AM6411: Questions : SDIO, mode parallel input of the PRU and WBS TMDS64GPEVM vs SK-AM64

Geoffrey Ficara

Part Number: AM6411

Hi,

Here are a few questions :

1) We noticed a difference between TMDS64GPEVM & SK-AM64 : that there are pull-up resistors at the SD card interface of the SK-AM64 and not for the TMDS64GPEVM, why and could it cause integrity signal issues if they are not present ?

2) When using the PRU-ICSSG in parallel input mode, page 3441 of the TRM, GPI0->16 are used and GPI17-> 19 are greyed out. Does it mean GPI17-> 19 can't be used as Direct Input ? If not, how can I use the PRU to receive other type if signal such as ChipSelect signal?

3) Can you confirm that they instruction wbs in assembly with the PRU is active on the rising edge and not on the state 1 ?

Regards,

Geoffrey

over 3 years ago

0 Andrei Aldea over 3 years ago

TI__Expert 3310 points

Hi Geoffrey,

I will need some more time to look into the historical reason why the GP EVM has DNId the resistors, but the pull recommendations are part of the SD Specification.

I will refer you to the schematic review checklist which covers SD use cases and gives recommendations:

https://www.ti.com/lit/an/spracu5b/spracu5b.pdf

For your remaining questions please allow me to route this to one of my colleagues

Best

Andrei

0 Geoffrey Ficara over 3 years ago in reply to Andrei Aldea

TI__Expert 3365 points

Hi Andrei,

Thanks. FYI this post is related to this one : https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1092902/tmds64gpevm-reduce-mmc1-clock-to-permit-integration-of-sdio-wifi-module

The goal is to communicate with a WIFI module using the SD card interface, and there were some signal integrity issues with the TMDS64GPEVM, which could be possibly explained by the presence or the absence of the resistances.

Out of all the question this is the priority as they can't progress in the evaluation of the EVM.

Regards,

Geoffrey

0 Andrei Aldea over 3 years ago in reply to Geoffrey Ficara

TI__Expert 3310 points

Hi Geoffrey,

Thank you for the background. I'm looking into it now, agreed that normally they should be populated. I'll let the owner of that thread go into detail for what's needed on the SW side you will be able to restrict mode and speed via a DT change in Linux, I believe you're on the right track there trying to lower clock speed.

Best,

Andrei

0 Andrei Aldea over 3 years ago

TI__Expert 3310 points

Also I'm still looking into answering your question #2 but currently I don't believe there is any limitation in using GPI17-19 as input/output. Will reply shortly with the definitive answer.

For #3: WBS (Wait until Bit Set) will come out of wait on the rising edge.

0 Geoffrey Ficara over 3 years ago in reply to Andrei Aldea

TI__Expert 3365 points

Hi Andrei,

Thank you for your feedback. Please let me know when you can give us the complete answer, especially for #1.

Regards,

Geoffrey

0 Andrei Aldea over 3 years ago in reply to Geoffrey Ficara

TI__Expert 3310 points

Hi Geoffrey,

Good timing, I've discussed this internally and we have concluded this was a design oversight on the part of the design team. They had used an older device to determine that the pulls would not be needed but on this current implementation you actually should populate these resistors.

Our final recommendation is as follows:

Use External Pull-ups on CMD and DAT0-DAT3 (do not enable internal pulls)
Use External Pull-down on CLK (do not enable pulls on CLK)

We will look at also updating future builds on the AM64x GP EVM to include this change and note it as an Errata in the next revision of the User Guide. I appreciate you raising this concern.

Best,

Andrei

0 Geoffrey Ficara over 3 years ago in reply to Andrei Aldea

TI__Expert 3365 points

Hi Andrei,

Ok we will try to hook 10K Ohms resistors up and keep you updated.

Regards,

Geoffrey

0 Andrei Aldea over 3 years ago in reply to Geoffrey Ficara

TI__Expert 3310 points

Thanks Geoffrey, I'll keep the thread open. Please make sure you disable internal pulls in the DTS (assuming you're using Linux) if you populate the external pulls.

Best,

Andrei

0 Geoffrey Ficara over 3 years ago in reply to Andrei Aldea

TI__Expert 3365 points

Hi Andrei,

Any reason why we should use an external pull-down resistor on the CLK ? Adding a resistor between the CLK and the GND does not seem to be straightforward, so can we add a pull-up, or simply leave it open ?.

regards,

Geoffrey

0 Andrei Aldea over 3 years ago in reply to Geoffrey Ficara

TI__Expert 3310 points

Hi Geoffrey,

The CLK pull is not part of the JDEC MMC requirement because they expect the clock to always be driven, but we recommend external pulls anyway because during reset the signal may not be driven. You may be able to get away with just leaving it disconnected, but my recommendation is you leave a resistor to the side of one of the pads and run a thin wire to ground. I don't recommend a pullup.

Best,

Andrei

Processors

Processors forum

AM6411: Questions : SDIO, mode parallel input of the PRU and WBS TMDS64GPEVM vs SK-AM64