I think I finally figured it out, and found info that may explain miscellaneous problems in the 8568, 8566, and other HP equipment.
The sweep stopping problem disappeared for a while - running rock solid for several days continuously, and seemingly independent of ambient temperature, unlike before. I figured I wouldn't be able to troubleshoot it anymore, so still in limbo, waiting to go bad again if I put it back in service. Then our dry spell ended, and we had lots of rain and increase in humidity. The next day, it crapped out again, really badly - it couldn't run for more than a few seconds before stalling out. I decided to dive in deep, and figure out what's going on once and for all.
After several hours of poking around in the digital control system and trying to figure out how everything is supposed to work, I couldn't find anything wrong in the brain parts. The processor did its thing, and the sweeps looked right, but it would just stop. I was running out of suspects in the logic, so revisited the A22 Frequency Control board sweep circuit. There was one last signal involved, related to resetting the sweep, from logic on A22. This is the last thing, apparently, that the processor could affect, sweep-wise.
Q21, driven by U31, can change the comparator threshold in the sweep circuit, but I couldn't tell why or when. I found it is located up near the top of the board, adjacent the SWP test point. I figured I could quickly check its base drive by poking the scope probe on the lead, through the test point hole, without having to remove the plastic cover over A22. I carefully jammed the hook-tip in there and grabbed onto it, and got the base signal on the scope.
The 8568 immediately started working properly! Something about hanging the probe there did it - I figured the stress on Q21s leads and package, or a bad solder joint, slightly (microscopically) moved something back into contact, or more remotely, maybe the probe capacitance did something. I let it run for a while, and was watching the base signal on the scope, and it suddenly disappeared. Oh oh, but no - the probe had popped loose, and the 8568 was still going strong. I left it running overnight, and it was still good this morning, so I pulled A22 to check it out.
Q21 and the surroundings look OK solder-wise, and are fairly clean - I was imagining a simple bad joint, but figured on replacing it anyway, just in case. So, I've been looking for the type to match. All I could find in the HP cross reference is that this part 1854-0071 is an Rohm brand CP4071. I suspected it may be special for high gain, since it seems so softly driven - a 100 k R from TTL to the base, while needing to pull about 3 mA Ic if it's supposed to saturate. However, there are a bunch of others in the area, so it likely isn't all that special. The actual parts are in TO-92, and marked with a symbol "2" with a circle around it, and "CT1200 138."
Now, here's what makes it even more interesting. While searching for "CP4071," I got some hits of discussions about woes caused by this part in some other HP equipment, specifically, random and thermally-related failures, which perfectly relate to this problem of mine. Here's one from a few years ago that has some detailed info:
Look at messages #30-32, in particular.
So, it looks like this part has some history of trouble anyway, and that HP may have used it a lot, so it could be the cause of other problems. It's going on my list of suspects for mysterious problems. I think that when I find and install a proper substitute, this 8568 problem will be gone - until something else comes up.
Now I'd like to find the data on this part and find a substitute. I could probably find the exact same kind from some junked HP boards, but after what I've been through, I think I'd rather have something else in there for long-term reliability. Does anyone have more info on this part type, or similar experience?
BTW I think the sweep stopping was due to the processor getting confused about the state of the sweep. It appears able to control the sweep start/stop, and the reset, but I couldn't find any feedback signal telling the processor whether it's completed or running - I could have missed something, with all the pages and pages of schematics and signals going every which way. It seems that normally each full sweep runs once, according to the settings the processor sends, but it does not report reaching the end of sweep, so the processor must know approximately how much time to allow, then does its updating activity, then the reset, then HSWP lets it start up again. I think if Q21 gets flakey, everything gets out of sync.
Ed
