Now I've done it. This is an LM3S6432 in a S2E circuit module based on the RDK-S2E circuit, using a JLink programmer. It is basically the same electrical circuit as ref design, except board form factor changed to fit our space.
Absolutely everything works, remember that. We can program in TI's own software, and we program our own version of the software, everything verifies and runs fine. Ethernet works and runs fine, LM Programmer runs fine over Etherent, etc. The S2E stuff is working just fine.
Except on our latest board rev, when the JTag connects: it reports a warning that the CPU speed is 15998 kHz, not 8000 kHz has expected. We are using the same crystal model as on the S2E design. When I plug the J-Link into the S2E module or our older rev boards, the Jtag will show the correct CPU clock speed of 8000 kHz.
The only thing altered was that the 8Mhz crystal is as close as possible to the MCU chip on the new rev. Trace length went from about 6mm down to around 5mm. Which theoretically should be better. The traces just pass over a ground plane, there are no other signals nearby.
There must be something else I'm not seeing. The other clue is the 8Mhz clock, measured on the S2E module crystal, is around 400mV pk to pk. On the new boards I'm seeing around 650mV pk to pk. I checked the load caps, and they are 18pF. I pulled the crystal off the working S2E board and placed it on our new board rev, and no change in symptom. I put the crystals from our inventory onto the S2E board and they seem to work fine, the JTags detects the correct CPU speed fine as 8Mhz. A sanity chack on the board traces shows no shorts or opens or high impedence traces around the oscillator area.
Any clue here as to what else might be going on? Where else to look??
I'm not TOO woried since everyting is apperently working, but I suspect there is a clock instability while the JTAG is trying to connect, and I want to make sure this problem is diagnosed before we send this batch of products into the field.
Thanks!!