• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 33 subscribers
  • Views 240 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.

DS280DF810: Settings for 25.8 Gbs

Ciprian Voiculescu
Prodigy 20 points
Part Number: DS280DF810

Hi E2E Teams,

Maybe you can support me on this topic, my knowledge is based on reading and studying and I have limited hands on experience.

At this moment I am trying to find the best settings for DS280DF810 for 25.8 Gbs Ethernet connection.

Between the processor and the Retimer the atentuation is about 3.5dB at 12.9 Ghz. The same configuration is on the second board (the boards are identical). The boards are connected  with a cable that has a atentuation of about 23dB at 12.9GHz. The cable connectors are QSFP+28.

Until now my results are:

Param(mode=2, maincur=22, precur=0, postcur=0, speed=100)
channel 0 - status: 0x30, heo: 0x0E, 0.438 UI, veo: 0x64, 312.5 mV
channel 1 - status: 0x30, heo: 0x10, 0.500 UI, veo: 0x5A, 281.2 mV
channel 2 - status: 0x30, heo: 0x10, 0.500 UI, veo: 0x5C, 287.5 mV
channel 3 - status: 0x30, heo: 0x0D, 0.406 UI, veo: 0x4E, 243.8 mV
channel 4 - status: 0x30, heo: 0x0F, 0.469 UI, veo: 0x5A, 281.2 mV
channel 5 - status: 0x30, heo: 0x0A, 0.312 UI, veo: 0x3E, 193.8 mV
channel 6 - status: 0x30, heo: 0x10, 0.500 UI, veo: 0x5A, 281.2 mV
channel 7 - status: 0x30, heo: 0x0D, 0.406 UI, veo: 0x3A, 181.2 mV

Even channels are TX, odd are Rx

Are these good values for HEO and VEO?

What can I do to improve the results? In which document can I find more info about the eye mask parameters for 100Gbs Ethernet?

Thanks in advance.

  • 0
    TI__Mastermind 35903 points

    Hi Ciprian,

    Can you please review the block diagram below to confirm my understanding of your system is accurate?

    On our 25G retimers, we typically observe good BER performance for HEO/VEO of at least 0.4UI, 200mV.  Most of the eye openings you shared look good, but channel 5 and 7 seem less than ideal.

    How are you configuring the retimer RX?  Which adapt mode are you using?

    You can also try adjusting retimer TX EQ on egress channels.  This may help improve the eye.  In general, we recommend conducting some greenbox testing to optimize the link.  This is essentially a method of trying different TX EQ values and measuring BER to determine the best EQ value.  Please see the following for more information:

    https://www.ti.com/lit/an/slyt678/slyt678.pdf

    In general, BER testing is one of the best indicators to whether your system is working well.  For 25G, you want to target BER of 1E-12 or better.  Have you tried any BER testing?  You can enable PRBS generator and checker on the retimer to help facilitate this.  Additionally, if your processor has PRBS generator/checker capability, this can help evaluate link quality from processor to processor.

    Thanks,

    Drew

  • 0 in reply to Drew Miller1
    Prodigy 20 points

    Hi Drew,

    Thanks for your feedback and the shared info.

    Yes, the block diagram it's accurate.

    These are some of the settings I am using. Yesterday I added also the DFE override to my script and I noticed some slight improvement.

            #Set Adapt Mode 2
            read_modify_write(bus, ds280_addr, 0x31, 0x40, 0x60)
            # DFE override
            read_modify_write(bus, ds280_addr, 0x23, 0x40, 0x40)
            # Enable the DFE and enable DFE taps 3-5
            read_modify_write(bus, ds280_addr, 0x1E, 0x02, 0x0A)

    Unfortunately I do not have acces to the SMBus in order to use SigCon or other tool. I am running everything in linux on a virtual machine with a python script to see all results and I am using other interface. I am still digging to find out how to perform some BER on my actual setup. And  I have read the documentation about green box test from day 1, very helpfull to start with.. Maybe I should implement this BER test also on a python script.

    So, basically if the HEO/VEO are of at least 0.4UI, 200mV the BER rate it's within limits. What about link stability, these HEO/VEO values will ensure it also? I need a stable link also to perform other kind of tests.

    Thanks,

    Ciprian

  • 0 in reply to Ciprian Voiculescu
    TI__Mastermind 35903 points

    Hi Ciprian,

    Thanks for the details.  If you don't already have access to the DS280DF810 Programming Guide, you can request access to this using the link below.  This may help you in performing BER measurements.

    https://www.ti.com/drr/opn/DS280DF810-DESIGN

    In your application, there is very low loss between the retimer and processor.  Because of this, I would consider manually setting CTLE to 0 on channels receiving from the processor (egress signal chain).  The retimer has around 7dB of boost at the lowest CTLE value, so there is probably not a need for additional signal conditioning.

    For channels receiving from the QSFP port (ingress), adapt mode 2 is well suited for this application.

    Empirically we have observed that provided at least 0.4UI and 200mV, a stable link and good BER can be achieved.  With that said, I would recommend conducting your own testing to ensure that you see good performance in your system.  Often this is done by transmitting PRBS31 signal and measuring BER.

    Let us know if you have additional questions regarding retimer configuration.

    Thanks,

    Drew