TI Team,
I was wondering if you could comment on the following problem we've been having with a batch of MSP430s. How do you suggest we fix this problem?
Problem Statement
We use the M430F5438A part. We noticed that with a new lot of MSP430’s that we failed 100% of our boards tested. Looking at this on the bench, it was narrowed down the TI chip since we noticed VCORE was dropping and causing a reset (BOR or POR). The supply for the LDO that creates VCORE on our system is called DVCC, and looking at this on a scope, it didn’t move when VCORE was dropping. So I assume this is internal to the chip.
We program our device to output a square pulse every 12.5 - 250 ms. Our firmware loads a DMA table up to reflect this state machine at the very beginning when programmed and then only wakes up to reload the timer to wrap the DMA table around to the beginning. We measured VCORE and it is dropping at a rate of 1380V/s, which off of a 470nF capacitor is ~650uA load . This coincides with when our state machine in the DMA wakes up and copies values to output a pulse on a few ports on P9 and P10.
The odd thing is that these pulses will come out fine for 100s of ms or even 10s of seconds and then all the sudden VCORE drops during one of them. We looked and this does not seem to correspond when the DMA timer is reloaded. We look at all other ports and nothing external is different during this time. The number of MCLK cycles is consistent with all of the good pulses.
We looked at the BGAs under a microscope and the solder joints looked fine. We also had our contract manufacturer do the same. After that, we had our contract manufacturer replace the MSP430 with a known good lot and the board works fine. The known good lot is 11CE03T and the known bad lot is 23C58TT. Both are Rev E. So far every part from the bad lot fails (we stopped at 25 units), and every part from the good lot passes (we’ve built 100s from that lot).
Again, what do you suppose is causing the problem?