• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 33 subscribers
  • Views 447 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.

TLK10034 HS routing on EVM

Uzi Ben-Zon
Prodigy 110 points
Other Parts Discussed in Thread: TLK10034

Hi,

I noticed that HS signals on the EVM board are routed to the SFP+ with relatively long traces while it seems that it could be done with a shorter traces. Would appreciate your explanation. Also, is it mandatory to use Rogers material rather than FR-4 like in the EVM board?

Thanks,

Uzi

 

  • 0
    TI__Mastermind 20820 points

    Hi Uzi,

    The reason that the EVM is routed with the longer traces on the high speed side is because we needed to introduce some ISI for the TLK10034 to avoid over equalization. The effects of the ISI aid the CDR in settling on the center of the eye. If shorter traces are desired in your design you can enable a function on the TLK10034 called HS_ENTRACK which forces the incoming signal through a lossy filter that will add some ISI internally.

    We used Rogers material on the EVM so to showcase the performance of the TLK10034 but FR4 is definitely a suitable end material.

    Regards,

    Mike

  • 0 in reply to Michael Peffers
    Prodigy 110 points

    Thanks Mike,

    Does TI have any recommendations for minimum and maximum HS side traces length on FR-4 material?

    Uzi

  • 0 in reply to Uzi Ben-Zon
    TI__Mastermind 20820 points

    Hi Uzi,

    Based on the EVM design and a scaling factor that I calculated between the dielectric constant of FR4 and the dielectric constant of Rogers (which both vary depending on the grade of material you buy) I came up with a length of 3.529 cm (1.389 in). I based these numbers purely on the EVM design because we have never collected any data for this device using FR4. The TLK10034 has many different modes of operation and these RX and TX lengths can vary depending on the mode of operation the IC is in. TI does offer an HSPICE model for the TLK10034 that you could use to simulate your design before spinning a board. You can find it here: http://www.ti.com/product/tlk10034.

    In your application will you be driving into a optical module or is will you be communicating through a backplane to another device? Are you intending on using the Link Training feature of the TLK10034?

    Regards,

    Mike

  • 0 in reply to Michael Peffers
    Prodigy 110 points

    Thanks Mike.

     

    Yes, I'll be driving into SFP+ optical module.

    Is the number you came up with (1.389") is the minimum or maximum length?

    I can simulate the design for minimum ISI, but you mentioned that in the EVM the ISI has been spoiled intentionally so I'm not sure that minimum ISI is the best situation for the receiver.

  • 0 in reply to Uzi Ben-Zon
    TI__Mastermind 20820 points

    Hi Uzi,

    I do not want to call anything a minimum because the TLK10034 has many different modes of operation so these things can vary. The number that I provided was simply based off the EVM design, we know this works on Rogers and it has been tested. I did a correlation between FR4 and Rogers and transposed that number from the length of the HSTXAN/P because that trace is the shortest going to the optical module. The RX trace is ~5.13cm and it induces ~1.4db of loss from my simulations results.

    Because of the type of CDR technology that is implemented in the TLK10034, ISI helps the CDR find the edges of the eye pattern so that it can readjust itself to the middle of the eye. So, planning to have some loss in your system similar to the EVM is good place to be.

    Regards,

    Mike