Microwave Test Equipment

Brooke Clarke, N6GCE 2001 - 2010

8410 Network Analyzer
Scalar Network Analyzer
Slotted Lines
Noise Figure
Spectrum Analyzers
Power Meters

HP 8410 Network Analyzer

Seeing the photo (External link) of the Boonton 33A "Admittance Bridge" front panel reminded me of the HP 8410 Vector Network Analyzer.
Prior to the 8410 transistors were specified using h or y parameters and these required testing the transistor using either an open or short circuit AC load.
As the cutoff frequency of the transistors got higher and higher in the early 1960s these loads resulted in oscillations that were hard to stop, hence the 8410 that uses 50 Ohm terminations on all the test ports. 

There were a number of accessories for the 8410 specifically aimed at transistor characterization.
There was an automated HP 8410 system in Palo Alto (or maybe Mountain View) where I took bunches of microwave transistors to get the characterized for S-Parameters.  These were analyzed using software I wrote that would combine the measured S-parameters with common circuit elements including transmission lines.  This was run on a time sharing computer in Palo Alto using an ASR-33 teletype in my office connected by an acoustic coupler type modem.  The billing was based on two factors, 1) how much CPU time you used and 2) phone connect time.  If the computer went down during your connection all the bill would be zeroed and start over.   There was a very large difference in the billing rates for CPU and connect time and the computer often crashed.  So it was my policy to leave the terminal connected after the program finished running in the hopes it would crash.  This paid off most of the time.  This was a big step up from driving to a different computer with a stack of punched cards to have the Fortran program run as a batch job, correcting any mistake in the deck of cards and resubmitting, etc.

The HP Journal for Feb. 1967 introduced the 8410A. Rereading that article in 2010 brought back the memory of attending the local HP introductory meeting to introduce the 8410A.  One of the reasons for it's development was to come up with a better way of characterizing transistors.  The then common "h" or ''z" parameters required either shorting or opening one of the transistor leads, which resulted in microwave oscillations on the new hot Germanium devices.  One way to avoid that problem is to terminate the transistor in 50 Ohms, hence the development of Scattering Parameters (S-Parameters).  I did a lot of searching of government documents looking for "Scattering Parameters" and got back many pages of 132 column computer paper that mostly had to do with radar cross section stuff.  On page 3 (Fig 2) in the above article shows a scope connected to the 8413 Phase-Gain unit and a BWO sweeper and there's a small box off to the right that I don't remember.  It turns out the small box is a 723A power supply (0-50V, 100, 200, 400 or 550 ma current limit) powering the filter under test.

I wrote a lot of computer code in FORTRAN and Basic to analyze S-Parameters.  At first this was done using punched cards and driving them to the computer in Palo Alto (that's how I got a dent in my 427 Cobra).  Later used an ASR33 Teletype in my office connected to Remote Computing Corp in Palo Alto to do the analysis.  They had a policy that if the computer went down while you were working (i.e. while you were connected) they didn't charge you for the prior computing time.  Since they went down often it was my policy to not hang up after doing a calculation, but rather leave the modem connected for the rest of the day.  The gamble paid off enough times that it was a good strategy.  When the RCC rep was visiting I told him about it as a motivation for them to have fewer failures.

My key reference document for S-Parameters was App Note 95 (an95.pdf, an95-1.pdf)

The design of microwave transistor amplifiers in terms of getting flat gain over the pass-band was straight forward.  The hard part was taming the out of band gain to prevent oscillations.  The narrow (L and S) band telemetry amplifiers designed by Bob Mouw were made of modules.  Each module had an quarter wave resonator as the input coupling element and another quarter wave resonator as the output coupling element.  When the module plates were installed into a rectangular housing the input and output coupling elements would line up.   That's to say there was no RF  electrical connection between the stages, just the coupling between the resonators.  The beauty of that is there was only in band coupling between stages.  This not only eliminates out of band gain/oscillation, but also eliminates distortion and makes the amplifier act like a narrow band filter. 

Prior to the HP 8410 microwave device Smith Charts were measured using a slotted line one frequency at a time.  This was a tedious process.  It took me about 1 day to get one Smith chart over a reasonable band of frequencies.  The HP 415E SWR meter is just an AC voltmeter that has a very narrow filter centered at 1.000 kHz.  Many of the signal generators had a built in 1 kHz modulation to support the use of the "VSWR meter". Also the testing methods for transistors that worked OK at low frequencies like "h" or "z" parameters did not work at higher frequencies because the measurement methods required either an open or short termination on the transistor causing it to break into oscillation.

