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.

CLK questions on Keystone

I was wondering if you can help with the following questions.

 

1.

I am concerned about our CORECLK and DDRCLK sequence, the datasheet states;

And

Our clocks are under software control and come up after POR but before RESETFULL. I believed that RESETFULL was the same as POR but I am now not convinced. Is our implementation okay?

 

2.  We have been encountering problems with pcie link up:

 - pcie fails our training, resulting in link down.

- problem is intermittently on 2 out of 7 boards, not yet seen elsewhere;

- the main DSP clock is available as soon as DSP boots, but pcie clock is available only during kernel boot. The main PLL and PCIe PLL both properly locked prior to link failure;

- the problem can be lessen or worsen by playing with the timing and delay between the steps of either side. Adding delays sometimes makes it better, sometimes worse. However I haven't found a solution that gives 10 consecutive success.

 Is there any known issues with this on single core. We did not see this with the eval cards.

  • Hi,

    1. As specified in the data manual, the clock must toggle at least 500 times before PORz is released. The delay between PORz and RESETFULLz is needed to be sure that the boot settings are latched correctly but you should start your clock before PORz is released. 

    2. I haven't seen anything like this reported. The EVM doesn't delay the PCIECLK until after the reset is released. The PCIECLK begins toggling at the same time that the CORECLK and the DDRCLK. While it isn't necessary to have the PCIECLK toggling until the PCIE is needed, it can be difficult to coordinate that transition. My recommendation to customer is to start all the clocks needed by the SOC before PORz is released. 

    Regards, Bill