OK, we have routinely used the JTAGICE-II pod to calibrate the R/C clock on ATMega32L parts. This has worked well until recently when we upgraded our Studio 4 installation to the most recent version, 4.18, which reprogrammed our pod. That is when things got interesting.
Our normal procedure programs an ATMega32L with the calibration routine, calibrates, tests for a non 0xff calibration byte in EEPROM, then load the program.
With the new installation we found that calibration would proceed normally and give a reasonable value. But, if we executed a read or read-verify of the device before erasing the calibration program from Flash, we frequently got a change in the calibration value; typically we would get 0x60 instead of values around 0xB5. Our firmware uses the calibration value to fine tune timing and the drastic drop in clock frequency raised merry hell with the application.
Rolling Studio 4 back from 4.18 to 4.12 fixed the problem (after the program reprogrammed our JTAGICE unit.) Anyone else seen odd behavior in the calibration program?