Noise Figure

The problem came when measuring Noise Figure.  The HP 340 Noise Figure meter has a 30 Mhz input.  We often used a Narda type-N coupler and one of the Tunnel Diode Detectors (made by us) as a mixer to convert the L or S band signal down to the 30 MHz input of the NF meter.  Between the mixer and NF meter there was a low noise AIL tube type IF strip. 
H.P. 349A Noise
The HP 349A Noise source was a rectangular box and we added a 10.0 (calibrated) attenuator to minimize the effect of the VSWR of the noise source changing between the on and off conditions.  This is a double side band measurement, i.e. the NF meter sees energy at both LO+30 MHz and LO-30 MHz.  When testing what amounts to a combined amplifier and filter with very steep skirts (many stages of amplification is the same as many poles in the filter) that's say only 100 MHz wide when the LO frequency gets within 30 MHz of the pass-band edge the NF starts to plummet.  QA inspectors didn't come to us with an understanding of how this works.

When testing the Log Video Amplifier used with our Limiter-Detectors made for Applied Technology (then in the Stanford Industrial Park just off Page Mill Rd) in Radar Warning Receivers I modified the HP 343A Noise Source (as far as I can remember) by bypassing a series capacitor so that the lower frequency limit was DC.
H.P. 343A Noise
The Log Video amplifier tested looks the same as the one in the AM-6536 / ALR-54.


The three columns are for the suffix letter.  The date is for the newest version of the manual, not the equipment date.
HP 8409A
                  System Components

110 MHz - 18 Ghz
Semi-Automatic NA
March 1978

HP 8410A
                  Fig 1-1
Manual NA
July 1979
September 1969
-100 System 0.11-2 Ghz
-200 System 2-12.4 Ghz
-300 System 0.11-12.4 GHz
HP 8412
                  Fig 1-1
Phase Magnitude Disp
April 1972
December 1981

HP 8413A
                  Fig 1-1

Phase-Gain Ind
May 1974

HP 8414A
                  Fig 1-1

Polar Disp
May 1973

HP 8418
                  Fig 1-1
Aux Disp Holder
May 1980
July 1982

HP 8740A
                  Cover image

Trans Test Unit
DC to 12.4 GHz
June 1972

HP 8741
                  Fig 1

Reflection Test Unit
0.1 to 2 GHz
July 1972

Reflection Test Unit
2 to 12.4 Ghz

HP 8743
                  Fig 1
Reflection-Transmission Test Unit
2 to 12.4 Ghz
July 1972

HP 8745A
                  Fig 1
S-Parameter Test Set
0.1 - 2 GHz
February 1982


1.8 - 18 GHz
Sliding Load
June 1984

1 - 18 GHz
Sliding Load
January 1979


11607A --

The 8410 would display the Smith chart in real time allowing tuning.  The 8411 Harmonic Converter has a reference and test input port and inside a variable frequency oscillator generating harmonics way up into the microwave region that acted as the local oscillator to the reference and test mixers.  The down converted Intermediate Frequency was fed back into the 8410 main frame.  The socket for the 8411 umbilical cord could be on the front or optionally n the rear panel of the 8410.  The 8407 is a lower frequency analyzer that covers 100 kHz to 110 Mhz.  There is no separate frequency converter, there are two direct inputs to the 8407.

There was a problem in automated systems because the 8411 might lock onto the harmonic just above the test frequency or just below.  Either one would provide the correct IF frequency.  One way to solve this was to measure the VCO tune voltage going from the 8410 to the 8411.  Another way was to use the 8410C and inject the correct LO frequency from an external synthesizer like the 3335.
4647847  Method and apparatus for eliminating harmonic skip March 3, 1987, 324/76.41; 324/76.43; 324/76.48; 324/76.62; 324/76.82

The 8411 fit on a shelf in the upper left the rear panel. On the upper right of the rear panel there are two APC connectors for the reference line.  If the electrical length of the DUT + the return coax was too long you could add a longer reference line here to balance the phase.

the 8412 Phase Magnitude Display CRT was one of a number of displays that plugged into the 8410, .  This was the display of choice for looking at gain or loss vs. frequency in real time.

8413 Phase Gain Indicator meter might be useful if you wanted to get more resolution in a manual measurement because you could expand the meter scales.


The 8414 Polar Display CRT is the one that I felt was the most useful.  There were Smith Chart overlays made with a number of different magnifications.  One was the normal Smith chart, one was magnified and the other was a wide angle.  You could also correct for the reflection coupler directivity by using a sliding load. 

Sliding Load

As one person pumped the sliding load back and forth you could adjust the IF attenuation on the 8410 and the X and Y position controls on the 8414 so that the dot was on a big circle centered on the CRT.  If you pressed the zero button after this you would see that the dot was not in the center.  This was a manual way to correct for the directivity of the reflection coupler.  In automated NA systems you just move the sliding load and then the system tests at all the scheduled frequencies, then move it again, etc.

