Wookey wrote:
> Just occaisionally, when you power it up, it doesn't start properly,
> stays in reset (red LED on) and draws shedloads of current (couple of
> amps, limited by PSU). The time I caught it and checked, it was the
> CPU that was getting really hot.
> 
> Cycle the power and the board works fine.

I've definitely seen this before, on probably less than half a dozen
Toby Churchill boards. Sometimes they'll latch up like that and recover
after a power cycle, and other times the consequences are fatal (for the
board, not the user). I've seen it happen to both the CPLD and the CPU.

I'd be surprised if it was a power supply sequencing thing: there's
effectively only one power rail (the 3.3V one), and the 1.8V CPU core
supply is derived from that and controlled by the CPU itself. The only
notable power supply weirdness I've seen multiple times is when one of
the feedback resistors on the core regulator is mistakenly fitted as a
capacitor (goodness knows how this happens) and the regulator
oscillates. That normally results in a board which won't run at more
than 133MHz, though.

Thinking about this, here's a hypothesis: suppose the PWR_EN signal from
the CPU was disconnected/faulty at its Evil Resistor Pack, as is not
uncommon, and the core regulator's enable input floated high, thus
powering the core before the CPU was really ready for it. It would be an
interesting experiment to try, anyway, but wouldn't explain CPLD latchup.

The fault has been sufficiently rare and hard to reproduce that it's
been easiest just to chuck the offending board into the
component-recovery box and forget about it, so there hasn't been any
earnest attempt to track it down on my part. If you've got a board which
does it repeatedly (and recovers ;-) then that's a good start...

Chris

-- 
Chris Jones - chris@martin-jones.com
Martin-Jones Technology Ltd, makers of Solidlights
148 Catharine Street, Cambridge, CB1 3AR, UK
Phone +44 (0) 1223 655611 Fax +44 (0) 870 112 3908
http://www.solidlights.co.uk/