Our implementation has a SECO designed Q7 module with an OMAP 3730 processor on it connected to a USB PHY over ULPI using the SMSC3320 PHY. The Q7 module installs onto our carrier and connects to an SMSC2517i USB hub. Periodically it looks like there are several re-tries between the 3730 and the USB hub and eventually we believe the 3730 issues a USB reset to our hub. We came to the conclusion that a reset is being issued because the PLL voltage generated by the hub does down when we start seeing error message reported in Linux. The PLL voltage going down is an indicator that a reset or disable command has been issued to the hub and the hub is going into a low power state. Given that our reset line to the device is not toggling and power is stable we believe that this is being commanded from the host.
We have noticed that the failure is rarely (if ever) seen when using a kernel from SECOs website which has no additional driver support for the USB hub. When we use a kernel built by Mercury which does include driver support for the USB hub we see these problems every few hours or so. On the surface this seems like it would be some kind of driver issue however, we noticed that when we did some testing at -40C the failures happen on the order of every 20 minutes or so as opposed to every few hours. So it certainly seems like there is some component level issue at play here as well.
I realize we are using the PHY that is listed in the devices errata as being problematic and our revision of silicon on the OMAP 3730 is susceptible to the errata but i don't believe that we are seeing the issue the errata describes.
Has anyone on this forum seen anything like this?