The 8418 Auxiliary Power Supply could be used if you wanted to have two displays like both the 8412 and 8414.  There was an option for the 8414 that grounded both channels just like the zero push button on the front panel.  This allowed the two DVMs to read the "zero" position of the spot.

Early automated systems had a problem in that the A/D converters in the 8412 (I think this is the one but not sure) were not so good.  The fix was to use the 8414 and feed the X and Y outputs to a good digital volt meter.  Older systems used the 59313A 4-channel A/D Converter.

There were a number of test sets all of which had reference and test outputs that matched the reference and test input mechanical locations on the 8411.

The 8740 is a DC to 12.4 GHz transmission test set that would be used with the 11605 Flexible Arm that was made up of rotary joints and hard lines.  If you used your own coax to complete the transmission path you might also need to add a longer reference coax line on the back of the transmission test set in order to be able to balance the phase plots (only if you were concerned with phase linearity).

The 8741 is a 0.1 to 2 GHz Reflection test set and the 8742 is a 2 to 12.4 Ghz Reflection test set.

8742  is a 2 to 12.4 Ghz Reflection Test Unit very similar to the 8741 except for frequency coverage.

The 8743 is a 2 to 12.4 Ghz Transmission and Reflection test set (Option -018 goes to 18 GHz) but has no provision for reverse S-parameters so you need to physically reverse the device under test to measure them.  In the upper left is the crank to control the electrical length.  Remember that the 8410 is a analog instrument, there is no microcontroller in it.  Microcontroller based network analyzers replaced the mechanical line streacher with math on the phase data. 


The 8745 S-Parameter Test Set is a 0.1 to 2 GHz S-Parameter test set that can measure all 4 S-Prameters with one insertion of the test device.  This allowed complete testing transistors with a single insertion and without reversing the test fixture.  The transmission return hard line and rotary joint return arm for the 8745 was the 11604.  There also was a rack width DC power supply for biasing transistors that may have been the 8714.

11607A Small Signal Adapter11607A Small Signal Adapter - an external coupler to allow the reference signal to be large and the test signal to be small.  AFAICR Good for testing devices that need a small signal so as not to be driven into non linear operation.  Note: most semiconductors should be driver at -20 dBm or lower.

8746 S-Parameter Test Set covers 0.5-12.4 GHz and has a built in 0 to 70 dB step attenuator (10 dB steps).


The 11605 is the Flexible Line used to complete the transmission return path, or you can use your own coax cable if phase is not important.  This line is awckward to use because it has limited degrees of freedom.  This line does not have the phase changes associated with many flexible coax lines.

8409 Network Analyzer

This was a computer controlled system based on the 8410.  We rolled our own versions of this system with improvements in the software and calibration methods.

I have read on the internet that the 8409S contained:
8620C based sweeper (although later ones had an 8350 sweeper)
3335A generator - to supply the local oscillator to the 8411 harmonic converter or to drive the Synchronizer???
8709A Synchronizer - to force the microwave generator (operating in single frequency mode) to be on the correct frequency
9845C Calculator with HP-IB interface
2 each 'S' parameter test sets
a switching box that used a single 8411 sampler
the 8410C with the auxiliary display holder
2 or more 6 foot tall rack cabinets

A computer controlled 8410 system was described in the Feb 1970 issue of the HP Journal.  There was a small business setup in Palo Alto that had this system and we would bring microwave transistors to them and get S-Parameter data for each serial number transistor.

AP App Note 221a (an221a.pdf) describes a two rack plus 9845T desktop calculator (HP called them calculators because there were purchasing restrictions on "computers") on it's own separate stand, so it took up the space of a three rack system.  It also used two 8620C sweepers one for each to the two test sets.  This is better than the multiple BWO (Backward Wave Oscillator) system, but eventually after using the 8350B the ultimate answer was to use a synthesized source.  This app note talks about using an HP 59313A Analog to Digital Converter.  11863D was the software package for this app note.

This system used the Source Phase Lock subsystem that consisted of an HP 3335A Synthisizer/Level Generator which even today is the instrument used to calibrate other instruments even though it's long obsolete.  It covers 200 Hz to 80 MHz in steps of 0.001 Hz and an amplitude range of -86.98dBm to +13.01dBm in 0.01dB steps.  The output of this synth was fed as the LO for the HP 8411 Harmonic Mixer.  The output went through an HP 10515A doubler and an HP 8447C RF amp then a 140 MHz band pass filter.  The HP 8709A Option H17 Synchronizer took the IF out of the 8411 and drove the FM input of the sweeper.  The 8411A Opt 018 allowed switching between normal stand alone operation (without all this extra equipment) or the Source Lock subsystem.
The problem with this is that there's a huge amount of phase noise on the output of both the 8620C and the 8350B sweepers.

