• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Subscribers 34 subscribers
  • Views 757 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.

TLK10232 RX Input sensitivity?

David Musoke
Intellectual 750 points
Other Parts Discussed in Thread: TLK10232, ONET8501PBHi: What is the input RX sensitivity of the TLK10232 for a BER
  • 0
    TI__Mastermind 20820 points

    Hi David,

    I am a little bit confused on your question, could you please elaborate?

    In the data sheet is states that the minimum voltage that can be received by the TLK10232 by the HS and LS receivers is 50mV differential. Does that answer your question or am I off base on this?

    Regards,

    Mike

  • 0 in reply to Michael Peffers
    Intellectual 750 points

    Mike:

    Thanks for answering but the question i'm posing is not to find out what the minimum input differential is, but what is thta voltage for the HS receivers that gives a bit error rate < 10E-12 at 10.3Gb/s for a PRBS31 pattern. All high speed limiting-amps (look at the ONET8501PB limiting amplifier for example) give a voltage spec for this condition.

    The ONET8501PB datasheet says that voltage is 5mV typically (see page 5 under Data Input Sensitivity). I need to know the equivalent voltage spec for the TLK10232 tranceiver. The very definition of input sensitivity is that voltage at the input of the limiting amplifier that gives a certain BER (at a defined data rate and pattern) at its output.

     

    Regards,

    David

  • 0 in reply to David Musoke
    TI__Mastermind 20820 points

    Hi Dave,

    We do not have a spec like that for the TLK devices. The ONET8501PB is a post amplifier and it is not the same beast as a SerDes so there is not a comparable spec. The 50mV differential minimum that I pointed out earlier is sort of like the ONET spec because it implies a BER of <10E-12 even though it may not be stated. 

    In actuality though you can throw a closed eye at the TLK10232 and still achieve a BER of <10E-12 as the device implements a FFE and a DFE on the receive side. These features allow you to take a 10G-KR signal that has a lot of ISI due to backplane traces and/or cables and achieve a BER like you mentioned above.

    Please let me know if you have additional questions and I will be glad to help.

    Mike

  • 0 in reply to Michael Peffers
    Intellectual 750 points

    Mike:

    If the 50mV differential minimum implies a BER of 10E-12 (and i hope it does), then why is LOS asserted at <75mV?

    Are you saying that the TLK10232 could assert LOS but still deliver a BER of 10E-12? Sorry, i'm confused...

     

    David

  • 0 in reply to David Musoke
    TI__Guru 62920 points

    Hi David,

    (I thought I would jump in since Mike is out of the office today.)

    Yes, the LOS assert threshold is higher than the minimum input voltage, so it is possible to have a link up and running with LOS asserted.  This may seem counter-intuitive, but it can be useful to have some indication of a signal problems before errors actually occur - especially since the device already has mechanisms to indicate bit errors at the receiver.

    As Mike pointed out, though, with such a complex receiver it is difficult to have a spec that bounds only one aspect of the input signal.  This spec was based on testing that provided a "clean" input eye (from test equipment) with a low input amplitude.  If there were significant other impairments (such as random jitter, noise, etc.) present simultaneously, the BER could degrade even if the (nominal) signal level was still above this limit.  On the other hand, a low BER can be achieved with a totally closed input eye if the signal impairments are primarily "equalizable" - that is, they result from physical media losses/reflections/etc. that can be compensated by a decision feedback equalizer (DFE).  This capability of the device is encompassed by the jitter tolerance specification, which indicates BER < 1e-12 operation in the presence of channel attenuation, additional random and deterministic jitter, and broadband random noise.

    Best regards,
    Max Robertson
    SerDes/PHY Systems Engineer
    Texas Instruments
    m-robertson@ti.com

  • 0 in reply to Miguel Robertson
    Intellectual 750 points
    Max:
    Thank you so much for your post to my dilemma. I tried to respond in the forum but the format is messed up which makes reading it difficult.
    In summary (please correct me if I’m wrong), the receiver will generate a BER < 1e-12 even in a LOS condition as long as the received signal impairments are equalizable?
    My design is simple as shown below:
    XUAI Host (4 x 3.125G) -> 12” 50 ohm cable -> TLK10232 -> 10G LASER DRIVER -> 10G LASER (TOSA).
    The receive path is even simpler:
    10G ROSA ->TLK10232 -> 50 ohm 12” cable -> XUAI HOST (4 x 3.125G).
     
    My question regards the EQ properties of the TLK10232 with respect to the cable.
     
    1. Does the TLK10233 offer independent EQ over its low speed XUAI interface for both the transmit and receive paths?
     
    2.  Does if offer the same independent TX/RX EQ over the 10G side as well?
    In my application, I need EQ between the host and low speed XUAI interface due to the lossy cable but I also need said EQ on the 10G side as well. The TX to deal with the distortions through multimode fiber via TX Pre-Emphasis.  The RX to minimize BER and improve sensitivity.
    Sorry for the newbie questions and thanks for your time...
  • 0 in reply to David Musoke
    Intellectual 750 points
    Just a couple points of clarification:
     

    ·        The chip can achieve BER < 1e-12 with an “open” input eye of amplitude greater than or equal to our minimum input voltage spec.  In this case, equalization doesn’t really play into it since there isn’t an uneven frequency response to compensate (or else the eye wouldn’t be “open”).  In this case, it is possible for LOS to assert even though the link is operating properly.

    ·        The chip can also achieve BER < 1e-12 with a “closed” eye provided that the eye is closed primarily as a result of something equalizable like electrical high-frequency loss.  In this case, LOS is unlikely to assert.  Most data will have a spectrum that includes low-frequency components as well, and these will not be affected as strongly by media losses.  This means that the eye opening may be 0 V (due to high-frequency losses and intersymbol interference) but the peak-to-peak amplitude could still be fairly large.

     
    I think the first condition I list above probably most closely matches your intended application.  However, I don’t think that our characterization of the minimum voltage corner included much jitter.  I will double-check with the engineer who did the testing to confirm.  Do you have an idea what the “eye” or time-domain waveform coming from your ROSA will look like?  The ROSA-to-TLK link is the only one you described that would concern me, since customers implementing fiber optic interfaces in the past have tended to uses pluggable optics (like SFP+) that contain a limiting amplifier following the ROSA.  Given the TLK10232’s sensitivity I could see how this would not be needed, but I don’t have any experimental data backing that up.  It is possible that the ROSA output is close to our characterization conditions, but if not I think it would be safest to do a little more evaluation to be sure.
     
    To answer your other questions - yes, there is independent EQ on all of the serial inputs (low speed and high speed).  Each serial output (low speed and high speed) has independent de-emphasis settings.
     
    Best regards,
    Max