• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 1 answer
  • Subscribers 33 subscribers
  • Views 499 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.

TDP142: Re-driver at RX side adds jitter in 3m cable 8.1Gbps scenario

Zhiwei Zhang
Prodigy 30 points
Part Number: TDP142
Other Parts Discussed in Thread: SN65DP141

Hi,

I'm building a 8.1Gbps DisplayPort TX+RX system, and 3m cable is used to transmit the signal. 

TDP142 Re-driver is added to both ends of cable, but I found the re-driver at RX side adds jitter to eye diagram, cause limited margin of jitter.

It seems high frequency 0-1-0-1 pattern and low frequency 0-0-1-1 pattern triggers TDP142 pre-emphasis at different time.

Do you have any idea on the part I should look into? Or should I try SN65DP141 as it support UHBR signal rate? Thanks!

Below is the eye diagram I measured:

1) FPGA -> TDP142 @ TX -> 3m cable -> measurement

2) FPGA -> TDP142 @ TX -> 3m cable -> TDP142 @ RX(EQ set to 15) -> measurement

Two config use same setup for TX side, only difference is re-driver @ RX or not.

Thanks!

-Zhiwei

  • 0
    TI__Guru**** 176281 points

    Zhiwei

    What is the difference between the two scope waveforms? 

    Why do you set the DP142 RX EQ to maximum of 15? Can you please dial down the EQ?

    Thanks

    David

  • 0 in reply to David (ASIC) Liu
    Prodigy 30 points

    Hi David,

    As you can see in the first waveform, we don't have much eye height margin due to cable insertion loss. So I'd like compensate it with re-driver. Jitter will become better indeed with lower RX EQ, but loss compensation will be insufficient.

    What confused me is, w/o re-driver(or EQ set to 0), eye diagram has a clear intersection point, but in w/ re-driver scenario, the waveform shifted in time domain. Is this caused by re-driver or something. Is there a theory describing this so I can learn from and better solve this?

    Thanks!

    Zhiwei

  • 0 in reply to Zhiwei Zhang
    TI__Guru**** 176281 points

    Zhiwei

    For the connection between the FGPA and the TDP142, 

    1) FPGA -> TDP142 @ TX -> 3m cable -> measurement

    a. How much is the loss between the FPGA and the TDP142?

    b. How much compensation does the FPGA TX provides?

    c. What is the EQ setting of TDP142?

    Thanks

    David

  • 0 in reply to David (ASIC) Liu
    Prodigy 30 points

    Hi David,

    a. Between FPGA and TDP142, there is only 5cm diff100 trace. We measured eye diagram before it enters TDP142 TX with high-impedance diff probe, which looks perfect.

    b. We don't enable pre-emphasis in FPGA, because if we enable pre-emphasis in both FPGA and TDP142 TX, the waveform sucks. So we only enable TDP142 TX pre-emphasis

    c. EQ of TDP142 TX is set to maximum, 11.

    After testing different cables, we found DP cable performs much better than the original used miniDP cable. We decide to move forward with DP1.4 cable and see how it woks. Let me get back if I found more problems with DP Cable.

    Thanks!

    Zhiwei

  • 0 in reply to Zhiwei Zhang
    TI__Guru**** 176281 points

    Zhiwei

    Thanks for the update. 

    For c. EQ of TDP142 TX is set to maximum, 11, is this the EQ setting of TDP142 that is connected to the FPGA? If this is the case, for 5cm, you may want dial down its EQ setting as it may be over-compensate the 5cm trace loss.

    Thanks

    David

  • 0 in reply to David (ASIC) Liu
    Prodigy 30 points

    Hi David,

    Yes, it's EQ setting that is connected to FPGA. The function of this TX EQ is to pre-compensate cable loss. I suppose there is not difference if I compensate before or after cable, is this correct? I have saw examples of EQ compensation before cable in LVDS application.

    Thanks!

    Zhiwei

  • +1 in reply to Zhiwei Zhang
    TI__Guru**** 176281 points

    Zhiwei

    The BR120 uses its TX Pre-emphasis to pre-compensate the cable loss. If the RX LVDS has RX EQ function, then you may want to dial down the EQ so the total compensation from TX Pre-emphasis + RX EQ do not over-compensate the cable loss.

    The DP142 is a linear re-driver so it does not have a TX Pre-emphasis. If you over-compensate on the input side with its RX EQ, some of the over-compensation will translate to the output as pre-emphasis. But in this case, it is better to dial down the DP142 EQ so it will compensate for the 5cm trace loss only and then let the 2nd DP142 RX EQ to compensate for the cable loss.

    Thanks

    David