- Ask a related questionWhat is a related question?A related question is a question created from another question. When the related question is created, it will be automatically linked to the original question.
Currently I'm working on a custom board with a TI Sitara AM3359, two DP83848I PHYs and some other devices. These two PHYs are connected via RMII. We use an external oscillator for the reference clock of the PHYs because of Advisory 1.0.16 in the AM335x errata sheet.
Now the problem is that the first PHY is found on Linux (using TI SDK 05.05.01), but it is impossible to get an ethernet connection. We assume the reason is an interference at the reference clock to the PHY.
As long as the Linux kernel isn't running, we can measure a clean 50 MHz frequency at the PHY reference clock input. But when the kernel starts, we see an interference. We guess this comes from the rmii1_refclk pin from the AM3359. It outputs a 52 MHz frequency even this pin is defined as input in the board file (this 52 MHz frequency is also there if this pin isn't configured at all). We suppose the AM3359 has to be telled that it shouldn't output this reference clock. The gmii_sel register, bit 6 and 7 seems to be the candidates for doing this.
Could this be right? What else could cause this behaviour?
