I have run into what appears to be a hardware problem with the 80-pin
"XMS430F5438" chips. I am sourcing ACLK from XT1 which is using a
32.768KHz crystal. Instead I am seeing an ACLK frequency of about
30.6KHz. This is observed from the RTC period and from the async baud
rate, both of which are driven from ACLK.
I originally had this running on the MSP-TS430PZ5x100 eval board using
the crystal supplied with the board.. The ACLK frequency was correct. I
am now running the same code on a board using the 80-pin chip (which is
still labeled X430F5438 and which identifies itself as an XMS430F5438)
and I am getting the incorrect ACLK frequency.
1) I am using the same crystal that was used on the eval board. (I
removed it from the eval board and moved it to the new board.). I
initially used another crystal which I believe was a 32.768 crystal with
the same results.
2) I have loaded exactly the same code image into the two chips. The
100-pin chip on the eval board gives the correct ACLK frequency, the
80-pin chip does not.
3) I've tried this on two different 80-pin chips with the same results
on each.
4) I have observed a sine wave on the crystal. As best as I can measure
the frequency with my scope, is slightly less than 33KHz, so it appears
to be oscillating at the correct frequency.
5) I also thought there might be something funny about the ACLK
selection, so I tried changing it first to use VLO and then REFO. In
both cases, the behavior was as expected. Neither resulted in a correct
32.678KHz ACLK. It is interesting that selecting REFO gave exactly the
same ACLK rate as did selecting XT1. This would tend to imply that XT1
is not working and REFO is being automatically used instead. This is
reinforced by the fact that the XT1LFOFFG bit in UCSCTL7 is set every
time I look at it after a reset. If I clear it, it stays cleared and the
frequency stays the same.