LTM4676A disappears from i2c bus and asserts both RUNx pins low while doing NVRAM programming

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!

Top Replies

  • Update: deadbugging the DC1613A to my board, I am able to see the LTM4676A moved to i2c address 0x7C.  Most of its register settings were corrupted, but I was able to manually program them to good values.  I also could force the ignore-external-pin-strappings and force the upper i2c address back to 0x4F before committing it all to the NVRAM.  However, the error flag is set that the factory-trim NVM area is corrupted/bad CRC.  Could you provide a programming file for LTPowerPlay (or just register{offset, value} list we can write over SMBus) for the factory defaults/calibration settings (reserved registers we don't have a way to know about) so this module could be used again without replacing it?  Thanks!  

  • +1
    •  Analog Employees 
    on Mar 5, 2019 7:09 AM in reply to sio2usa

    Support was taken off line. This is considered closed.

Reply Children
No Data