This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DRA821U: PCIe Serdes I/O and Power Sequencing

Part Number: DRA821U

It is my understanding that, in general, voltage should not be sourced into an I/O line (or at least current should be significantly limited) until the I/O bank power rail is powered up. Otherwise, the ESD structure clamping diodes can shunt the voltage to the power rail and prematurely raise its voltage.

I have a scenario where we have a SOM module with the DRA821U and PMIC Power solution (PCIe master) that is then inserted into a carrier board (PCIe slave) with its own separate power sequencing requirements (FPGA) that will start asynchronous to the DRA821U power sequencing. In this scenario, it is going to be extremely difficult to sequence the SERDES power rails on the DRA821U at the same time as the SERDES power rails on the FPGA. It is also not possible to add any significant series resistant on these high speed lines to limit clamping current.

I'm wondering if there is a sequencing concern with the SERDES I/O and power rail sequencing on the respective devices coming up asynchronously. If for example, the SERDES lines do not enable their outputs, or if the outputs are driven low, until the software initializes the PCIe and SERDES peripherals, then I do not expect any issue with voltage being present that would clamp to the power rails through the ESD structures.

The SOM and carrier board share a common ground reference.

Please advise.

Thanks,

Stuart

  • Since my original posting, I have been investigating this a bit on my own. I have some additional understanding, but I would like some confirmation that it is correct.

    It seems that the PCIe data lines (SERDES) should be AC coupled with DC blocking capacitors placed near to the transmitter side. Given that this is the case, it would seem that power sequencing between the master and slave is less important because the series capacitors should block any DC leakage between the two devices' SERDES I/O lines. As long as neither side begins transmitting data before both sides are powered up, I think we should be okay, regardless of sequencing.

    Can I please get confirmation that my understanding is correct?

    Thanks,

    Stuart