Frederick, "A 'Voltmeter'
for the Microwave Engineer," Hewlett-Packard Journal,
November 1972.
John J. Dupre
Jack Dupre is section manager for
sweep oscillators at HP's Santa
Rosa (California) Division. He's
been designing YIG-tuned oscil
lators and other components for
sweepers and spectrum ana
lyzers since he came to HP in
1964. just after receiving his BS
degree in electronic engineering
from California Polytechnic Uni
versity (San Luis Obispo). Along
the way he's earned his MSEE
degree at Stanford University
(1967) and been awarded a
patent for a YIG-tuned Gunneffect oscillator design. He's a member of IEEE. Jack was born
in Louisville, Kentucky. He and his wife and three children live
in Santa Rosa and enjoy traveling and photography.
Carl J. Enlow
Carl Enlow has been with HP since
1967, designing various parts
of the 8620 Sweep Oscillator
family. He's now project manager
for the 8620A.86290A interface
ana is responsible tor introducing
the 86290A to production. Born in
Kilgore, Texas. Carl spent three
years as a guided missile tech
nician in the U.S. Navy before
entering the University of Illinois
(Champaign-Urbana) to work for
his BSEE degree, which he re
ceived in 1 967. Now living in Santa
Rosa, California, Carl is married
and has a son and a daughter. He's an active tennis player and
golfer, likes to play the stock market, and enjoys seeing new
places with the family and participating in Indian Guide ac
tivities with his son.
Cyril J. Yansouni
Y a n s o u n i i s e n g i n e e r i n g
manager for sweep oscillators and
^^^~ network analyzers at HP's Santa
Rosa Division With HP since
1967, he's been a sweeper de
signer, YTO and amplifier project
leader, and sweeper section
manager. Born in Alexandria,
Egypt. Cyril graduated from the
Catholic University of Louvain,
Belgium in 1965 with a degree in
electrical and mechanical en
gineering, and in 1967 received
his MSEE degree from Stanford
University. Cyril is married and has
two children. He enjoys traveling with his family, and usually
encounters no language barriers; he's fluent in English, French,
Italian, and Arabic. A tennis player and music lover, Cyril serves
on the board of directors of the Santa Rosa Symphony
Orchestra and on the electronic advisory committee of Santa
Rosa Junior College.
© Copr. 1949-1998 Hewlett-Packard Co.
C y r i l
The Dual Function Generator: A Source
of a Wide Variety of Test Signals
FM signals, AM signals, dc levels, tone bursts, pulses and
ramps, in addition to sine, square, and triangular waves,
are produced by this new dual-source function generator.
by Ronald J. Riedel and Dan D. Danielson
audio, ultrasonic, video, and RF fields require
the engineer to have more than one signal source
available. Such would be the case when sweep-fre
quency testing amplifiers and filters, testing modula
tion and detections systems, evaluating discrimina
tors, and simulating transducer outputs.
With this in mind, we set out to design a truly
general-purpose signal source that in many cases
could replace combinations of sine-wave oscil
lators, signal generators, sweep oscillators, and
pulse generators now used. Such an instrument, be
sides the obvious cost savings, would enhance user
convenience by making the various test signals
pushbutton selectable so the user wouldn't have to
sort out and reconnect wires and cables. Neither
would he have to worry about signal level incompati
bility or logic threshold mismatch.
The result of this design effort is the Model 3312A
Function Generator (Fig. 1). This instrument has two
independent waveform generators that can be used
separately, or in combination with one modulating,
gating, sweeping, or otherwise controlling the other.
The main generator operates overa range of 0.1 Hz to
13 MHz and the other generator, called the modula
tion generator, covers a range of 0.01 Hz to 10 kHz,
Fig. 1 . Model 331 2A Function Generator combines two signal
sources to produce a wide variety of test signals. Total fre
quency coverage is from 0.01 Hz to 13 MHz.
giving a total coverage of 9 decades for the two
Like other HP function generators, the new Model
3312A produces sine, square, and triangular wave
forms of high quality (Fig. 2a). It also generates
pulses and ramps with continuous control of the
Fig. 2. Oscillogram (a), made by
a sampling scope of the triangle
waveform, demonstrates pre
servation of good waveform shape
at high frequencies, in this case
10 MHz (sweep time is 20 ns/div
and vertical deflection factor is
1 V/div). Multiple exposure oscillogram (b) shows the main gen
erator output at 100 kHz modu
lated by various internallygenerated waveforms at 1 kHz.
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 3. Multiple exposure oscillogram (a) shows pulse trains
generated in the single-cycle
and multiple-cycle modes: the
bottom trace shows the internallygenerated gating waveform.
