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.

DS250DF230: can a 625MHz clock input generate a clean 20Gbps PRBS signal?

Part Number: DS250DF230
Other Parts Discussed in Thread: DS280DF810

Hi,

I have few questions and uncertainties about the DS250DF230 with regards to PRBS generation with differing clock inputs.

I would like to generate a PRBS-31 signal at a data rate of 20Gbps as shown in the image below. I would like to do this with a 625MHz clock emulating an input NRZ data stream of 1.25Gbps. 20 / 1.25 is a 16x division.

A diagram of a cableAI-generated content may be incorrect.

 

Some forum posts suggest that this can be done. Evidence is found in posts 5, 6, and 7 (links and summaries below).

Other forum posts suggest that this cannot be done. Evidence is found in post 4 (link and summary below).

Page 28 of the DS250DF230 datasheet itself shows that data rates of less than 4.9Gbps by-passes the CDR, and so this suggests that a minimum clock speed of 2.5GHz (to emulate a 5Gbps data stream) is needed for accurate PRBS generation.

A screenshot of a data sheetAI-generated content may be incorrect.

Will a 625MHz input clock be able to generate a 20Gbps PRBS signal?

Will this clock be treated as NRZ data with a data rate of 5Gbps so that one clock cycle looks like 00001111… repeating data?

Will jitter be much less if an input clock of 2.5GHz is used to generate a 20Gbps signal so that the data looks like 01… repeating data?

 

Forum post 1

e2e.ti.com/.../faq-ds280df810-why-is-25g-28g-retimer-prbs-generator-not-working

The DS280DF810 is very similar to the DS250DF230 and this FAQ states that a recovered clock from the CDR re-timer is needed for PRBS generation.

 

Forum post 2

e2e.ti.com/.../ds250df230-the-divider-setting

For the divider setting, does the divide by 8 and 16 setting bypasses the CDR? If this is done and the PRBS generator uses a recovered clock from the CDR, does that mean it will not be able to generate a PRBS signal if the divider is set to 8 or 16?

 

Forum post 3

e2e.ti.com/.../ds250df230-how-to-generate-a-prbs-signal

It is mentioned that to generate a 10.3125Gbps PRBS signal, a 10.3125Gbps / 8 = 1.2890625GHz clock is needed at the input. Does this mean that the clock must be fast enough to look like a div by 4 data stream of 2.578125Gbps?

 

Forum post 4

e2e.ti.com/.../ds250df230-ds250df230

On this post, it looks like a 322.265625MHz clock input cannot produce a PRBS signal output at 10.3125Gbps. /8 and /16 both bypass the CDR due to high sub-divisions and so no PRBS can be generated?

 

Forum post 5

e2e.ti.com/.../ds250df230-use-retimer-to-generate-and-check-prbs-signals

A TI engineer on this post states that a 25.78125Gbps PRBS signal can be generated if the input has an 805.66402MHz clock signal (equivalent to a 1.611328Gbps data stream). This is divided down by 16x.

 

Forum post 6

e2e.ti.com/.../ds250df230evm-prbs-configuration-at-6-375gbps

This post shows that a 6.375Gbps PRBS signal can be generated from a 398.4375MHz clock input. This is equivalent to a 796.875Mbps data stream. This is a division of 8.

 

Forum post 7

e2e.ti.com/.../ds250df230evm-a-problem-of-prbs-testing-with-ds250df230evm

This post answered by a TI engineer states that a PRBS signal can be generated down to a clock of 25.78125GHz / 64. Does this mean the minimum input data rate is 805.664Mbps (corresponding to a clock speed of 402.832MHz)? If so, this is a 32x division.

Regards,
Liang

  • Hi Liang,

    The "Forum post 7"  that you referenced is the most relevant post for your setup. Based on the testing done previously with the DS250DF230,  we recommend targeting 20GHz/32 = 625 MHz or higher for the input clock. Therefore, your setup should work. Regarding the next questions, the data is treated by the retimer as the data rate between 19.6 and 25.8 Gbps corresponding to the sub-rate being used. Since the retimer is configured to 20 Gbps, it will be 00000000000000001111111111111111 repeating data. 

    Best regards,

    Greg

  • Hi Gregory,

    Thank you for your reply.

    Do you happen to have characterisation data on input dummy clock speed versus rms jitter of the generated PRBS signal?

    If you do, is this information shareable with the public?

    Regards,

    Liang

  • Hi Liang,

    We do not have this characterization data available. For device performance, we would recommend at least a 625 MHz clock to generate a 20 Gbps PRBS signal.

    Best regards,

    Greg