Lpc2000

I recently posted about sourcing the lpc2106 due to burning a few of 
them lately. Leon Heller asked as to how they got fried so I put some 
information together. Now this has occured with the lpc2106 board 
from olimex, both the 40pin header and proto board. Im using 
Crossworks.

The wiggler-compatible JTAG connector (Olimex) seems to be causing 
sporadic mishaps, which in turn fry the LPC2106. The normal 3.3V 
supply current to the JTAG connector itself is around .7mA, going as 
high as 1.7mA when downloading a program to the '2106. When the fault 
condition is present, however, the supply current increases to 37mA. 
I set up the JTAG connector to be supplied by by a sourcing-only 
power supply (op-amp with a transistor instead of pulling a wire from 
the headerboard since I thought it was sloppy to wire to the supply 
on the 40-pin header), and it appears as though the 3.3V supply 
current is being sent from the supply through the JTAG connector, to 
the LPC2106 straight to ground, since the supply's transistor always 
becomes hot when the '2106 becomes hot. The only way the 3.3V supply 
and the '2106 are connected is through the JTAG connector. I am 
speculating that an output on the JTAG connector is trying to pull a 
pin high that leads to the '2106, which in turn is trying to pull 
that same pin to ground. I suspect that this might be a flaw in the 
JTAG connector itself, since nothing but the JTAG is connected to the 
processor. I don't know which pin on the JTAG cable might be at fault 
here, and I cannot reliably produce the fault condition that fries 
these chips. On occasion, some action on the software side, such as 
disconnecting or processor reset causes the fault to occur.

I greatly appreciate any ideas or comments on this issue.
Thank You!
Eric