Oscillogram (b) shows how the
waveform is a/ways completed in
the multiple-cycle mode (top)
and single-cycle mode (center):
the gating waveform is at bot
tom. The start/stop phase is se
lectable over a range of -80°
to +90°.
ratio between up and down slope of the ramps or the on
and off times of the pulses (see box, page 20). Also
generated are one-volt rectangular waves with a rise
time of less than 10 ns, available from the main genera
tor's SYNC output. These are useful not only as triggers
for oscilloscopes and counters, but also as a source of
fast rise pulses for testing logic circuits and for lowlevel testing of the transient response of wideband
A 60-dB step attenuator and a 20-dB vernier give
the main generator an amplitude range of 1 mV to
10 V into a 50Q toad (20V into an open circuit). A dc
offset voltage can be added to the output. The output
of the modulation generator is fixed at 1 volt into a
high impedance.
gle-cycle or multiple-cycle modes of operation.
These modes greatly expand the variety of wave
forms that the new Function Generator can produce.
Narrow, low-duty-cycle pulse trains can be obtained
by setting the main generator to a relatively high fre
quency and using the trigger mode with a low-fre
quency gating signal, as shown in Fig. 3a. Pulse
bursts, useful for testing counting circuits, are ob
tained by using the multiple-cycle mode.
When operating in the single cycle or multiplecycle modes, the starting phase can be adjusted over a
range of +90" to —60". The generator always com
pletes the last cycle and stops on the same phase on
which it started (Fig. 3b).
These modes can be used at the same time as the
AM and FM modes, making it possible to generate a
wide variety of waveforms such as tone bursts or the
"chirp" waveform shown in Fig. 4a.
The modulation generator can also sweep the
main generator up to two decades in frequency at
sweep rates as slow as 1 sweep per 100 seconds, use
ful for plotting frequency response on X-Y recorders,
or at rates up to 100 sweeps per second for oscillo
scope display (Fig. 4b). The gating mode can be used
in combination with this mode to provide retrace
Multiple Combinations
The two generators can be used separately for tests
of multiple input devices such as modulators. On the
other hand, the main generator can be amplitude or fre
quency modulated by the second generator (or an ex
ternal source), giving the user a built-in choice of
sine, square, or triangular waveform modulation
(Fig. 2b). Square-wave modulation in the FM mode,
for example, produces an FSK (frequency-shift-key
ing) waveform. Triangular modulation provides a
convenient waveform for testing detector linearity.
The modulation generator (or an external source)
can also gate or trigger the main generator to give sin-
What's Inside
A simplified block diagram of the new function
Fig. 4. 'Chirp" or swept-burst
waveform, upper trace in (a), is
made by sweeping the main
generator with the ramp from the
modulation generator, shown in
the lower trace, and gating it off
during retrace. Sweep is from 7 to
70 kHz. Oscillogram (b) shows the
output of a series R-L-C circuit
driven from 3 kHz to 1 MHz by the
main generator (vertical scale is
100 mV/div). The oscilloscope
horizontal axis was driven by the
ramp output of the modulation
© Copr. 1949-1998 Hewlett-Packard Co.
Variable Symmetry with Constant
The versatility of a function generator can be increased by
equipping it to generate pulses and ramps in addition to its
other waveforms. Pulses and ramps are produced by varying
the duty cycle of the square and triangular waveforms How
ever, it would also be desirable to maintain a constant repeti
tion rate as the duty cycle varies so that frequency may be read
directly from the front-panel controls. This can be done using
only a single control with the circuit in the diagram below.
In the diagram, the period T, of the positive slope of the
triangle is proportional to 1/1, while T2, the period of the nega
tive slope, is proportional to 1/I2. \-¡ and I2are in turn linear func
tions of e, and e2 thus:
k/T, = e, and k/T2 = e2
Where a is the fractional rotation of the symmetry control, and
ranges from 0 to 1 .
Substituting (4) and (5) into (3), we have:
where k is a constant determined by the frequency range
To maintain constant frequency f0,
1-a) R, +Ra
e0R2 kf0
Combining terms yields:
TI + T2 = 1/f0
Therefore from (1) and (2), we have:
1 + 1 = 1
e i e o k f
Thus, the frequency f0 is completely independent of a, the
symmetry control setting.
e, and e2 may vary over the range:
To maintain a constant frequency while the symmetry is
varied, e, and e2 must be varied in sucha way as to keep the left
side of equation (3) constant. Since the relationship between
e, and e2 is not linear, this requirement could represent some
difficulty in realization
The rather simple circuit shown in the diagram solves this
problem nicely. It is easily seen that:
,(ore2) =ï e0
In the Model 331 2A, this range was selected to provide a duty
cycle range of 20% to 80%.
tor Cl through constant-current sources II and 12, a
technique similar to that used in other HP function
generators.1 2'3 The direction of current flow is deter
mined by the diode current switch under control of
the comparator. The comparator, in turn, is set or
generator is shown in Fig. 5. The heart of the instru
ment is the main triangle generator, shown in more
detail in Fig. 6.
