Part number: S9KEAZN64AMLC
Clock: External 8MHz solid state clock that runs a KEA128 and KEA64 at the same time.
ICS: FEE mode to run the bus clock at 16MHz. Problem occurs the same way when I bypass the FLL completely.
LVD: Problem occurs whether or not LVD is enabled.
Given the right set of conditions, my KEA64 will refuse to boot AT ALL. As you deviate from -10C, to say +10C, it will take about 1 second to boot. Then at room temperature, it boots instantly. Again, given the right conditions, which are as follows:
Temperature soak module at -10C while unpowered. You must power the module initially by plugging in a banana jack into the power supply. I have a throw switch that I tested with that seems to bring up the power with a lower slew rate due to bouncing, and this CANNOT cause the issue. It will ONLY happen if you plug the module in with a banana jack, apparently achieving a very high slew rate. (On the 5V bus, banana plug in is 0 - 5V in 20usec, and throw switch is 0 - 5V messily in about 120usec). Keep in mind, I have not officially tested that a fast slew rate is needed to cause the problem. I only have two test cases and that's my best guess as to the contrast between them.
Also, the module must not have been powered in the last 30 seconds, no matter how brief.
I monitor the 5V and reset line while this happens, and both are active almost instantly, with no processor response. I monitor the processor's boot progress with a GPIO pin. Processor expert allows you to put your own code before PE_low_level_init(), which is the very first function in main(). I used it to initialize my GPIO pin (to high), and this happens in the __init_hardware() function. __init_hardware() runs before main(). I still see a low on the GPIO pin until the processor comes up and runs properly, which could be seconds later, or never.
I want to believe that I'm doing something wrong, but it seems silly that something like this can literally vary by temperature. This is REPEATABLE from module to module. It should also be noted that I've never seen this issue occur the same way with the KEA128 that is also on the board.
Any ideas, anyone ever see something like this before?