I've some 10 similar boards with ATmega169PV - 8MU - 0642 manufacturing lot. I am using SPI - ISP to program the flash.
This chip uses internal 8MHz oscillator, and it is divided by 8 internally (thru DIV8 fuse), so the core is intended to run at 1MHz.
Vcc is stable 3.3V, obtained from a lab bench-top power supply.
Reset line is on board externally pulled up to 3.3V by a 47k (47k)resistor.
All the boards have virgin ICs, and I use STK500 to do the ISP.
Problem I am facing is, in many boards, initially I was able to program the chips, but later on, the signature is read as 0x7F 0x7F 0x7F. I programmed a simple LED blink code, and then, say after an hour, the chip is not getting recognized by STK500.
I tried to probe the SPI signals thru o'scope, and I see no noise or ringing. I've 100 Ohm resistors in series with SPI lines to tame the ringing.
I also noticed that, during "Read Signature", the MISO line gives some signal between 0 and 3.3V (I mean proper signal levels), but STK 500 reads the signature as 0x7F.
This problem is getting peculiar. In three of the ISP-problem boards, I tried to program the chip thru JTAG mkII, and it works perfectly fine. Even, the fuse bits are correct and intact for ISP operation. But even after that, SPI-ISP interface doesnt work.
SPI signal from STK500 physically reaches ATmega169PV chip.
All good care for ESD has been taken care, and all the above are tested in lab environment.
I came to know that ATmega...PV is rolled out by a new silicon process to reduce the power consumption (picopower), so I am suspecting even the batch of ICs delivered. I am using 64 pin MLF package, so its really hard to probe the IC pins, except the associated PCB traces.
I've sought Atmel tech support, nevertheless, Any of your help is highly appreciated. Thank you.