| Electronics > Repair |
| Series defect on agilent 167xx boards? |
| << < (32/50) > >> |
| ahakman:
Here's some context, if it can help someone else. I labeled the couple chips I know for sure by fault injection and running the self test (again, I stress what the self test reports on a 16903A 3ch Windows XP mainframe - I have some 16702B's I could try the test in as well in HP-UX - the 16752A cards are the only cards I have though that are new enough to work in the newer mainframe, which boots faster, has a better screen, and the hard drive doesn't sound like a jet engine running). C0 P1 L = Chip 0 Port 1 Low Word |
| MarkL:
Wow, I can't see any problem with that joint on the resistor pack. A good reason to always do end-to-end continuity checks in areas with corrosion or on traces that transit corroded areas. Another common location for breaks that are hard to spot are those solder blobs which are probe points. The traces leading up to it are exposed for a very short distance after they come out from underneath the soldermask. So, after reflowing does the card work now? Thanks for the update on the nomenclature. All my information is using a 16702B, and clearly they changed the chip 8/9 designation to chip 0/1 for the windows OS models. And from your investigation, it sounds like they got much of the Uxx identifier reporting just plain wrong in the windows version. Thanks, Agilent. Each pod pair has 32 bits of data that is acquired and stored. Plus, there are two extra bits for the clock/qualifier inputs that is also acquired and stored. These are the Bonus bits, and I think are stored in U19/U47 (the DRAM chips that don't have a partner on the underside), or in whatever logical grouping that includes those two chips. Some people have installed SCSI2SD adapters to get rid of the "jet engine" hard drive in the 167xx units. But the drive noise is nothing compared to the chassis fans, IMO. |
| ahakman:
I didn't scrape the corrosion off, repair the trace and try it yet - it was about 3AM when I finally found the right chip by moving my induced fault around. Hopefully tonight I will get that fixed along with the other memory error on the other chip (after figuring out which port is the one that's failing) and look into what's causing the error withe the Bonus bank on Chip 1, then we'll see how many self tests still fail. Yes, this card is very tricky compared with the first one I fixed. The first one had very obvious corroded traces that looked green under the solder mask. Scrape the mask off until you get to good copper on both sides, and solder in a jumper - problems solved. This one is all about the ends of the traces where they meet pads being corroded, which are MUCH harder to spot visually. And the corrosion is further away from where the runners were. This card almost looks like it was above another card that was off gassing or something, and the corrosion is much more widespread. As I said earlier, ALL of those 33 ohm resistor packs looked just downright awful, and I can see now that some of them still need some attention. They didn't look the best on the other card that's working either, but better than this one. Long term, I probably need to remove, clean / rehab the pads, and re-solder ALL of them on the other card too - but that can wait until later. I do want to get this up so i can actually use it for the project I have in mind. I should probably just stick the pair of 16550A cards that don't have runners and thus don't have any corrosion into a 16702B chassis and use that. For some reason I initially thought those weren't supported in a 16702B chassis, but looking at the compatibility matrix again, looks like they are - they're one of the 165xx series cards that works in the 167xx series mainframes (in the same way that the 16752a is one of the 167xx series card that works in the 169xx series mainframe) The next issue is going to be cables - I see now that the 167xx series cards use a wider plug (not to mention that spacer built into the back of the card) than the 165xx cards do. I have some cables for the 165xx cards, but none for the 167xx series cards :( |
| ahakman:
I was dragging my feet working on this again as I was waiting for my new soldering microscope to arrive. I fixed all of the connections on all of the 33 ohm resistor packs, and now all my memory errors are gone! Now I only have 2 self test failures left (down from about 10 self tests failing before): Comparators and ZoomChipSel These are obviously in a different area of the board - back to the microscope to do some detailed inspection... Edited to add: I found the comparitor problem - I think it was a trace I repaired previously, but there wasn't enough solder on my bodge wire and it wasn't making a good connection, or it could've been one of the vias I reflowed and got a bunch of weird looking junk out of). Sweet, both of my 16752A cards are passing self test now! |
| MarkL:
Great! Glad you found the problem. The comparator control lines seem to be a common victim of corrosion. It's likely the comparator failure(s) caused the zoom failures since the system uses the comparators to set up test data patterns for the zoom chips. For reference for future readers, the comparators are 1NB4-5036 next to the pod connectors (top and bottom), and the zoom chips are 1NB4-5040 next to them (top only). |
| Navigation |
| Message Index |
| Next page |
| Previous page |