Repairing the Amrel PPS-2322

There is a 16V8B series PLD (Programmable Logic Device) on the computer board. The PLD's data-sheet indicates that the device is rated for 20 years of data (configuration) retention. As these instruments are now going on 30+ years, this poses a vulnerability. Another blogger and I worked together to determine the necessary logic equations to replicate its function. His power supply was a Motech PPS-1002 and while there are differences his map of the computer board is largely identical to the PPS-2322. As such, these equations ought to work for everything in the PPS series.

The design files used by WinCUPL and the Burn-Ready .jed file are available as files linked to this project. While I tested this using old Lattice Semi GAL16V8B chips, it should also burn to the ATMEL version(s) of the 16V8B available as new stock from the typical electronic parts catalog companies.

Having found that the LCD display was leaking noise back into the analog section I installed some low frequency Manganese Zinc Ferrite chokes on the supply lines to the display and the lines from the transformer to the computer. No joy, the noise remains.

Seperatly,  I found that room lighting (LEDs) were presenting ground line noise. I installed chokes on the AC and ground lines, that effort was effective.

I tack-soldered film caps a few of the MLCCs (also replaced two with film capacitors as I was trying some desoldering for off-board tests) and then removed the new low ESR electrolytic from the -15 V regulator output, tack-soldering a cheap 0.1 uf electrolytic cap in its place. Interestingly, the intermittent ripple didn't return, even after removing the film caps tacked to the MLCCs (the two film caps I fully installed remained). This gives some weight to the possibility that there were multiple failures (voltage/power stage output cap, supply voltage regulator cap, & op-amp decoupling caps) that caused the problem.

While researching capacitor types I learned a few interesting tidbits.

- MLCCS are generally available in two classes at present with Class 1 being the highest stability, highest cost, and least dense of MLCCs. While typically used for oscillators and not cost-effective as supply caps, there can be additional reasons to use these (voltage coefficient of capacitance loss, piezo-accoustics, long term aging).

-Assuming the original (mid 1990s) MLCCs are Class 2,(possibly even Class 3), testing the repair options by tack-soldering film caps to the terminals of the old MLCCs subjected them to intense heating that could have "de-aged" those MLCCs.

-Class 2 MLCCs can exhibit bias voltage dependent capacitance loss.

While Class 1 is roughly 4x the price, I'm proceeding with them instead of film caps or the cheaper Class 2 caps on CH1 to give this effort to minimize the remaining ripple its best chance of success.

Continuing to hunt for the cause of the intermittent 0.8 MHz sinusoidal ripple remaining on the negative 15 volt rail of CH1, I replaced the electrolytic capacitor at the output of the 1 uf negative 15 volt regulator on the rectifier/supply board with a low ESR 10 uf that I had on-hand

Finally, that fixed the problem!

There is some random ripple remaining and so I'm going to look at replacing the ceramic caps on the board. Not sure if i'll stick with MLCC (ceramic) caps or 'upgrade' to polyester film. The ceramic caps on the computer board are standing off a bit which allows for some strain relief when the board flexes or expands. The MLCC caps on the main board are fully inserted and may be prone to cracking as a result of thermal cycling stress. My attempts to (gently) de-solder four have left none intact and I'm increasingly suspicious that one or more failed which demanded more suppression from the electrolytic caps.

I replaced the D-A converter, but the  0.8 MHz sinusoidal output ripple remained.  More surveying, with the D-A converter and a few other ICs pulled, demonstrated that ripple was present across the groundline itself, observable by probing the ground line at the rectifier board and the ground point in the vicinity of the power transistors. Lacking a good reference point on the main board threw me off when I was investigating the D-A converter and saw ripple at every Analog point (should have been obvious the problem was more fundamental).

So, let's consider some vintage equipment repair fundamentals 1) look for burned resistors and transistors (done) 2) check your electrolytic caps (not done, oops). Checking electrolytic caps includes checking ESR (or just replacing key electrolytics with modern low ESR parts if they are cheap enough).

There's only one electrolytic per channel on the main board, at the output of the power/voltage gain stage. I pulled the output capacitor and noted that CH1 was 100 uf while CH2 is 200 uf. (There are mechanical reasons for the difference, but there are also workarounds, so there's an opportunity for improvement).  replacing with a decent low ESR cap solved the problem of ripple on the ground line entirely. But there's now intermittent ripple at the output and at the negative 15 V rail.

With the ripple reduced in magnitude and intermittent I've re-probed the main board but can't find any signal matching timing or phase with the intermittent blips of ripple.

While waiting for mt D-A converter to arrive I copied the UV-EPROM (TMS27C512) binary using a TL866II+ tool, results were unremarkable and a coursery inspection matches blogger Kerry Wong's binary. I have another vintage UVPROM in the mail. Nice to have a backup in case of bitrot. I'm now wondering if I could edit the binary to implement *IDN in place of the present non-standard command as the GPIB commands are visible in the ASCII interpretation of the binary.