Spectrum Analyser TR4172

[Alan Bain]

By some micacle I saw the capacitor which had pinged out of the black blob on the board. It is connected between base and ground of the "bias" transistor. I was able to solder leads on and measure on a bridge 10nF in a 0805 package. Leaving this here in case of use to anyone in future.

[Matt kd4pbs]

Good find! For sure, far less than I thought it would be.
I do hope you get this going, Alan. It's a wonderful SpecAn... when it works. Super low noise, great filtering, and a high resolution, even by today's standards. What's not to love? Oh yeah... the age-related issues... I forgot

[Alan Bain]

Yes, replacing smelly electrolytics in the RF section is a fairly slow process. Almost back able to do some actual RF testing!

[Alan Bain]

Back together and starting to test to find what I forgot to connect - running into a few manual discrepancies; manual makes reference of a T.G CNTR button but when I look where that should be I find TUNED AMP. Is this just a new name for the same thing (as in count the actual signal not just calculated centre + offset)?

[Alan Bain]

Here is the repaired first IF bias board (after and before pictures!). I have to admit that it was tricky. The board material conducts heat away very effectively and the gold layer appears to be pure gold (so no nickel in the alloy) and doesn't wet with solder. I found I had to use a powerful iron to be able to work quickly. The resistive layer underneath does wet with difficulty. I was very reluctant to use anything corrosive to clean the surfaces given the comments in this thread about glue cure by products damaging trace adhesion. New parts deliberately not straight to avoid the damaged trace sections. One trace to decoupling 10n cap replaced by a wire (single strand of 7/0.2 hookup wire).

Afterwards now drawing 13mA instead of 22mA+ and voltages look sane. Time for more analyser testing.

[Alan Bain]

After testing further I've noticed that the directional coupler is terminated in 50ohm on input side but measures about 600ohms looking back from C1. While I don't have any values any design for such a coupler that I have seen has had both ports terminated equally. Fortunately I have some precision 0402 50ohm terminating resistors, but does anyone have any ideas of a flux that allows the gold to be wetted?

[Matt kd4pbs]

Can't help you there, Alan. I noticed the same thing with the insanely difficult to solder properties of the substrate, and with my iron at a heat setting that would overcome the thermal conduction from the board, it would melt the traces.
I think I do remember having the same issue that you see... the directional coupler/filter not being equally terminated.
Ultimately, after bodging the SA enough that I got it working and calibrated, after a few weeks I noticed it went deaf again. Likely a leg of the mixer corroded again. I have since pulled it from the shelf and set it in the morgue awaiting either the day it gets carted to the dump or I find the required parts to make it right.

[Alan Bain]

So having mended the IF amp, remaining deafness hard to trace without an SA which covers 2-3GHz. Thanks to a kind loan, the deafness traced to one of the stripline 2.046GHz bandpass filter sections which is an approximately 60db attenuator. Careful inspection shows one section of stripline has no gold so no great surprise and I should have noticed sooner! For the moment I have put in a piece of 50ohm semi-rigid coax to bypass this section and with this bodge I can get it to calibrate.

I vaguely recollect doing the J-admittance inverter calculations to design one of these filters for a Chebychev response and I have a quick response measurement from the other so that shouldn't be so hard although this looks like an Alumina substrate around 0.72mm thick ... don't think the cheap Chinese PCB people do Alumina, or Rogers 3010! And I also seem to remember some iterative length adjustment was needed to the lines....