It could be the state of the SYS_BOOT pins which regulate the boot order of the SOM. Sometimes they are shared with other peripherals and if they are held at the wrong values, the SOM may be trying to boot from something other than NAND.

There is a design checklist here:
https://support.logicpd.com/DesktopModules/Bring2mind/DMX/Download.aspx?portalid=0&EntryId=965

you'll need to be logged into the support site in order for the link to work.

I would review section 5 and 5.1. It has some suggestions.


The Hardware Spec also has some design suggestions:
https://support.logicpd.com/DesktopModules/Bring2mind/DMX/Download.aspx?portalid=0&EntryId=971

We also offer a support contract that includes a schematic review where two engineers at Logic PD would examine your board schematic and look for any areas of concern. Would you be interested in something like that?

adam

Hi Adam:
Thanks for quick response. Will review references you sent. Let me provide a little more background. Have been running with this design for a few years now. We did switch over to the industrial version of the SOM late last year. (I will check on your suggestion about a schematic review. I inherited this and will see if we did a schematic review near the start of the project.) Some of the customer complaints seem to be correlated with the switch to the industrial SOM, but not sure this is conclusive. We did have some assembly issues on htis batch of PCBs - SOM connectors had loose pins, but have touched up each pin. Believe this problem exhibited on more than the board that I have been working with - we did send board to another of our offices in preparation for sending to customer for approval (we did a board spin - can provide more details but most related to peripherals) and they observed similar reset cycling - theirs was not continuous as I observed. We boot from NAND. Will look to see if I can correlate start of reset cycling with any boot up events.

Hi Adam:
Section 5.1 mentioned above did help. We have LAN9221 on this version of PCB and somehow the solder stencil omitted ground pad (correct on previous versions) so we had to pull part and resolder. Looks like some of the uP_D0-D15 or uP_nWE or uP_nOE lines may have been compromised. If I tried to boot with SD card it consistently started (may not have finished boot) but that was the clue. Need to put the LAN9221 back on rest of the boards (and finish chasing any other manufacturing defects) but have three with good signal levels to all uP pins to LAN9221 and they boot properly. Thanks for your help. Any more insight into what is happenning when SOM gets into this reset cycling? Assume it is re-trying but do you have a reference that might give a bit more info?