As shown in Fig. 6. a triangular waveform is gener
ated by alternately charging and discharging capaci
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 5. the diagram of the Model 3312 A Function Generator. The frequency of the main
generaioi is controlled by a front-panel knob, by the modulation generator, by an external vol
tage, or by any combination of the three.
reset as the triangular waveform reaches a positive or
negative limit. Positive feedback around the compar
ator provides hysteresis that sets the limits and thus
the peak-to-peak amplitude of the triangle.
Frequency ranges are selected by switching capaci
tor Cl and changing resistors in the current sources.
Frequency control within each range is accom
plished by varying the control voltage applied to the
current sources. As shown in the block diagram of
Fig. 5, this may be done with the front-panel fre
quency dial, by applying a voltage from the modula
tion generator, or with an external voltage (applied
through a rear-panel connector). An external voltage
can control the frequency over a 1000:1 range, thus
enabling the generator to serve as a highly linear vol
tage-controlled oscillator in a phase-lock loop or in
voltage-to-frequency converter applications.
High-Quality Waveforms
The triangular waveform is shaped into a sine
wave with less than 0.5% distortion (below 50 kHz)
by a 12-diode nonlinear shaping network.3 The per-
To Function
Switches <«< j I,
To Sine
Shaper and
v v
Feedback to Comparator
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 6. Main generator produces
triangle waves and square waves
as part of its normal operation. The
square wave is divided down to
1 volt peak for the sync output.
Asymmetrical control of currents
n and 12 provides a means of
varying the up-down ratio of the
triangle (see box on opposite
formance of most diode sine shapers tends to de
grade seriously at frequencies of several megahertz
and higher, particularly with respect to third harmon
ic distortion. In the Model 3312A, the sine shaper is
followed by a low-pass LC filter that has a sharp cut
off at 14 MHz. This filter attenuates third harmonics
substantially at generator frequencies above 5 MHz,
enabling the instrument to produce sine waves with
all harmonics more than 30 dB below the fundamen
tal up to its maximum frequency.
The output amplifier has separate parallel paths
for the high and low frequencies. This gives it wide
bandwidth and a high slew rate to maintain good
square and triangle wave shape without compro
mising dc stability and low offset. An integratedcircuit operational amplifier is used for the low fre
quencies while the ac-coupled high-frequency path
is optimized for wide bandwidth. The two paths are
summed in the final gain stage and then buffered
through cascaded emitter-followers to the output.
The sine wave is derived by shaping the triangular
wave as in the main generator.
Currents for charging and discharging the integra
tor are supplied through the diode-resistor networks
at the integrator input. The range of current provided
by the FREQUENCY control gives a 100:1 frequency
span on any range. The symmetry control, however,
affects only the run-down time so frequency is also af
fected by this control.
In the sweep mode, these networks supply un
equal currents, giving a 90:10 ratio of up to down
times. At the same time, the reference input to the
comparator is changed so that the ramp operates be
tween — 9 and 0 volts, rather than the symmetrical
±5-volt range of the triangle. This places the top end
of a frequency sweep at the frequency indicated on the
main generator's frequency dial. The bottom end is
then determined by the setting of the modulation
level control. In the FM mode, on the other hand, the
modulating waveforms are centered on the zero axis
and thus give frequency modulation (up to ±5%)
centered on the frequency set on the main tuning
When operating in the sweep mode, the symmetry
control affects the "retrace" time. This can be extend
ed up to 10x its normal time.
To facilitate setting up a frequency sweep, a "0
Hz" position is provided on the modulation fre
quency range switch. In this position, the modula
tion generator ramps down to its lowest level and
then stops. Then, when the frequency range switch is
moved to one of the other positions, the main genera
tor starts at the lower limit and sweeps up.
The "0 Hz" position is obtained by the circuit driv-
Slow Ramps
The design of the modulation generator differs
from that of the main generator because it has to
generate very slow ramps for the sweep function
while it does not have to operate to frequencies as
high as the main generator. Therefore, it was de
signed around an integrator that effectively gives the
large capacitance needed for the very slow ramps.4
A block diagram is shown in Fig. 7. The triangular
wave is generated by integrating the square wave.
