Mini board repeatedly looses programming

I am using the Mini Board in an outdoor art installation. It works great, but seems to loose it's programming at the drop of a hat. The boards work for an indeterminate priod of timw - sometimes as long as days, other times just mnutes. Then, with the least disturbance - sometimes even just the power cycling on/off causes the board to stop working. The symptoms are that I see the three flashes as it starts, then it either does nothing, or else the LED on pin 13 begins blinking, and the board can't be reset at all (the reset button does nothing) the only way to reset it is to power down/up.

Invariably - reburning the code causes it to instlantly be healthy again. Until it happens all over again. I wouldn't mind if the board reset occasionally - but to totally loose its programming is a big problem.

This has been making my life miserable - if you have any ideas as to why this is, I'd love to hear them. At this point I'm considering a solution using a high-voltage burned ATmega168 which I would certainly think should work. Perhaps faricating a PCB board with the chip, a regulator, and an oscillator that would be the size of the stamp, and I could just swap in. But the arduino surely isn't this unstable - I must be doing something wrong.

Some details - I'm powering the Arduino Mini with 9VDC regulated external power at the +9 pin. I've put a 10K resistor between Rx and ground, just to be safe. I've tried shorting together the RX and TX pins as well (after programming) but this has no effect either. I'm using 5 boards simultaneously, each with a PCB I've designed that interfaces to a set of small speakers I am sending pulses to - all 5 boards have the same problem.

Thanks in advance for any ideas you might have.

Well I am quite new but these are my thoughts... questions....

Sounds like it is a software problem, but have you tried running some standard program to see if the are any hardware problems? Have you tried a different power supply, or ruled power supply problems? What is the program you are running? Could a variable be becoming to large? What sort of pulses are you sending them, could they be the problem?

I would try adding 100 Uf or 470 Uf filter capacitors to the +5 supply and the regulator input. This kind of thing was reported a while back and the only thing we ( the forum types) could trace it to was the fact that one of the bootloader versions did not have the brownout detection enabled. We talked to Atmel and it seems that sometimes sudden power loss can corrupt the Flash. Of course this is all unverified, but you could try the caps.

I can't remember the details, so see this thread:

http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1184252197;start=all

D

Thanks for the suggestion on the Caps - I in fact, on the newest version of my board went to larger caps on the power supply, and the Problem does seem to have improved at least somewhat. I suspect that in the end the power supply must somehow be the cheif suspect. I Haven't followed the link yet, but will, and will report back if I figure anything else out for sure -