HP 8720C Phase Lock Lost and A58 M/A/D/S

[mankan]

A while ago I got an HP 8720C VNA 50MHz-20GHz which was a nice extension of my labs VNA capabilities (HP 8753D and NanoVNA SAA2 Plus).
It come with the 001 (1 Hz frequency resolution), 006 (solid state transfer switch) and H89 (mixer test set) options.
It also came with some issues: the port operator checks had some intermittent attenuator failures, in particularly on low frequencies. And some occasional phase lock lost messages, which I found that I could avoid if I set the lowest frequency above 2.6GHz, but sometimes I also hade to limit the highest frequency to around 19GHz.

A lesson learned from my repair adventures of my HP 8753D was to first make a EEPROM backup so I started out by building a GPIB disk emulator. After that the repair paused almost a year and used it for some measurements and avoiding the phase lock lost messages and so on.

Some weeks ago I resumed the repair, starting with enabled the SW option 010, Time Domain analysis. This was the first time I successfully enabled an option using a code as you are supposed to instead of having the manipulate the EEPROM. Next up was another small repair, replacing one of the missing bias fuses and its holder. I managed to find the make and model of the original parts but ended up with a cheaper option. New to the left.


Now it was time for the phase lock lost issue. I started out with running the corrections procedures but that did not help so it was time to dig into the troubleshooting section of the service manual. First up was the "low band phase lock lost" parts even though the Analog Bus Test 9 showed some intermittent high band failures as well.


After various measurements I found that the low band (< 2.55GHz) output was a tad low around 1-1.1GHz so I started to suspect the A53 mixer amp since both YIG oscillators provided enough power. But the next day I realized that power measurements was made with a rather lossy cable so I ruled the A53 mixer/amp module working as intended.

To be continued...

[mankan]

The next step was to follow the broadband phase lock lost section and the high band part. Started out checking the PSU leds and also measured the voltages delivered to the lower section of the unit. All looked good. So I checked the output power in CW mode on the ports. Due to the solid state switch option the output power is limited to +5dBm (some versions of the specifications says +3dBm), however I got around -25dBm instead. Ran some more tests like the open loop power which I concluded was not working on my machine due the mixer test set option so I connected directly to the A58 M/A/D/S R path output and ran the source DAC tuning. -30dBm on high band and no signal on the low band. I'm lacking at least 17dB. So now the A58 M/A/D/S module is the main suspect which led me to this thread with some useful info.

However since the A58 M/A/D/S module seemed to be made of unobtainium (or at least costing 6-8 times more than I paid for the entire machine) I continued to do some more diagnostics like checking A12 reference board and the A14 Frac N, which actually indicated a failure. But anyway, I continued to check sampler voltages and source control and test set voltages, all seems within specs. So all evidence indicated a faulty A58 module.

Got some help from a friend that is a professional HW engineer (SW is my game) and also a RF geek. So with info from the forum thread mentioned above we got the idea of bypassing the A58 M/A/D/S module with a simple resistive splitter to see if we would get a phase lock again.


During this session the Rifa X capacitor in the PSU decided it was time to let the magic smoke and its stench out, luckily the power cable was easily accessible and we managed to minimize its damage.


This incident sidetracked the repair for a while since I had to go through my other instruments to rid them of Rifa capacitors and net filter modules with Rifas inside. Seven of twenty instruments had Rifa capacitors, high score to my Agilent E4407B spectrum analyzer with five Rifas and a notoriously known netfilter model from Schaffner.

[mankan]

Now I got the following idea: maybe the A59 source control board that controls A58 M/A/D/S" is faulty. So I started to draw its schematics, too bad that it seems to be no HP CLIP available for the 8719/20/22 series. After a while and some measurements I concluded that the A59 source control board is working properly.

So it was time to try to take a look inside the A58 M/A/D/S unit, however my first two attempts ended up with broken tools so I diverted to desoldering and read the EEPROM with a programmer as well put in another EEPROM. Also started to read out the FW EPROMs and uploaded them to the KO4BB site. Tried replacing the EPROMs with flash memories as I did on the HP 8753D but no success yet, I have some ideas on howto to proceed but it was to get into A58 M/A/D/S again. So after talking to another friend about this and looking at some eBay pictures I concluded that the lid was glued and some more force was necessary.

With help of a hammer and the broken screwdriver from the first break-in attempt I managed to get the lid off.


In short M/A/D/S consists of an attenuator in series with four amplifier steps and then RF taps for the R sampler and the ALC.


After several measurements it was found that amplifier 3 had 8 Ohms resistance instead of 4 Ohms as the other amplifier steps had in their positive bias paths and when powered on it measured as an open. One can also see that the trace have been hot:


So the plan was to solder a wire to feed positive bias to the other end of the trace and thus not using the poorly connected path, however a tiny moment of "I should just touch up this soldering a bit" a RF path connection was lost. Like the small net seem in the upper right of this image:


Bummer!

So the way forward was now to bypass amplifier 3 with a small wire instead.

To be continued again...

[mankan]

Now it was time to put the A58 M/A/D/S back into the machine and give it a spin.

Yes, no phase lock lost messages, Analog Bus test 9 looks perfect, A14 Frac N is also successful. What about the output power then?

After doing measurements every GHz and on multiple power levels I concluded the RF output power is about 10-12dB too low.

So next up is power adjustments procedures.

Low band power was adjusted from -0.84dBm to 0, as it should be.
Blanking adjustment increased from -28 to -17, also according to service manual.
Power linearity, -15 level was -17 but adjusted, -5 was -13.6 and also adjusted, 5dBm was -8.4 but the M/A/D/S got into compression at +2dBm.

After the adjustments the results are:


Which actually are within one version of the level spec (up to 15-16GHz at least) that says max power +3dBm for option 006, with flatness is ±4dB.

Knowing these limitations will be useful when doing measurements in the future. I've decided I'm done for now and now the operator checks works passes on both ports as well  . Just did some final measurements on the A58 M/A/D/S control signals now that it is "working" again, until some of the other amplifiers gets too hot as well.

[mankan]

After some thinking I decided to give A58 M/A/D/S some more input power in order to see if it could remedy the power drop on high frequencies.
To achieve this I replaced the A68 attenuator between the S2 and S3 relays. The theory of operation section of the service manual says it is a 20dB attenuator but the spare parts section says it is 6dB. So I popped the hood once again and found the attenuator to be 6dB, replaced it with a 3dB one.

I did not do any power adjustments since they are using too low frequencies, 50MHz and 2GHz, for the A68 attenuator to come in play.

Anyway without further ado, here are the new measurements:

I'm happy