MPC7448 - Checkstop immediately after releasing HRESET

[jabreen]

We have an MPC7448-based board where the '7448 is waking up dead. No bus activity. We connected a Green Hills Probe (GHP) via JTAG, but from the GHP's vb command, it appears that TDO is either driven high or tri-stated (it's pulled up on the board).

The only visible activity is that CKSTP_OUT is negated (high) while HRESET is asserted; then as soon as HRESET is negated, CKSTP_OUT asserts, which I assume means that it had a checkstop/machine check immediately after coming out of reset.

I can't find any likely cause for a machine check before the first bus cycle. It appears that most causes have to be explicitly enabled, or at least occur during a memory access. If there are other "implementation-dependent" causes, I can't find anything on them.

I should mention that this is one out of a number of working boards of the same design, so it's likely a manufacturing issue.  But I still need to find the cause.  We've checked the power rails, input clock, resets - just about everything we can access.

Has anyone seen anything like this?

[jabreen]

[Posted in wrong place, and I can't figure out how to delete it.]

[ufedor]

Please X-ray the board and check memory subsystem, also it is possible to compare the board with an operational one.

Can it be that /TEA is asserted?

[jabreen]

[Originally posted in wrong place]

Thanks for your reply!  Yes, we plan to X-ray it, and/or reflow the PPC.  We were trying to debug at a remote site, which didn't have that capability.  Now we'll have to wait until it gets shipped back to us.

I thought about /TEA, but we don't have access to it on the board (it's driven directly by an FPGA pin).  But would /TEA being low affect the PPC before it even starts an address tenure?

A few updates:

• I'm thinking now that the failing JTAG is more indicative of the problem than the checkstop.  I can't think what would make JTAG and the PPC itself inoperable.
• With a lot of playing around, we were able to get it to come up semi-reliably.  It takes a combination of hooking up scope probes and pushing on a couple of areas of the board with probes.  This would seem to confirm a solder problem, although just pushing on the board without other probes hooked up wouldn't cause it to fail, which points to some electrical effect.
• In a few cases, we got it where the JTAG would respond but the PPC would be halted.  Could it be some sort of power-on reset failure?  The resets look good at the board access points.

Once we're able to X-ray it, we'll probably be able to figure it out.  But I'm still curious what (presumably) single-point failure could cause complete inoperability.

Thanks again.