And the noise looks digital in nature... I mean, it seems to jump between two discrete levels, in addition to the wobble/drift (it drifts within a couple of seconds, not sure what to call it ).
Because of the complex CCD based sampling system this oscilloscope uses, problems can manifest as noise spikes as shown without it being a digital logic problem. While from a DSO using a design closer to a modern one, the photographs below are from a completely *analog* problem.
I wouldn't actually know; this is a dumpster rescue I was diagnosing the power lines when suddenly that capacitor blew up, and some of the lines went back to normal. I then suspected there were other faulty capacitors, which I confirmed and, upon removing them, the voltages went back to their nominal values. It would usually fail some combination of CCD, PA and TRIGS, until I replaced the caps. Now it passes self-test, but the trace does that weird stuff.
Which other capacitors did you replace?
The aluminum electrolytic output and maybe input capacitors of the switching power supply are old enough to be suspects when odd problems start to crop up. They are a lot more work to change though compared to cleaning the hybrids.
I will try cleaning up the hybrids, then. Are they in any kind of weird fragile package? Can't really see under those heatsinks
I have not had them off myself but from descriptions, I think they are just pressed onto the printed circuit board and held on by the visible fasteners at their cornerss.
Are the NVSRAMs particularly prone to failure? Sometimes it comes up with weird waveforms instead of those previously stored, I found that somewhat suspicious. Other settings seem to be stored normally otherwise.
The NVSRAMs have a rated operating life of 10 years. Mine lasted almost 20 years. Not all of these oscilloscopes used NVSRAMs however. Older ones used a lithium backup battery and low power SRAMs.
The two NVSRAMs (or SRAMs) are located on the bottom of the processor board which is the top board which tilts from one side for easy access. Inspection will reveal if a battery and SRAM or NVSRAMs were used and the date codes on the NVSRAMs if present will be visible.
If the calibration constants stored in memory are lost, it is not a huge deal. External calibration can be done with a modicum of lab equipment; it just requires some accurate voltage levels and a fast rise or fall pulse; the sync output on a function generator is good enough. For non-critical applications the external calibration can be ignored.
Also, do not operate this oscilloscope with the cover removed *without* air blowing across the hybrids on the bottom circuit board. Otherwise they will overheat!