This system was found to have some problems.  One was that the same frequency needs to be tested many times.  For example during calibration of using a short, open and load then during the testing of the DUT for S21, S11, S12 & S22.  The results of all these tests are combined mathematically to get the final results.  If the RF source does not exactly repeat each test frequency then the result is degraded.  So instead of a sweeper, first an EIP frequency counter was used to phase lock the RF sweeper and later frequency synthesizers were used.  Another problem was the quality of the analog to digital converters that were reading the 8414 Polar Display.  These were replaced with a system DVM that was both fast and accurate.

In the Typical Results section of an221a they compare the 8409A (without Source Phase Lock), 8409B (with Source Phase Lock) and 8542B (metrology system).
When measuring an air line the 8409A shows gain up to 0.3 dB whereas the 8409B shows 0.04 dB loss.  Air lines can be embarrassing to measure.

Trivia:  You can make a more accurate measurement of a low VSWR using a scalar analyzer than with a network analyzer.  This required a special Wiltron bridge and an air line plus a termination that's slightly off from 50.0 Ohms.  When a swept VSWR measurement is made over a wide enough frequency range compared to the length of the precision air line, the trace has small oscillations caused by the slightly off termination circling the air link 50.00000 Ohm impedance.  By finding the mean trace you can get a more accurate measurement than that provided by a standard network analyzer.  I wonder if this could be incorporated into the software of a NA?

Trivia:  This reminds me that one way to correct for the S11 of a reflection test port is to use the 8414 Polar display and a sliding load on the test port.  One person pumps the sliding load back and forth while another person uses the X and Y position controls to center the circle on the display.  Now when you press the center button you are seeing the S11 of the test port.  Note:  This is a manual test method.  In an automated system there's a relay that presses and holds the center button and the ADCs for X and Y are read and stored so later when measurements are made the "centered" values are used to find the vector of the measurement.

Trivia:  In an221 there's a footnote "APC-7 is a registered trademark of Bunker-Ramo Corporation".

Trivia:  There is no HP-IB port on the 8410.

We used various models of the HP RMB HP-IB computers to drive these systems.  The programs were written in RMB.

Scalar Network Analyzers

Scalar analyzers work with magnitude only (no phase information).  This means for normal use they are not as accurate as the Network Analyzers.  But for metrology grade testing they can be more accurate.  This involves the use of precision transmission lines and the use of loads that are slightly different from 50 Ohms.  Wiltron made the bridges, air lines and loads for this type of testing, but it wasn't used for normal component testing.

The 8755, 8756 and 8757 are scalar analyzers that we typically used with the 8350B sweeper.  The 56 & 57 have two HP-IB ports on the back, one to connect to the sweeper and the other to connect to a computer.  Much of the automatic test software that I wrote in HP Rocky Mountain Basic would allow the sweeper to be connected to either the computer directly or to the 56 or 57.  When the sweeper is connected to the 56 or 57 then all commands to it must be passed through the 56 or 57.  When the sweeper is connected directly to the 56 or 57 manual operation of the system is much more user friendly.

These analyzers were a big improvement on using couplers and detectors with an HP 120 (later Tek 5104?) scope and grease pencils to mark the scope face.

There are two modes of operation when combining a sweeper with one of the HP Scalar Network Analyzers.
System Interface
In one mode the sweeper is connected to the System Interface HP-IB connector and there is some functionality linking the two instruments without a computer.
Note the HP SNAs require a 27.8 kHz square wave modulation on the microwave signal, which is a built in feature of the 8350B or can be done using an external modulator.
Also required is a 0 - 10 Volt sweep ramp.

Standard HP-IB Interface
There is also a standard HP-IB connector on the SNA that allows computer control.  In the systems I designed this was the connector that was used with a computer and so any sweeper equipped with an HP-IB connector could be used.

One feature of the RMB code used to control this system was a check of the HP-IB connection status.  For example, if you loaded the software into an RMB computer and ran it without any equipment connected it would prompt you to configure the HP-IB address of each equipment and connect it.  If later one of the HP-IB cables was disconnected or failed that same prompt would show up allowing the tech to figure out what was wrong and correct it.

