• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 33 subscribers
  • Views 645 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.

Application note on measuring performance of TLK3131

Joey Grasty
Expert 4490 points
Other Parts Discussed in Thread: TLK3131

Hello, team:

Do we have an application note or other documentation on measuring the performance of the TLK3131 in a customer's system?

I plan to start looking at the eye diagram associated with the TLK3131 part and I am wondering if TI has any app notes or similar material about how to make proper eye opening measurements.

One of the questions I have is do I trigger the scope on the reference clock (much slower than the actual serdes signal) or on the serdes signal itself?

I will talk to some others around here about it, but I am a little new to this type of measurement.

Our design has some serdes signals that are short in length and some that are much longer.  I am trying to determine if we can set each of the serdes up with the same parameters  (transmit amplitude, pre-emphasis, equalization) or if I need unique settings for each part based on routing characteristics.

  • 0
    TI__Intellectual 1810 points

    Sorry, we don't have an application note with details on TLK3131 performance measurement in a customer's system.

    The best way of measuring the TLK3131 serial eye-pattern is through a scope that is equipped with a clock-recovery module. Here at TI we use the Agilent 86100C Infiniium DCAj wideband scope equipped with the Agilent 83496A Clock Recovery Module and Agilent 86112A 20GHz dual channel electrical plug-in module.

    The second best way is to use a scope trigger with the same frequency as the serial data rate (e.g. 2.4576GHz trigger for a 2.4576Gbps serial rate). The trigger has to be synchronous to the reference clock.

    The third best is to trigger the scope with the replica of the reference clock, phase aligned to the reference clock. (e.g. 122.88MHz trigger for a 2.4576Gbps serial rate).

    Indeed it is possible to have one set of TX and RX SerDes settings that are suitable for a wide range of interconnect lengths. If the longest link is not long enough to fully close the received eye, indeed looking at the received eye-pattern can provide guidance on the suitable settings. However, if the link causes the received eye to fully close, then the best way to determine appropriate settings is to use the built-in pattern generator/checker. That option is even more appropriate if the tests are to be performed on TX and RX SerDes devices mounted on boards that are not well connectorized for connection to a scope and a signal generator. The pattern generator/checker can be activated through the MDIO interface.

    Please let us know if further information is needed.

    Best regards.

    Hassan.

  • 0 in reply to HA
    TI__Expert 4490 points

     

    Thanks very much for the feedback.  I really appreciate it.
     
    Due to our limitations I am pretty much left with the third option for testing the serdes eye pattern.  Given our fairly benign environment I think I will be OK.
     
    I do have a follow-up question or two.
     
    Our maximum serdes distance is around 40 cm of FR-4 PCB.  I am looking for a procedure to test this link and determine if pre-emphasis or equalization or both are beneficial in improving the quality of the eye pattern.
     
    In the other direction I am trying to figure out if I can benefit by adjusting the transmitter amplitude in the TLK3131.
     
    My approach right now is to try different settings and see what looks best on the scope.  I am hoping there is a better way.
     
    Let me know your thoughts, thanks!
  • 0 in reply to Joey Grasty
    TI__Intellectual 1810 points

    You're welcome!

    Yes, you will be OK with the third option especially if your reference clock source is clean (has little jitter).

    40cm of FR-4 PCB is easily handled by the TLK3131. Even if the shortest link is less than 5cm, you can have the entire length range covered by just one set of TX settings for just swing and de-emphasis. The receiver equalization is adaptive so the default RX settings may just be fine with you for the entire length range. As a matter of fact, you may even not need any TX de-emphasis for your length range - the RX equalizer is pretty powerful.You should certainly try to run the SerDes TX with various link length and look at the received signal on the scope to verify how the SerDes performs. You should then pick the TX settings that gives best eye opening.

    Please let me know if you need further info.

    Best regards.

    Hassan.