MSP430F2132 programming issue

Additional info from the customer;

We bought a full reel of 1000 from Digi-Key which they packed 9 June 2011, and the lot code on the Digi-Key label is 1616957ML1 date APR-2011.

The markings on the packages are

MSP430
F2132
TI  14KB
DXKT  G4

We had 800 Emitter Driver cards assembled, with one micro on each board. 

Of the 800, 592 were programmed once and tested 'good', that is, the program ran satisfactorily for the test.

Of the remaining 208 installed micros, 8 were damaged in testing by accidental application of 7 volts power, and they are presumed to have been good, otherwise.

The are about 5 boards where the micro is running OK but there are other unrelated problems with those boards.

The 195 'problem' micros out of 800 installed include about 25 or 30 which have been programmed once, but which will not program again, even with repeated tries.  This is the strangest problem.  We have two programming pods, and the problem seems to exist with both pods.  The remaining 165 or so boards will not program at all.  Once we noticed this problem of not being able to program repeatedly, we began programming 3 times before accepting the board for test, but we still have the 592 boards which were programmed only once before we knew of that problem, were tested, and soldered into larger assemblies, so I am worried that some of those boards may end up not being able to be reprogrammed.  We will find out when we go to reprogram them in the larger assemblies.

Of the 165 boards which will not program at all, we have so far replaced the micro on 24, and of those 24, 18 could be programmed 3 times, and tested 'good', and 6 could be programed once, but not repeatedly, and are therefore 'bad'.

So, the evidence so far suggests that about 1/3 of the micros may be 'bad' in the sense that they will program once, but not again, or, will not program at all.

-Tom-

Correction from customer;

Of the 165 boards which will not program at all, we have so far replaced the micro on 24, and of those 24, 16 could be programmed 3 times, and tested 'good', and 8 could be programed once, but not repeatedly, and are therefore 'bad'.

So, the evidence so far suggests that about 1/3 of the micros may be 'bad' in the sense that they will program once, but not again, or, will not program at all.

Comments......

-Tom-

Do you use SBW protocol for programming?
Do you have additional circuitry on the RST pin?

SBW uses RST line for data transfer, so capacitance on this pin will make SBW unreliable, preventing SBW communication.
However, come capacitance (and a pullup resistor) are recommended for a proper startup, especially on slowly rising VCC.

The TI recommendation is to have at most 2.2nF onthe RST pin. My own suggestion, since 2.2nF are quite low for some slower power supplies for a reliable startup (FET not attached!), to separate the capacitor/pullup combo from the RST pin and SBW signal by a series resistor >1k. This way, the capacitor is decoupled form the high-frequency SBW signal but still does its job when the FET is not controilling RST.

So if the resistor/capacitor combo is on RST, capacitor tolerance might be the reason why some boards work and some don't.