DP83869HM: Minimum frame size

Hi Steven,

Thank you for submitting your query, I'm ready to assist but I don't think I fully understand the question.

Could I ask why a small frame size is desired? How is the frame size being set/controlled?

Regards,

Alvaro

Hi Alvaro,

A small frame size is desired for this application to reduce latency. We are using TI's Sitara AM64x processor to send data from one processor to another. The frame size is controlled using the AM64x (specifically we use the PRU to generate a variable-length Ethernet frame, so we can control exactly how long the frame is to minimize latency). 28 bytes would include an 8 byte preamble, 16 byte payload, and 4 byte FCS.

Hi Steven,

Thank you for providing more details. In this case since the frame is still the same, only variability is the payload, it should be fine. I believe this is supported by the IEEE standard, by adding a PAD if the payload is too short. Have you tried testing this out yet? 

Regards,

Alvaro

Hi Alvaro,

Unfortunately padding increases latency (more bytes transmitted = higher latency), so that doesn't help with this problem.

Also we are aware that other TI PHYs have not strictly complied with IEEE standards. For example DP83867 does not support IPG <12 bytes even though IEEE 802.3 requires IPG down to 8 bytes. I am also aware that other TI PHYs have been compatible with frames smaller than 64 bytes, so hoping DP83869 is similar.

What is the minimum frame size specified for DP83869?

Is the minimum frame size different in 1000BASE-X mode compared to 100BASE-FX or 100BASE-T?

I'll also add that my understanding is that minimum frame size is a legacy requirement from the 1980's when collision detection was required. With modern packet switching, minimum frames aren't necessary.

That said, I'm not sure if there are technical reasons why DP83869 might require a minimum frame size. This is why I'm asking the question. Thanks for your help!

Hi Steven,

I candidly don't know if it can support it, so instead I'll try recreating this experiment in lab for both 100M and 1000M speeds, copper and fiber. I will reply again before end of day 01/26 at the latest.

Regards,

Alvaro

Thanks!

Hi Steven,

I tested this by setting the DP83869EVM into reverse loopback (packets received from the MDI side or looped back through the MAC pins, back to the connector, done by setting Reg 0x16 = 0x00020). The link partner in this case was a SmartBits 2000.

Smart Bits <-- Ethernet Cable --> DP83869EVM ↵ Reverse Loopback

In normal operation I was able to see the number of transmitted packets = to received packets on Smartbits. When adjusting the length of frame, Smartbits would display the length in red when set to <60. At 60, everything still worked, however 59 bytes and below there were no packets received back to smartbits. I am unsure if this is a Smartbits issue or a DP83869 issue. 

Regards,

Alvaro

Hi Alvaro,

I've tested frames down to 32 bytes without issues over a 24 hour period. I need to put my test setup back together before I can test smaller frames.

But the larger question is just because this works on 2 boards in development, should we expect this to work on all DP83869's across the voltage and temperature ranges in the datasheet? Or are there other technical reasons a minimum frame size should be used with DP83869?

Hi Steven,

I don't see why it wouldn't work. Please let me know the results of your test once you bring up the test set up again.

Regards,

Alvaro