Also used one or more of the 8350B service HP-IB commands as part of the above routine to know what options were in the 8350B.  This may be related to the power calibration.
Here are some historical microwave test methods:

1) BWO sweepers with poor frequency calibration used with absorption wave-meters which produced a notch at the set frequency provided the test signal.
A tunnel or crystal detector driving an oscilloscope with a high gain DC coupled input like the HP 120
If a VSWR measurement a Narda directional coupler and another detector were also used.
A grease pencil would be used to draw a calibration line on the scope.
A variable attenuator could be used to move the trace by the VWSR (translated to return loss using the HP microwave cardboard slide rule).

2) The storage normalizer was to replace the grease pencil but I don't remember seeing more than one of them.

3) The scalar network analyzers (SNA) became the workhorses and we had dozens of setups using the 8350B and 8756 or 8757 SNAs.  A good number of these with a computer.
    These and the 8350B also support offset sweeping which is needed for mixer testing.

4) The 8350B supports the use of an external detector to provide power leveling.  For many things this is OK, but for more precise power control a computer can be added to either do point by point power control (based on a calibration using a power meter) or to measure the actual power at some number of points as controlled by the leveling loop.

Slotted Lines

Prior to the HP 8410 Network Analyzer slotted lines were used to plot a Smith Chart of a component.  A typical setup would be:
CW Signal generator (1 kHz square wave modulation) - high quality pad - slotted line (either coax or waveguide) - Device Under Test (DUT).
The carriage on the slotted line has a detector feeding an HP 415 which is a very narrow band 1 KHz AC voltmeter.
By moving the carriage to the peak and adjusting the 415 to full scale then moving the carriage to the null and noting the 415 reading and the carriage distance you get a data point.
If the 415 amplitude drops many tens of dB then the DUT has excellent return loss (very low VSWR) and the data point is at the center of the Smith Chart.
If the 415 amplitude drops a small number of dB then the data point is not far inside the outer circle of the Smith Chart and it's phase is the difference between the short null and the new null normalized by the wavelength.
Note the distance between nulls is related to the wavelength of the test signal and if the slotted line is long enough you can measure the distance between adjacent short nulls to get a half wavelength distance.  Getting at least two nulls during the short calibration is required.  Since slotted lines are typically 50 Ohm coax air lines they get longer as the test frequencies get lower.  Then come slab lines, two flat plates with a rod halfway between them, then waveguides with a slot in the center of the broad wall.
Calibration involves measuring a load and a short (the null locations are the basis of Smith Chart rotation when the data points are measured).
It can take the better part of a day to make a Smith Chart plot of a DUT at 11 frequencies and much longer if you want to measure a few DUTs to see what an average unit looks like to do a new design.


Early test setup using the 8410 was driven by octave band HP 690 series sweep oscillators that used Backward Wave Oscillators (BWO) in plug ins.  To go with each plug in there was a plastic ruler that snapped over the frequency pointers to give you a rough idea of the frequency.  But you would need a wave meter (tunable cavity that puts a narrow suck out on the scope display) in series with the RF setup so you could really know the frequency.  The HP BWO sweeper was about 1/3 the size and weight as the Alfred unit that it replaced (a true boat anchor).  HP had a combiner box that would hold three of the plug ins and a controller plug in that went into the 690 series main frame. 

Microwave detectors were used to convert the microwave signal into a DC coupled low frequency signal that was fed to an oscilloscope.  The HP 120 scope was used for many years and later I switched to the TEK 5000 series scopes which were purchased as three parts, the mainframe, the vertical plug-ins and the time base plug-in.  By doing that each PO would be under the dollar limit so no high level approval would be needed. 

Tek Low Cost Scope

Model    Description
5103N   Main Frame
5A13    differential 2MHz
5A14    Four Trace plug-in, 1 MHz
5A15    single 2MHz 1mV-5V/div
5A18    dual 2MHz
5A19    differential 2MHz 1mV
5A20    differential 1MHz 50uV
5A21     differential/current probe amp 1MHz
5A22    differential 1MHz 10uV
5A23    single 1.5MHz
5A24    single 2MHz 50 mV-1 V/div w/prototyping area
5A26    dual differential amp 1MHz
5A24N    V. Plug-in
5A38    dual 35MHz
5A45    single 60MHz
5A48    dual 50MHz 1mV/div
5B10N    timebase 100ns-5s/div
5B12    dual timebase 100ns-5s/div
5B13    timebase to 1us/div
5B25    digitizer time base, used w/5223
5B31    digital delay timebase
5B40    timebase 50MHz 10ns-5s/div
5B42    delaying timebase 50MHz 10ns-5s/div
5B44    dual timebase to 50ns/div
HP8693B BWO Plug-In

Then the 8690 and 8690B mainframes, 8693A bwo plug-in..  This combination allowed sweeping across a frequency range covered by the three such as 2 to 4 then 4 to 8 then 8 to 12.4 giving a 2 to 12.4 Ghz sweep. 