The triangle level is compared to the square wave
level and when the two are equal, the comparator
switches the square wave generator to its other state.
0 Hz Control
TTL Signal
Gating Logic
Fig. frequency Reset generator uses an integrator lor very low frequency operation. Reset level and
symmetry are altered for ramp waveforms.
© Copr. 1949-1998 Hewlett-Packard Co.
From Main
Generator Comparator
From Modulation >
Enable To Integrating Capacitor
C1 of Main Generator
From Triangle Buffer
Amplifier of Main Generator
Fig. operating signal control circuit prevents the main generator from operating when a gate signal is
not present. (The switches are on the rear panel.)
the waveforms may be applied to the modulator, it is
possible to have sine, square, or triangle modulation
of sine, square, or triangle carriers.
If none of the modulation function buttons (sine,
square, or triangle) is depressed, an external wave
form may be applied to the front-panel connector for
amplitude or frequency modulating the main genera
tor, as selected by the modulation mode buttons (AM
or FM).
ing the FET switch Q2 on the integrator shown in
Fig. 7. When the range switch is set to "0 Hz", one in
put to gate 1 goes high. Assuming the other input is
high, the output of gate 1 goes low, Ql is turned on
which in turn turns on Q2, clamping the integrator at
its lowest level. Gates 1 and 2 form a flip-flop that pre
vents Q2 from being turned on until the modulation
generator completes its current cycle.
Versatile Modulation
Gate and Trigger Functions
The waveforms generated by the modulation
generator are available at a front-panel connector. In
AM or FM operation, the signal at this port is the
sine, triangle, or square wave selected by the modula
tion function buttons. In the sweep mode, the output
is the sweep ramp which is then available for driving
the horizontal axis of a scope or X-Y recorder during
a frequency sweep.
For AM modulation, the main and modulation sig
nals are routed to an integrated circuit balanced mod
ulator. The modulator output is amplified and then
routed to the main output amplifier. Because any of
The versatility of the Model 3312A is greatly en
hanced by the gate and trigger functions. These al
low the main generator to be gated on and off under
control of either the modulation generator or an exter
nal source.
A block diagram of the trigger/gate circuit is
shown in Fig. 8. The gating amplifier shown closes a
negative feedback loop around the main triangle
generator, preventing oscillation. Whenever the gat
ing amplifier is disabled by the gating signal, the
main generator is released to generate the output
± ION Rise or fall t.me (tos to W.j • 10 nsec Duty cycle vanes
HP Model 3312A Function Generator
DC OFFSET -10 volt:
Main Generator
S-ne square triangle - ramp pulse sweep trigger gate AM and FM
RANGE 0 t Hz to 13 MHz m B decade ranges
DIAL ACCURACY zSSol tul scale
SQUARE WAVE RISE OR FALL TIME (10*. to 90", j 18 rvec at ful rated
VARIABLE SYMMETRY 80.20» to i MHZ Pustvng me CAL but»» results
SINE WAVE DISTORTION • 0 5'= THD from 10 Hz » 50 kHz 30 OB be*O*
'uncamemai from 50 kHz to 1 3 MHz
LEVEL 20 V p-p <nio open orcuil 10 V p-p mto SOU
LEVEL FLATNESS (SINE WAVEl • 3*. from 10 Hz to 100 kHz at lull rated
output M kHz reference) • 10". from 1 00 KHz to 1 0 MHz at Ml rated output
ATTENUATOR 1 1 101 100 1 ana 1 000 1 Vemer gives 10 1 corWrtuOus
SYNC OUTPUT 1 V p-p square wave into open cvcuM Impedance 5011
«tne DC
ir setting Pusnmg me CAL button n
offset Instantaneous ac vortag* • v dC offset kmfted to sifor
RANGE 10001 on any range
INPUT REQUIREMENT Witnctal setal 10. Oto -2V ± 20*» linearly t*
frequency iOOO 1 An ac voltage modulates the frequency about a o»
setting within me kmts I 1 • t- 10) « range setting
LINEARITY Ratio of ou«>ut frequency to mput voRage (At AV) is knear «rm.fi
0 5*c over a tOO 1 frequency range
TYPES AM FM Sweep Trigger Gate or Burst internal or e
DEPTH 0 to 100°.
EXTERNAL SENSITIVITY 1 0 V (K> f or 1 00% w
CARRIER ENVELOPE DISTORTION 2*. at 70*. sine wave modulation w*i
DtSTQRTION 35 dB al tc 10 MHz. '„, = 1 kHz 10*. n
© Copr. 1949-1998 Hewlett-Packard Co.
