Thanks Kelvin and Jason.
OK - I think we have a work around for the temp sensor issue.
We'll find a way to jumper it to a valid level or disable it
via software and offer the buyer a TBD discount of some sort.
but now the 1V5 supply issue...
Boy I hope this isn't another strange big Xilinx BGA issue
like what happened on Roach2 010203.
It's probably best if I visit Digicom and help you debug in person.
As Jason says if we use the roach_monitor to watch the temperatures
and voltages
we may see the FPGA or PPC heating up due to a short or someother
voltage going wonky.
or feel a logic translator or maybe QDR VTT&VREF generator or
QDR memory getting hot, ...
that is check all the loads of the 1V5 supply and look for something
amiss.
could there be something like as happened on at least 1 BWRC Roach1 board
?
The large current sense resistor R296 had failed somehow and
with use would open up so there'd be too much Vdrop w/ load.
fix was to replace R296.
Do you see valid 1.5V right a power up (unprogrammed big FPGA so as
to minimize current demand) ?
I'll still think about when I can visit in person...
Matt
On Wed, 2 May 2012, Jason Manley wrote:
As suggested by you, we shorted pin 2,4,6 & 8 of J31 together (you
mentioned 1,3,5 & 7 but I thing you mean 2,4,6,8)
Nope, I definitely meant 1,3,5,7 (1&3 and 5&7 should already have been
shorted) as shown in the v1.3 schematic. This would have shorted all the
return lines from the temp sensors together. It's unfortunate that this
didn't work. If the purchaser of this board doesn't mind (maybe you can
offer a discount or something), I'd suggest simply disabling the
hardware safety checks (by setting that software flag using
roach_monitor.py).
Next thing was we ran burn-in of the board. The board ran for 30 minutes and
suddenly the protection kicked in, it powered up and down again and shut
down afterwards. We monitored the low voltage supply and observed 1V5
dropped to 0.9V. I believed something that killed the U51 before was killing
it again. At this point we don't know how to proceed. I have tried shorting
1,3,5,7 of J31 together but it didn't seem to work. Do you have any idea?
What caused the shutdown? Was it definitely the 1v5 rail? Again,
roach_monitor.py should tell you the reason for the last shutdown. Is
there something on the board that was getting hot before it shutdown?
That PSU can supply a lot of current, so I'd imagine any solder hairline
traces would have burnt through already.
Jason
