Electronics > Repair

Series defect on agilent 167xx boards?

<< < (12/50) > >>

nikodem:

--- Quote from: MarkL on January 20, 2021, 10:22:12 pm ---In the pv output "Chip 8" and "Chip 9" refer to the acquisition ASICs.

Please post (attach) your output from pv with "d r=9" turned on so we can have some context.  At the moment, the output from anlyBusTest and clksTest will do.

--- End quote ---

I will provide it as soon as possible, for now I only have a screenshot with d=256, r=100 of the anlyBusTest:



and part of the clksTest:



for the TEST 1: Master Clock from CHIP 9 all the clocks it's as shown:

--- Code: ---Stage 1 is 0x0,000,0000 0x0,000,0000
--- End code ---


for the TEST 1 I've also found this:

 

The same thing goes for TEST 2, 3, 4 and 5... the rest seems fine.

And U50 and U59 are a pair of SDRAMs on the board, that constitute a 32-bit word. They are both correctly connected to the VDD, VDDQ and VSS and have their control signals and addresses tied together (I've measured that). But I have no idea, where they should go after that. I'm guessing, that one of the control lines of the SDRAMs might be severed, and because of that, the whole bank is dead. Correct me if I'm wrong  :palm:.

The acquisition ICs are U22 and U45 (the big ones under the heatsinks)? Virtex FPGAs next to them are the glue logic to the memory?

MarkL:
U45 is one of the acquisition ASIC and is known in pv as "Chip 9".

I think you're correct that the Virtex FPGAs are glue logic for the memory.

I think you're also probably right that there's a common signal for U50 and U59 that is dead and is causing them both to read as all 0's.  If so, it's probably going to be between U50/U59 and U52, or between U52 and U45.

I would examine ALL the traces on the bottom for breaks between U45 and U52.  If nothing is obvious, I would double check each trace with a continuity tester from VIA TO VIA.  You need two very fine and sharp probes to poke sideways at about a 45 degree angle into each via.  Start where you removed a runner (there's a label "AF2" there) and work your way towards the edge of the card.  You'll probably need a microscope (I do).  It's easy to lose your place, but you can mark your progress with a fine Sharpie or small removable labels.

The corrosion especially likes to attack a tiny ring of exposed copper around the small solder pads.  It's impossible to see which is why it's important to test end to end by probing the vias.

Another approach would be to turn the chassis upside down and access the card from the bottom while it's running.  You can remove the mouse/keyboard card and put it aside.  All signals, even those on top, can be probed from the bottom on the small solder pads (at least I haven't found any exceptions to this yet).  You would need to figure out which ones are CS#, WE#, CLK, RAS#, CAS#, etc. for U50/U59 with a continuity tester.

Then use a script to repeatedly run anlyBusTest.  Compare the signals on U50 and U59 with another pair of memory chips that aren't having an error.  You might be able to identify which signal(s) are missing and trace it backwards that way.

Example script that runs anlyBusTest on slot E once a second:


--- Code: ---#!/usr/local/bin/bash

# export PVRESULTLEVEL=9
# export PVDEBUGLEVEL=9

(
  sleep 5

  while true; do
    echo "s e"
    echo "x anlyBusTest"
   
    sleep 1
  done

) | pv

--- End code ---

I have bash loaded on my system.  I'd highly recommend it, but if you don't use it you can adapt the script for the default system shell.

MarkL:
Another thing you might try...

You'll notice that there are numerous 33R resistor networks between U52 and the memory.  They are for series termination for the control and data signals leading to the memory.  I've had several cases of open single resistors in other areas, so it might be worth checking these resistor networks.  The control signals appear to be mostly on the resistor networks directly under U52 (mounted horizontally).  The data signals appear to be mostly on the ones mounted vertically on the top and bottom.  But I would probably check them all anyway near U52 since it's very easy to do.

nikodem:

--- Quote from: MarkL on January 23, 2021, 04:00:33 pm ---Another thing you might try...

You'll notice that there are numerous 33R resistor networks between U52 and the memory.  They are for series termination for the control and data signals leading to the memory.  I've had several cases of open single resistors in other areas, so it might be worth checking these resistor networks.  The control signals appear to be mostly on the resistor networks directly under U52 (mounted horizontally).  The data signals appear to be mostly on the ones mounted vertically on the top and bottom.  But I would probably check them all anyway near U52 since it's very easy to do.

--- End quote ---

I've found out, that the resistors are terminating all(?) the signals to the memory ICs, but didn't knew, that thy migh be faulty! I will check that first, as this is ultra-simple  :-DMM

nikodem:
Terminating resistors seem to be fine. I've also checked RAS, CAS, CS and WE lines, and they are fine. Still hunting for the CLK, CKE, DQMH and DQML lines - they have to be somwhere on the resistor packs :-DMM.

I'm thinking about lifting the RAMs and/or exchanging them with some else RAMs, there is pleny of them.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod