• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 2 answers
  • Subscribers 33 subscribers
  • Views 886 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

Original question:

DS100MB203: DS100MB203 : MODE select pin description datasheet

DS100MB203: DS100MB203 : 10GbaseKr link failing

Giridhar Reddy
Prodigy 180 points
Part Number: DS100MB203

Hi,

In one of the queries asked on 10GBaseKr links with DS100MB203, the reply from TI is "The MB203 is designed to compensate for attenuation in long and lossy channels, when the channels are shorter link training algorithms can be confused by over-equalization.".

When you say long and lossy, what is the approximate length and loss the MB203 assumes. And if we have a shorter channel, is there any other settings that we can change in DS100MB203. (we are already setting it to zero in our design and the link with shorter channel is not working and the other link with longer trace is working fine).

  • 0
    TI__Mastermind 43795 points

    Hi Giridhar,

    I have seen changing the output MODE to 10GbE style can help to improve the system response.  Also make sure the output DEM is set to 0dB. Do you have register access?  With the registers you can make the change on individual channels.

    Regards,

    Lee

  • 0 in reply to Lee Sledjeski
    Prodigy 180 points

    Hi Lee,

    We have tried changing the output mode to 10GbE and it did not help. Yes we have access to registers and we have tried varying the DEM values for individual channels. DEM is set to 0dB and still the link has issues.

    The only difference we can see between the working channel and non working channel is their length and hence some loss in them. The shorter one with less loss is not working.

    In the datasheet of DS100MB203, page 16,  table 2 (Equalization settings),

    EQ bits 0x00 settings is suggested for a channel of less than 5 inch and 5.3dB loss is given at 5GHz.

    1) Is this 5.3dB minimum loss  assumed by the device at 5GHz ?

    2) And if so, if we try to make our channel to have 5 dB loss, will the performance improve?  (As of now the loss in that channel as per simulations is very less and we are thinking that over equalization is creating the issue.)

     

  • 0 in reply to Giridhar Reddy
    TI__Mastermind 43795 points

    With additional attenuation between the 10G-KR Tx and the DS100MB203 input, the performance is likely to improve.  If it is possible to force the 10G-KR device into a non-KR mode and set the Tx equalization to 0dB, the link may work as it is currently designed.

    I would recommend to look at the 10G-KR training logs from the working channels to make sure the Rx equalization is finding a consistent range after training is complete.  If this is the case, modifying the short traces to the long trace length is likely to give similar results.

    Regards,

    Lee

  • 0 in reply to Lee Sledjeski
    Prodigy 180 points
    Additional attenuation between 10GKr tx and DS100MB203 will improve the performance. Can you quantify as to how much attenuation is expected approximately at the DS100MB203 input. From the datasheet, 5.3dB @5GHz is mentioned for channels of FR4 less than 5 inch when we set EQ to 0dB. Does that mean we should try to achieve 5dB attenuation at the minimum?
  • 0 in reply to Giridhar Reddy
    TI__Mastermind 43795 points
    For the MB203 a good minimum in a 10G-KR application is about 8dB. This gives the transmitter a little space to provide Tx equalization as well.
    Regards,
    Lee
  • 0 in reply to Lee Sledjeski
    Prodigy 180 points

    Based on the above suggestion, we can adjust the length of RX trace (MB203 is the receiver) to attain the required minimum of 8 dB. 

    Is it suggested to increase the loss on the other side of MB203   as well (i.e TX from MB203)  or we can leave the present traces as such. 

  • 0 in reply to Giridhar Reddy
    TI__Mastermind 43795 points

    I recommend to keep the traces connected to the MB203 output at the same length as they are.  I would not make them longer.

    Regards,

    Lee