We have a design with LTM4676A, 4675, and 4678 on the same bus. We make several register changes from how the modules ship from the factory, then save NVRAM. In a seocnd step, we do a restore from NVRAM, then write a serial number USER_DATA_00 and USER_DATA_02, and then an NVRAM save. This last sequence is in compiled code, and we wonder if we are doing it to fast and corrupting a reserved register. About half of the units we did this, they assert both their RUN0 and RUN1 pins and will no longer show up on the i2c bus. We tried writing 0x4f to 0xe6 in case it was just the MFR_I2C_ADDR upper bits, but that doesn't help. Power cycle also doesn't help.
So far we've found no way to recover and want to avoid having to solder new parts. Is there a way we can reprogram factory defaults and try to recover these? We have you eval kit and dongle and LTPowerPlay, and have succesfully bridged the eval board to our board in the past--if that helps. Thanks for any advice what reserved registers or other mechanism by which the 4676A can get in this state. Thanks!
