We had a problem similar to the 'USB PLL drift' reported on the Advisory 2.1.21 of the C6748.
The problem ocurred when temp was rising from -30 and the USB clock was drifting starting around -20Celcius.
I captured Tektronix USB scope reports every few degrees. (I have the files and Excel spreadsheet shoing the drift.)
I think there should be something in the C6747 errata sheet too about this.
Reported problem:
TMS320 USB port stop working in ALT when temperature rise from -40 to around -5 or 0C. ALT use TMX parts.
Cause found:
TMS320C6747 clock drifting issue in SAM ALT.
Background:
During ALT, the SAM Product USB port would disconnect for no apparent cause and stay unresponsive until reset (Power down from outside chamber or with Watchdog reset command if UART was connected).
Sometimes, it appear the UART access would also fail. [This could not be verified]
The Tektronix DPO7254 scope and USB sQUIDD were connected to the USB bus to check the signal quality while the temperature was rising. Two tests sequence were made to determine if time/temp were causing problems.
First test (SignalRateDrift.xls)
The first test proved that the USB data rate was beginning to drift around -20 Celcius while the temperature was rising. This appears to match the ALT symptoms, even if the cycle was started at -33 instead of -40, the drift seem to begin around -20C. The USB EYE diagrams did not appear degraded during the rise.
Second Test (SignalRateDrift2.xls)
In this test, the chamber was put on temperature hold at -15 Celcius for 35 minutes to check the USB data rate behaviour with a stable temperature. The clock error seem to stabilize at the wrong value. Resume the chamber cycle (rising to +25c), do a power down near 0 Celcius.
It appears the drift is solely dependent on temperature change, it will stay stable (even bad) if the temperature is stable.
Power cycling/reset seem to fix the problem when the temperature is in range.
Conclusion:
For the SAM, the operating range is 0-40 celcius (commercial part).
The fact that the part is not working properly at -20 Celcius should not be an issue, as long as it does work properly in the expected range +/- 5 Celcius.
For the Accelerated Life Test process, A short step was added around -5C to power down the units to clear the USB drift. This corrected the problem of the lost connections.
Notes:
The additionnal test were done with production TMS320C6747, but not sure which die revision.
