sidehack
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
December 23, 2013, 08:42:35 AM |
|
I've got a few like this, have yet to narrow down a problem. Spent most of today and yesterday working on one particularly problematic one. I had a blade reading as 00000000xxxxxxxx00000000xxxxxxxx and turned out to be a bad solder joint on an adress line for a chip select decoder, that was fun to track down... but the other one, with all X, I'm really not sure. ASIC I/O appears to work, clocking is good, address decoding is good, even swapped controller ICs to make sure that wasn't an issue and no dice. Might be one of the chips is spitting out erroneous data at the wrong time and interrupting regular comms, since all 32 chips share a data bus. Will probably only know by systematically removing every ASIC and buffer until it starts to work.
I did get one reporting all X during an overclock once when the new oscillator wasn't soldered properly and no clock signal was getting to the ASICs. That's an easy fix. With all X, it's such a complex system that there could be any number of things causing a total failure like that. As they say in lasers, you can't tune zero.
First thing I would do is look all the chips over closely to make sure none of them have any pits or craters indicative of having released the magic smoke. Just in case. If you got a scope, check that the oscillator output is firing, and making it through the 7404 to the buffer ICs next to each ASIC (74126?). I don't have a pinout in front of me, but I think the clock input is pin 2 and outputs on pin 3 into something like pin 7 of the BE100 chip. Also contact the reseller about a replacement is a good idea. Who did you buy it from?
|