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.

Which way provide bypass mode for DS100RT410

Other Parts Discussed in Thread: DS100RT410, DS100BR410

I actually tried putting the TI Retimer DS100RT410 into Ref_mode 0 mode per the TI datasheet, but I have my doubts as to whether it is bypassing the retiming mechanism.  The output of the retimer does not look like the input – the output signal is significantly cleaned up as if it is being retimed. It is not possible on our board to physically bypass the retimer, then I ask if we are doing the right thing to make the Ref_mode 0 (bypass mode) work? 

Generally the question is: is it possible to bypass DS100RT410 mechanism such way that it'll be transparent for input signal? If "yes", please send me exact procedure we need to do. If "no" is it reasonable to use DS100BR410 pin compatible device instead or any other solution you can recommend us? 

  • Hi,

    The reference mode is not related to the bypassing of the CDR. The reference mode selection changes how the device locks to the various data rates.

    To change the output from retimed to raw data you will need to enable the override for the CDR output mux and program the mux to output raw data. This is discussed on page 27 of the datasheet. You can find the Output mux settings in Table 9.

    For a quick recap:

    1. Set bit 5 of channel register 0x09 to be 1.
    2. Set bits 7:5 of channel register 0x1E to be 0.
    3. Set bit 7 of channel register 0x3F to be 1.

    Mike Wolfe

    DPS APPS / SVA

  • Hi Mike.

    Several different people have tried register writes above, and still get a great looking eye out of the retimer.  attached are eye diagrams showing 10G input vs output with retimer on, and programmed in bypass as you show above.  Note how clean the output looks even in programmed bypass mode??  Shouldn't  the output look very close to the input eye when in bypass mode?

    In the pics below, the Retimer input is shown on left.  The two pics on the right are with Retimer operating normally in retime mode, and with the mux bypass programmed .  I don't see much difference.

    Matt Carlson

    Freescale Networking Division

  • Hi Matt,

    Putting the retimer in bypass mode only bypasses the CDR. The data is still being equalized and any programmed de-emphasis is applied at the driver output. So, if the input data did not exhibit noticeable jitter due to a dirty, noisy clock at the signal source, the retimed result may only look marginally better than the bypass result.

    Thanks,

    Michael

  • thanks very much..that clarifies what we thought might be happening.  So is there a way in this device to perform an internal bypass so that the input pretty much flows directly thru to the output?

    Matt

  • Hi Matt,

    The recipes I listed previously are intended to be implemented in the channel register sets.

    What channel are you using for your testing? What values do you have register 0xFF set to?

    Mike Wolfe

    DPS APPS / SVA

  • I intend on operating on all four channels, so 0xFF is set to 0c.  If I understand it correctly, writes will affect all channels, correct?

    -Matt

  • Hi Matt,

    Sorry for the delay in response. To answer your two questions:

    1. There is not a way to have the retimer input pass through to the output without being equalized and redriven at the least (raw data).

    2. Yes, setting 0xFF to 0x0C will write to all channels.

    To show some matching information to what you are seeing on your end, please reference the images below:

    I have the following test setup:

    PRBS11 @ 10.3125Gbps => 10" stripline trace => DS100RT410 => DCA 86100 Scope.

    Retimer input:

    Retimer output, retimed:

    Retimer output, Raw:

    Notice that there is not much difference between Jitter p-p, because the incoming data jitter is not dirty enough to cause noticeable differences when a CDR is used, even though CDR improves the jitter performance.

    Now, if I used a 40" trace instead of 10", the value of the retimer CDR becomes more noticeable, due to more jitter introduced into the system:

    Retimer Output, Retimed:

    Retimer Output, Raw:

    Michael