I got a supposed-to-be-working HP 8904A Synthesized AF generator which passed self tests and had no obvious problems, so I took the time neccesary to carefully clean it before doing further tests. After hooking it to the scope, surprise…
Sample & Hold Amplifier
Lower half of the output waveform was not as wide as the upper half. The waveform at TP1 was not as expected. Looking at the schematic, something in the circuit that drives the sampling diodes was defective. There was no visual damage, and it took a lot of work to find that Q201 was just semi-shorted and not doing his work right. Being unable to find an exact replacement for Q201, I replaced it with Q207. This one is identical but used in a simple 10mA constant current source, which makes it a good candidate for replacement with a common part. I used a BC237 in its place and just had to change R209 from 750 ohm to 910 to bring the current back to 10mA. After soldering the ex-Q207 in place of Q201, the signal at TP1 showed the right waveform and the output was fine… appart from the fact that I did not have a blue BC237 at hand 🙂
When the output was grounded everything was fine, but when output was set to float, signal levels became completely screwed. Good news, I found the problem was simply R471 open. It did not show external damage so I lost some time probing here and there. Bad news, this is a 0,1% 25.42 ohm precision resistor. I tried to locate a replacement but gave up soon as I realized it is 100% unobtanium. The alternative was to build a resistor from a combination of others. As you can see, I used 1% precision SMD resistors in a series combination of parallel 41,2 ohm with parallel 9,53 ohm, resulting in a theoretical 25,36 ohm which is quite close. Because of temperature coefficients, I decided to replace also the mirror R470 to preserve the simmetry of the signals. Minimum order of each value was 10, so I was able to pick the best samples among the batch for the closest match, keeping in mind that getting R471 as close to R470 as possible was more important than getting close to 25.42 ohm.
I measured the lithium backup battery and it was alarmingly weak. Having read horror stories about trying to put this instrument to work again after backup battery depletion if options are fitted (which is the case), I wasn’t going to leave it there. Unable to get an exact replacement, and after getting good advice from nice guys at the hp_agilent_equipment Yahoo group -thank you-, I decided to replace the to-be-soldered battery with a socket plus standard battery which should make it easier to replace it in the future. I chose a thick CR2477 and a socket that after a bit of modification was fitted in place of the original battery. The service manual includes detailed instructions to provide power to the RAM IC while the battery is replaced so it does not loose the contents.