Kruse Stork model 5000 Microwave
Later the Kruse Stork Model 5000 (later Systron Donner) on highway 101 in Mountain View came out with a small solid state sweeper that could sweep 1 to 18 GHz using a combiner box similar to the HP unit but much smaller and with leveling.

Kruse Stork Patents:
3397365 - 1967, Oscillator with separate voltage controls for narrow and wide tuning -
3377568  - 1968, Voltage Tuned Oscillator
3416100 - 1968, Voltage Tuned oscillator with resistive and capacitive tuning diodes (Varactor and PIN diodes)

Later HP came out with the 8620 then the 8350 sweepers which we used for all kinds of microwave testing.




From the HP-Agilent mailing list:
86222B 0.01-20Ghz 20 mw max.out.
86290B 2.0-18.6 Ghz 10 mw max. out.
86240A 2.0-8.4 Ghz 40 mw max out.
86240B 2.0-8.4 Ghz 20 mw max. out.
86240C 3.6-8.6 Ghz 40 mw max.out
86251A 7.5-18.6 Ghz 10 mw max. out.
86235A 1.7-4.3 Ghz 40 mw max. out.
86241A 3.2-6.5 Ghz 5 mw max. out.
86242D 5.9-9.0 Ghz 10 mw max.out.
86245A 5.9-12.4 Ghz 50 mw max. out.
86250D 8.0-12.4 Ghz 10 mw max. out.
86260B 10.0-15.5 Ghz 10 mw max.out
86260A 12.4-18.0Ghz 10 mw max.out.
86260C 17.0-22.0 Ghz 10 mw max. out.

From somewhere else:
86220A  0.01 to  1.3 GHz

86222B  0.01 to  2.4 GHz
86230B  1.8  to  4.2 GHz
86235A  1.7  to  4.3 GHz
86240D  5.9  to  9.0 GHz
86241A  3.2  to  6.5 GHz
86245A  5.8  to  6.5 GHZ
86260A 12.4  to 18.0 GHz 10 mW
86290B  2.0  to 18.6 GHz

HP 8350B Sweep Gen 8350B Mainframe

This was the workhorse at Aertech/TRW Microwave/FEI Microwave for production testing of all kinds of products.  Mostly used with Scalar Network Analyzers but sometimes with 8410 Vector Network Analyzers, like for tuning the RWR modules.

The 8350B has support for mixer testing where two 8350 boxes can sweep with a constant offset (the mixer IF frequency).  This feature was critical for mixer testing and was not supported by otherwise competitive sweepers.

There was a problem with automated systems based on the 8350 (and all the earlier signal sources) in that when you programmed the microwave source to go back to some frequency it would be off just a little.  Since an Network Analyzer error corrected measurement required measuring a number of standards and then the device under test a number of times, and then combining all the results to remove the errors any change in the frequency would introduce an error.  By using and EIP 575 (EIP App Note 5 "Using the EIP 575B/578B Source Locking Counters with the HP 8350B and Its family of Plug-ins") Source Locking Microwave counter you could program the source and counter to a frequency and the counter would output a DC feedback voltage that could be fed into the 8350 DC coupled FM modulation input to frequency lock the source.  This way you would get the same test frequency every time making the test results more accurate.  Modern network analyzers use a synthesized source so this is no longer an issue.  But the cost for the EIP counter and the 8350 sig gen was much lower than the cost of a synth.

I don't remember trying the EIP + 8350 combination for the mixer spur test system, I think for that application the synth was needed to get the phase noise down low enough.

There are HP-IB (IEE 488) commands used for service that are not listed in the user programming manual that are useful for production automated testing.

Plug-Ins - came in suffix letters A/B/C/? and with option numbers for things like a built-in 0 - 70 dB step attenuator.
83522A  0.01 to  2.4 Ghz 20 mW
83525A  0.01 to  8.4 GHz
83540B  2.0  to  8.4 GHz 40 mW
83545A  5.9  to 12.4 GHz
83550A  8.0  to 20.0 GHz
83554A 26.5  to 40.0 GHz
83570A 10.0  to 26.5 GHz
83572C 26.5  to 40.0 GHz
83590A  2.0  to 20.0 GHz
83592C  0.01 to 20.0 GHz
83594A  2.0  to 26.5 GHz
83595A  0.01 to 26.5 GHz
83596A  2.4  to 40.0 GHz
83597A  0.01 to 40.0 GHz
11869A adapter allows use of 86200 series Plug-Ins


