10 Mar
2016
10 Mar
'16
4:39 a.m.
On Wed, 09 Mar 2016 21:59:27 +0100, Robert Jarratt
<robert.jarratt at ntlworld.com> wrote:
I think one day I will have to equip myself and learn
how to desolder
and resolder surface mount chips. I don't know how many chips implement
the B-CACHE, but perhaps you could replace all of them, assuming you
know which ones they are. Perhaps some careful probing of the board with
a scope might show if there any chips that are perhaps completely dead.
You could do the same for the DSSI controller on the other board of
course, if you can identify that.
Our lab has an EE department with some pretty fancy SM gear, I just have
to practice on it (which would come in handy anyway). The B-cache consists
of 18x CY7C166-20 SRAM plus 5x CY7C170A-25 tag RAM, so they're easily
identified (the KA-675 manual infact points them out).
Replacing 18 SRAMs doesn't sound like a lot of fun, and I have no idea how
to probe the board in the cardcage without disassembling the chassis, in
which case the thing may overheat; opening the CPU bulkhead infact
triggers a prompt reaction from the fan controller.
FWIW, here's a thermography (hope the link works) of the B-cache section
of the KA-675 after being powered up for ca. 30 mins:
https://www.dropbox.com/s/2nsx1dfngp1jfq5/TH710065.BMP?dl=0
Note that the rightmost chip just below the CPU heatsink has a pin that's
ca. 2 degrees warmer than the others. I don't know if that says anything,
but it *is* reproducible. Maybe I should start with that one.
The dead DSSI controller on the other board is easily identified, as it
was physically destroyed in transit. I'm not inclined to transplant a
160-pin PGA as my first foray into SMT... :^\
Thanks for the feedback,
--GT
--
"END OF LINE" [MCP, 1982]
"... nowhere in the standards is it specified that 'programs that use a
lot of memory may randomly crash at any time for no apparent reason'"
[Stackoverflow forum, 2012]
