• State TI Thinks Resolved
  • Replies 5 replies
  • Answers 2 answers
  • Subscribers 33 subscribers
  • Views 127 views
  • Users 0 members are here
Support feedback
Options
Options
Related

SN75LVPE5412: SN75LVPE5412 setting question

Sean Liao (Yu-Sheng)
Expert 3790 points

Part Number: SN75LVPE5412
Other Parts Discussed in Thread: DS320PR410

Hi sir, 

My customer Avalue is using SN75LVPE5412, here are some questions:

  1. TXAn(P/N) output, is the differential= 85 ohm? (only need to use redriver, no need to use Demux)
  2. In Table 6-2 EQ SET, TYPICAL EQ BOOST & Table 6-3 Flat Gain SET, what's L4 length default (means redriver to PCIe slot length)? 

 

 

 

  • 0
    TI__Intellectual 1046 points

    Hi Sean,

    Please see my comments: 

    1. The output differential impedance of the TX driver is 100 ohms. 

      The customer may choose to route the PCB to 85 ohms following the PCIe CEM spec (72.5 Ω to 97.5 Ω differential). This is OK. Our device can support driving a 85 ohm differential impedance. 
    2. Could you please clarify your question? If you are asking about what L4 setting does in table 6-3 for Flat Gain: L4 is the default and recommended setting which adds minimal DC gain. 
    3. Since you are not using the demux functionality, perhaps you could also look at DS320PR410. DS320PR410 is our 4 channel redriver which has a smaller package compared to SN75LVPE5412.

    Best,

    Charles

  • 0 in reply to Charles Wang
    TI__Expert 3790 points

    Hi Charles, 

    In brief, customer use it in the riser card, and the IC board length is as below, thanks a lot!

    TX0~TX7

    TX8~TX15

    RX0~RX7

    RX8~RX15

  • 0 in reply to Sean Liao (Yu-Sheng)
    TI__Intellectual 1046 points

    Hi Sean,

    Thanking you for sharing the customer's system. What is the insertion loss pre and post redriver? I see that you provided trace lengths. I have estimated worst case loss budget based on 0.5 dB/in insertion loss of traces and 1.5dB loss from connector. Please use customer system and find actual loss values to verify loss budget. 

      CPU 5" trace Conn 0.75" Redriver 3.5" trace Conn Total Loss
    Insertion Loss (dB)   2.5 1.5 0.375   1.75 1.6 7.725

    Based on these estimates, redriver should be able to handle the estimated 7.725dB added insertion loss.

    Best,

    Charles

  • 0 in reply to Charles Wang
    TI__Expert 3790 points

    Hi Charles, 

     

    CPU

    5" trace

    Conn

    0.75"

    Redriver

    3.5" trace

    Conn

    Total Loss

    Insertion Loss (dB)

     

    2.5

    1.5

    0.375

     

    1.75

    1.6

    7.725
    Charles Wang said:

    Based on these estimates, redriver should be able to handle the estimated 7.725dB added insertion loss.

    1. Is the 7.725dB set at the EQ site?

    2. Is the Flat gain L4 set at +0.6db, what does it mean, does it mean the output allow total 16dB cable loss? 

    3. Customer PCB is  NPG-170D  (Low Loss, Megtron 4, DK:3.3 DF:0.007). Is 5” Insertion Loss 2.5 dB too large or it's acceptable?

    Thanks a lot!

  • 0 in reply to Sean Liao (Yu-Sheng)
    TI__Intellectual 1046 points

    Hi Sean,

    Please see my comments:

    Sean Liao (Yu-Sheng) said:
    1. Is the 7.725dB set at the EQ site?

    The 7.725dB is my estimate of total loss of system. The actual EQ index used will need to be tuned based on eye opening at the endpoint. 

    Sean Liao (Yu-Sheng) said:

    2. Is the Flat gain L4 set at +0.6db, what does it mean, does it mean the output allow total 16dB cable loss? 

    DC gain is the gain applied over all frequencies. It is recommended leave at default value of +0.6dB. The gain from the CTLE EQ index will compensate for the 16dB of allowable cable loss, not the flat gain.

    Sean Liao (Yu-Sheng) said:
    3. Customer PCB is  NPG-170D  (Low Loss, Megtron 4, DK:3.3 DF:0.007). Is 5” Insertion Loss 2.5 dB too large or it's acceptable?

    Insertion loss of Megtron 4 is estimated around 1.2dB per inch @ 14GHz. Total insertion loss is ~12.25dB. This should be OK. 

    For better understanding of EQ tuning, you can read snla461. 

    Best,

    Charles