An early attempt to build an automated mixer test system was based on using a pair of HP 8350B sweepers.  The mixer offset cable was not used since the HP RMB Computer would set the frequency of both sweepers.  An HP 8566B Spectrum Analyzer was used to look at the mixer output.  This system did not work because the phase noise of both of the sweepers spread the output signal so much that it was very difficult to measure.  The solution was to use a pair of synthesizers and to connect the 10 Mhz reference output from whichever instrument had the best internal oscillator to the other two so all three ( 2 - synthesizers, 1-spectrum analyzer) were using the same reference.  Even if the referenceis off all three will see a signal at the same indicated frequency.  This allows the use of very narrow IF bandwidth in the SA and either zero span or very narrow span instead of waiting for a full sweep, greatly reducing the time to make a single measurement.  The down side of this is that synthesizers cost maybe 10 times what sweepers cost for the same frequency and power capability.

Spectrum Analyzers

The most often used model was the HP 8566B.  Saw one at the HP Santa Clara facility that was running water fall software and the sweep was covering all of the amateur 2 meter band.  When a repeater was keyed you could see vertical lines for both the input and output frequency that would start and stop at the same time (vertical axis).  I think this was an IBasic program running inside the SA, if  you know more about it let me know.

My home SA is the HP 4395A which is a "combo" instrument from HP Japan, i.e. it can be a Network Analyzer (NA), or Spectrum Analyzer (SA) or Impedance Analyzer (needs optional firmware) (ZA).  The prior combo box was the HP 4195A which used conventional analog circuits, but the 4395A uses a digital IF system that allows it to display much faster, also allows displaying true power rather than a signal's peak, and allows true 1 Hz RBW measurements.

While working on a proposal for some satellite hardware I studied measuring non harmonically related spur detection.  The idea was that we would need to use a spectrum analyzer with a vary narrow resolution bandwidth in order to detect very weak signals.  This requires an automated setup working for many many hours is not days.  The HP 70000 series was the fastest way to make the test.


100 Hz - 2.9 GHz SA
HP 70004A color display
70904A RF section
70902A IF section
70900A LO module

Color Display full rack

2.4 GHz Freq Counter

5" display 3/8 rack width

9" full rack width
PC display


70210A Prec 10 & 100 MHz Freq Ref 70900-90286
20 Hz - 2.9 GHz RF Track Gen
70900-90286 Inst_70300-90107
10 Mhz - 18 GHz Microwave Track Gen
70900-90286 Inst_70301-90039
Prec Freq Ref

1.0 - 20.0 GHz Sig Gen

10 Mhz - 1 Ghz extension "

70428A 2.4 - 26.4 GHz Microwave Source

2.7 - 22 GHz Preselector
Switch Driver
2 - 22 GHz Pre Amp
10 Mhz - 2.9 Ghz Pre Amp

1 - 26.5 GHz Pre Amp
70900-90286 70620-90036
100 kHz - 26.5 GHz Pre Amp
Digitizer 20 MSPS

100 kHz - 2.9 GHz Pre Amp
1200 - 1600 nm Lightwave
750 - 870 nm Opt B
Microwave Transition Analyzer


IF 10 Hz - 300 kHz RBW
70900-90286 70902-90021
IF 100 kHz - 3 MHz RBW
70904A RF 100 Hz - 2.9 GHz
RF 50k - 22 GHz RF
RF 50k - 26.5 GHz RF
Ext Mixer I/O (mm wave)
70900-90286 70907-90029
RF 100 Hz - 22 GHz w/preselector
RF 100Hz - 26.5 GHz w/preselector
RF 100 Hz - 22 GHz w/preselector 70900-90286
Ultra Wide IF 10 - 100 MHz RBW


70000 Series Spectrum Analyzers & Related



100 Hz - 2.9 GHz

70001A, 70310A, 70004A, 70900B, 70902A, 70904A
50 kHz - 22 GHz

70001A, 70310A, 70004A, 70900B, 70902A, 70905A
Dual rack 100 Hz - 26.5 GHz

70001A, 70310A, 70004A, 70900B, 70902A, 70903A, 70909A

100 Hz - 22 GHz

70001A, 70310A, 70004A, 70900B, 70902A, 70903A, 70908A, 70810B
1200 - 1600 nm Lightwave

70001A, 70310A, 70004A, 70900B, 70902A, 70903A, 70908A, 70810B
71400C-850 750 - 870 nm Lightwave

