AM620-Q1: Changing MDIO frequency manually (Errata i2329)

Hello Til Schoengarth,

Wasn't the below link from previous thread helpful on the subject:

(+) AM625: Changing MDIO Interface Clock Frequency on Am625 Processor - Processors forum - Processors - TI E2E support forums

Let me know if you need additional guidance from the software team.

Thanks

Kind Regards,

Anastas Yordanov

Hello Anastas,

no, it is still unclear how to increase the frequency - the OEM requires more than 700kHz and it is not possible to change the frequency as of now.

From the E2E you shared:

" As Dominic has pointed out there is a silicon errata that requires using the kernel mdio/mdio-gpio.c and mdio/mdio-bitbang.c code. Looking through the mdio-bitbang.c I see some hard coded values that appear to be delayed based bit shaping that seems to setting tx at 2MHz and read is 1.5MHz. This is community code that is being used by TI and to my knowledge it is not modified "

-> We do not observe this frequency. What needs to be done?

BR,

Til

Hi Til,

I understand.  

I will contact the software team to follow-up on the topic. Please expect a feedback early next week but some delay might be possible.

I appreciate your patience !

Best Regards,

Anastas Yordanov   

Hello Anastas,

friendly reminder.

BR,

Til

Hello Til,

Anastas will be able respond tomorrow at earliest due to holidays.

Thanks,

Stan

Hi Til,

I did an extensive research but could not find a MDIO manual mode solution obtaining  MDC freq. =  1.5 MHz - 2.5 MHz to be covered "step by step"  in the E2E forum. I have contacted a Linux software expert  and I await the response. Please expect that some delay is possible because of the upcoming new year holidays. I will let you know as soon as his answer is available.

Thanks for your understanding !

Kind Regards,

Anastas Yordanov

Hello Til, 

1. What SDK version are you using?

2. Are you currently developing on a custom board or a TI AM62x EVM?

Til Schoengarth said:

What we see is that the frequency is at around 600kHz and cannot be increased - can you hint us on how to increase it to 1MHz for example or 2.5MHz?

3. When you observed the MDIO clock frequency at 600KHz, did you observe that this signal was continuously 600KHz or did it only show that frequency occasionally/sporadically when, for example, data is being written or read via MDIO?

Please give me some time to research/discuss internally if it is possible to change the clock frequency generated from MDIO manual mode/bit bang.

If you do not hear by Tuesday next week, please kindly ping this thread.

-Daolin

When you observed the MDIO clock frequency at 600KHz, did you observe that this signal was continuously 600KHz or did it only show that frequency occasionally/sporadically when, for example, data is being written or read via MDIO?

- Since Til not yet answer, I will answer this because we requested the MDIO Freq to be increased to 1MHz or 2.5MHz. 

- I cannot answer this question because I don't know. We want is to increase the Clock Freq to 1MHz or 2.5MHz for our customer. Pls share with us the step. 

Hello Junaidi,

Junaidi Simon said:

When you observed the MDIO clock frequency at 600KHz, did you observe that this signal was continuously 600KHz or did it only show that frequency occasionally/sporadically when, for example, data is being written or read via MDIO?

Til Schoengarth said:

What we see is that the frequency is at around 600kHz and cannot be increased

Just to clarify, how did you determine that the frequency was at 600KHz? If you measured this frequency, can you clarify if this was a continuous clock signal?

Junaidi Simon said:

- I cannot answer this question because I don't know. We want is to increase the Clock Freq to 1MHz or 2.5MHz for our customer. Pls share with us the step. 

After checking internally, it appears that the MDIO clock frequency is not something that can easily be modified to your expected frequency. Since it is in bit-bang/manual mode, the frequency that is generated can often times be impacted by how many other processes are running on the CPU cores.

-Daolin

Who can share the steps to increase MDIO clock freq to 1MHz or 2.5MHz. 

Hello Junaidi,

can you please share more details as requested above:

Daolin Qiu said:

Just to clarify, how did you determine that the frequency was at 600KHz? If you measured this frequency, can you clarify if this was a continuous clock signal?

And as mentioned by Daolin, it appears that the MDIO clock frequency is not something that can easily be modified to your expected frequency. Since it is in bit-bang/manual mode, the frequency that is generated can often times be impacted by how many other processes are running on the CPU cores.

Daolin Qiu Is there something else that can be done to achieve a higher frequency?

I cannot attach the waveform. but I think it is continuous

Please find the waveform above Daolin Qiu 

Hello Til, Junaidi, 

Thanks for sharing the waveform.

Til Schoengarth said:

Daolin Qiu Is there something else that can be done to achieve a higher frequency?

I can double check with our software developer regarding MDIO bit-bang but the key places in the code that I can think of are 

1. https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/drivers/net/mdio/mdio-bitbang.c?h=ti-linux-6.12.y-cicd

You can try adjusting MDIO_DELAY and/or MDIO_READ_DELAY to see if that could adjust the clock frequency. It's no guarantee that it will, as I explained full control over exactly what the resulting frequency will be is difficult when bitbang mode is used.

Please note the following from https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1295623/am625-changing-mdio-interface-clock-frequency-on-am625-processor? 

This is also in accordance with the 802.3 specification that says: "MDC is an aperiodic signal that has no maximum high or low times."

2. https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/drivers/net/ethernet/ti/davinci_mdio.c?h=ti-linux-6.12.y-cicd#n110 

I believe this section is responsible for the source clock for the MDIO module, not necessarily the actual bit-bang generation of the MDIO clk but it might be something additional to add some debug print statements in the function in the link to print out the value of mdio_in, div, mdio_out_khz, and access_time.

-Daolin