OUTPUT LEVEL • 1 0 V p-p mÃ-o 10 «1
SPECTRAL PURITY Sin» Wave Distortion • 2". THD Irom to Hz to 10 kHz
SWEEP WIDTH tOO 1 on any range
SWEEP RATE OOTHzlolOOHz ,9010rarrp . 0 rtt provxM* manual Mtkng at
Sweep Start with Modulation Generator operation suppressed
SWEEP MODE Repetitive linear sweep between start and Mop frequency
settings Retrace time can be increased wrth symmetry control
FREQUENCY RANGE 0 1 Hz to 1 MHz ISKigte or mufapte cycles)
Internal 0 01 Hz to 10 kHz
Ertemal DC to t MHz <TTL compatible input level)
OPERATING TEMPERATURE: O C to - 55 C specrfccalions apply from 0*C to
POWER '00120220240V • 51. 10". swncnao* 40 Hz to 440 Hz - 25 VA
DIMENSIONS: 2t3 • 102 • 377 mm (8 375 . 4 • 14 825 in}
PRICE IN U SA.. 3312A $900
P O Box 301
81 5 Founeenffi Street S W
Lovetand Colorado 8053'
waveform. When the gating amplifier is again en
abled at the end of the gating signal, diode Dl is ini
tially back-biased, so waveform generation contin
ues until the waveform reaches a level set by the
START/STOP PHASE control, at which time diode Dl be
comes forward biased, closing the negative feedback
loop and stopping the main generator.
When in the trigger mode, the one-shot multi
vibrator, fired by the positive transition of the gating
signal, generates a short pulse that disables the gat
ing amplifier only long enough to allow waveform
generation to get under way. The modulation genera
tor then produces one full cycle and stops.
Product design and mechanical layout were done
by Bob Moomaw. Industrial design was by Jon Pennington. Thanks go to group leader Noel Pace for his
guidance throughout the project, and to section man
ager Bob Dudley and initial product manager Jerry
Dan D. Danielson
From the city of Grand Junction on
Colorado's western slope. Dan
Danielson crossed the continental
divide to earn a BSEE degree at
the University of Colorado( 1 972).
Fromthere. he went straight to
work for Hewlett-Packard in func
tion generator development.
Dan's spare time activities include
the standard Colorado outdoors
(skiing, fishing, hiking)
Estes for help in defining the instrument. Also, spe
cial thanks are due Doc Hadley, who moved on to 1C
production before the project's completion, but
whose early design efforts, particularly in the sym
metry control circuits, exerted a positive influence
on the project.,-'
1. H. Heflin, "Compact Function Generator with En
hanced Capability/Cost Ratio." Hewlett-Packard Journal,
July 1973.
2. R.C. Hanson, "Compact Function Generator Covers
0.0005 Hz to 5 MHz," Hewlett-Packard Journal, June 1969.
3. R.L. Dudley, "A Voltage-Programmable Low-Fre
quency Function Generator with Plug-in Versatility,"
Hewlett-Packard Journal, November 1965.
4. R.H. Brunner, "A New Generator of Frequencies Down
to 0.01 GPS." Hewlett-Packard Journal, June 1 <):>!.
Hewlett-Packard Company. 1501 Page Mill
Road. Palo Alto. California 94304
Ronald J. Riedel
A Missouri native. Ron Riedel
earned BSEE and MSEE degrees
at the University of Missouri in
Columbia. He then joined HewlettPackard (1972). going to work on
function generators. Like other
residents of Loveland, Colorado,
he skis, fishes, and goes hiking
but he's also learning classic
guitar and spends time as a coun
selor in church youth work.
Bulk Rate
U.S. Postage
MARCH 1975 Volume 26 • Number 7
Technical Information from the Laboratories of
Hewlett-Packard Company
Hewlett-Packard S A.. CH-1217 Meyrin 2
Geneva. Switzerland
Yokogawa-Hewlett-Packard Ltd . Shibuya-Ku
Tokyo 151 Japan
Editorial Director . Howard L. Roberts
Managing Editor • Richard P Dolan
Art Director. Photographer • Arvid A. Danielson
Illustrator • Sue M. Perez
Administrative Services. Typography • Anne S LoPresti
European Production Manager . Michel Foglia
/"^ I delete A our I please off) /^\ [3 A I I f") [Z O O . To change your address or delete your name from our mailing list please send us your old address lable (it peels off)
O (l r\ I Page Palo 94304 \J I AA \—J LJ ll L O O . Send changes to Hewlett-Packard Journal 1501 Page Mill Road Palo Alto California 94304 USA Allow 60 days
© Copr. 1949-1998 Hewlett-Packard Co.
Was this manual useful for you? yes no
Thank you for your participation!

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

Download PDF