1200 - 1600 nm Lightwave

70001A, 70310A, 70004A, 70900B, 70902A,70904A
71401C-850 750 - 870 nm Lightwave

Microwave Transition Analyzer

70004A with 70820A
Microwave Source

100Hz to 26.5GHz
Surv Rx

70001A, 70004A, 70900B, 70902A, 70903A, 70911A, 70910A, 70310A

Power Meters

Patents related to Power Sensors 
Agilent app note AN64-1C has a lot of good information about power measurements including some history.  Alos see Chooosing the right power meter and sensor 5968-7150E.
Bolometers - measure power based on resistance change caused by temperature change from RF heating.
Most accurate non NIST sensor type.
Barretter - A thin wire (might be a 10 mA fuse) is one type of Bolometer with a positive temperature coefficient.
Thermister - semiconductor device with negative temperature coefficient.  By applying DC power to the thermistor when no RF is present and then reducing the DC power to keep the thermister at the same resistance the reduction in DC power allows determining the amount of RF power that's heating the thermister.  There is also another thremister in the sensor to compensate for ambient temperature.  The 478A and 8478B are this type of sensor.

Since NIST will only calibrate Thermister type power sensors (neither Thermocouple nor diode sensors are NIST traceable) the 432A Power Meter is still a current model number because it allows the most precise power measurement.  The related thremister sensors are also current model numbers.

Thermocouple - More sensitive than bolometers and have DC out proportional to RF power in, i.e. square law detection.  Introduced in 1974.  But since there is no longer any DC power substution they need a 50 MHz reference power for operational calibration.  The 8481A is of this type.

Diode - Offer -70 to +20 dBm dynamic range in a single sensor, but the peak power level needs to be below -20 dBm in order to get correct results when measuring complex waveforms.  The -20 to +20 dBm range is good only for CW signals.  To measure non CW singals with average powers in the -20 to +20 range use a thermocouple type sensor.  In order to have the 50 MHz calibration about in the center of the power range a special 30 dB attenuator that's optimized for use at 50 MHz is inserted between the sensor and the 50 MHz cal port.

In 1975 the 8484A diode sensor was introduced.  This was a single Schottky diode that has a small error when the signal has even order harmonic distortion (the + signal is not the same as the - signal and the diode only detects one side).

A newer type based on GaAs material and Molecular Beam Epitaxy (MBE is avery expensive process).  These are the 8481D series sensors and they use two diodes to correct the even order harmonic problem and add a number of benefits related to balance.  Much higher output than the single Schottky diode type because of the improved materials.  The power range is still -70 to +20 dBm.

E Series Sensors - These combine the GaAs dual diode sensor with a EEPROM in the sensor to automatically load the cal data into the power meter.
Tube type
Solid State
Manual Operation
introduces auto
temp compensation
with 10 kHz ref

TM 9-6625-2469-15
 Solid State
Manual Operation
A analog meter
B Digital meter
no 50 MHz cal reqd
old cable
Hi Precision (0.2%)
Vref & Vcomp out
for highest accuracy
DC ref
Manual operation 
single channel
50 MHz cal

TM 11-6625-2969-14&P
Single Channel
50 MHz cal
new cable

Agilent Manual
Single Channel
Sensor cal memory
50 MHz cal
new cable
Dual Channel 
Sensor cal memory
50 MHz cal
new cable

Power Meter Sensors & Cables

Old style, Bolometers

HP 478 & MX-7772/U Bolometer N(m)
10 MHz to 10 GHz
200 Ohms
30 mW max 5 W uS pulse max

8478B N(m)
0.01 to 18 GHz

8478B-11 APC-7
0.01 to 18 GHz

Waveguide Bolometers:
S478A 2.6 to 3.95 GHz
G478A  3.95 to 5.85
J478A  5.3 to 8.2
H478A  7.05 to 10
X478A  8.2 to 12.4
M478A  10 to 15
P478A  12.4 to 18
K478A  18 to 26.5
R478A  26.5 to 40

New Style, Thermocouple

Model        FS Power Range  Freq
HP 8481A -20 to +20 dBm  10 Mhz to 18 GHz
HP 8482A -20 to +20 dBm  100 kHz to 4.2 GHz
HP 8483A -20 to +20 dBm  100 kHz to 2 GHz
HP 8481H  0 to + 35 dBm  10 Mhz to 18 GHz
HP 8482H  0 to + 35 dBm  100 kHz to 4.2 GHz
HP 8484A  -60 to -20


HP Diode Test Set K79 0981C Diode Test Set - internal HP box for maybe for testing microwave diodes that went into something back in the tube days, noise diodes?
Microwave Network Analysis by Mark Kahrs -
ARFTG - A HISTORY OF ARFTG, 1972-1980 - Automatic Radio Frequency Test Group
N2PK's Amateur Radio Projects Page - home brew VNA - How it Works -

Back to Brooke's Products for Sale, Home, Rack and Stack Systems, Past Projects page

page created 17 Sep 2001.