So I picked up a cheap HP 8133A pulse generator, non-working as is, on eBay. The photos showed something on the display, which led me to think it might be a decent repair project.
On the bench, power it up, it mostly worked. Output was limited to 2.4GHz after which is faded quickly. I left running for a few hours, and it didn't seem to have any thermal issues. So I moved it around a bit, which made it stop working (no output), then fiddled with it a bit more and the output returned. So, clearly something mechanical, like a loose connector or corrosion.

First step was to open it it up to see what kind of messes there might be inside. Corrosion, debris, dust piles, devulcanized rubber, leaky caps, exploded tants... who knows until you take it apart.
Off comes the top panel, side panel, and the top shield, exposing the MCU (MC68HC008 @ 8MHz) and main control board. Under the second shield is the card cage. Under the logic board is the power supply. I removed 9 screws to get this far; this thing is seriously solid iron!

Battery looks super clean, measures 3.14V, so I'm not going to bother desoldering or replacing it for now. In fact, there was barely any dust on it; this unit must have spent its life (it was a rental unit) in immaculate lab conditions! Some of the MCU board chips are dated wk 39 1999, so I suspect this unit shipped around 99/00 or so.
Now I've removed the cage top cover, exposing the card cage. The board seen is the channel 1 output (opt 002/003 add a channel 2, which requires a different front panel). Underneath this is the main timing board. I didn't get to that board today; the card cage is accessed by removing the front panel, and I only got to removing the rear to check the power supply. Maybe I'll pick up a nicer front panel and replace it; at that point I'll take a look at the timing board. 14 screws held the cage cover, and we're now up to 23 screws removed.

Super squeaky clean, like the MCU board. Just a hint of dust, which I vacuumed off.
On the bench, service guide on the laptop.
Overview of the channel 1 board:

Output switches. It has complementary differential outputs that can be individually turned on/off.


Delay unit. This has traces for 8 coarse delays. Finer delays are created with a DAC biasing a diode, routing the signal through one of the traces in the delay unit.



The top board position is used for opt 002 or 003. Adding it isn't as easy as just dropping in a board though, as it requires a front panel and the proper connectors and output assembly.

MCU board removed. 6 screws. Checking the voltages is easy since they're immediately accessible on popping the shield. All good here.

Pulling the PSU requires removing the back portion of the chassis. 10 screws. Plus 4 on the left (seen from the rear) to free the rear frame. 37 screws removed so far.


Detaching the PSU from the chassis. Another 9 screws. 46 screws.
And there, finally it is. Super clean, all Nippon Chemi-Con, all looking just fine. No leakage or problems. I took notes to order replacements to re-cap at some point.

That's a monster toroidal... Clearly no expense was spared in the design of this thing!

A small machined block for the power trannies.

So, cleaned out all internal dust and reassembled it. 46 screws... and some fiddling to get the cage cover to fit right. Nothing overly difficult, but just took a bit of care. It's one of those jobs where you wish you had a third hand!
Cleaned off the front panel and checked that it still works. Now it gets to 2.9GHz before fading. I couldn't get it to drop the output either, so I seriously suspect it's the jumper in the back that's a bit crusty and could use a good cleaning. I had to remove the rear block assembly to get the rear off, and just removing and reinstalling the jumper probably scraped some of the crust off. I didn't think the cause was going to be anything related to the front switches since both complementary outputs were identical and faded similarly. The fading was measured with my Signal Hound SA. So next on the agenda after this is to give the rear block a good cleaning. The instrument specifies 3GHz minimum, 3.5GHz typical, so it's still a bit on the low side.
What's with patching up panels with stickers? Okay, I can see a calibration sticker someplace, but how about removing old ones? Ick. So much nicer without that mess on the front.


Spec is 0.2% frequency accuracy, so good enough.

Gave the outside panels a good IPA scrub too, followed by a wet rag (water).
