HP 3577A Network Analyzer odd confidence test failure

[dzseki]

We have this puppy at work, we only use it ocassionaly and the usage is rather limited. Today I was playing a bit more with it, and found out that using input B the self confidence test fails, see picture. The input is not completely dead but as you can see there is a (nice) dip in the freq response (both in linear and log mode) all other tests are passing on all channels. Any idea?

[capt bullshot]

Does the frequency of that dip vary with cable length? Just a wild guess: The culprit could be a failed 50 termination or the relay switching the termination on/off.

[precaud]

Was this with one short cable output to input? I've seen response similar to that from just having a Y on the output with one cable hanging unconnected (and unterminated).

I've repaired several of the input boards, and the most common failure is the first shunt R in the 20dB input attenuator. It's only 1/8 watt (61.1 Ohms). Try comparing sweeps with and without attenuator engaged.

edit: its worth mentioning, unless you're using the reflection test set, you only need two channels for most VNA measurements. I actually removed the CH B board from mine.

[dzseki]

Yes just a short cable between the terminals, nothing else. I've tried with different cable lengths, the hole remained at the same frequency.
Engaging the -20dB input attenuator made no difference. In fact I would have expected that with -20dB on the line gets -20dB down, but no, it was at the same place as before. I don't know if this is right?

[precaud]

Engaging the -20dB input attenuator made no difference. In fact I would have expected that with -20dB on the line gets -20dB down, but no, it was at the same place as before. I don't know if this is right?

Yes, it's right, the input level hasn't changed, just the attenuator.

Mark which is which, and try swapping the Ch A and B input boards and see if the problem follows the board or stays in the slot. If its the slot, it is probably the LO signal from A4.

If its the board, you'll have to put it on an extender and trace signals. Do you have the service manual?

You could check the output of the buffer amp (R115) with another channel to see if its flat. After that, its the filtered outputs of the mixers.

I've only encountered one such problem and it was the LO signal, and affected all three boards. It's still waiting for me to fix it    but that unit has pretty much become a parts donor.

But as I said earlier, if you're just doing freq response and other normal VNA work, I'd just unplug the bad B channel board (or just cap the input and don't use it) and carry on with just the R and A channels. The 3rd channel was an option with the 3577B, and no changes or jumpers are required when its present or not. The software just ignores the missing channel if you don't try to use it. The "A" behaves the same way...

[dzseki]

Thanks! For now I think I'll leave as is, we really use it to tune in one notch filter to 70MHz, for that even this B channel is usable, apparently the channel can be compensated as well. Not to mention that it is currently built into a rack...
I was just curious because this surely looks like an odd problem.

[precaud]

OK, sounds like you have it covered. That symptom could be a tricky one to track down.