DS160PR410: TX test Fail

Hello-

The DS160PR410 is a linear redriver which passes PCIE TX presets for link training.  However, as expected when dealing with any real word device, its linearity is not ideal.  As a TX Preset (signal pre-shoot and / or de-emphasis) passes through the redriver, it is slightly attenuated. 

If your PCIe TX is marginally passing the TX presets, once you add a redriver in the data path, it is possible that the TX presets would be marginally failing the spec.

For the DS160PR410, reducing the VOD improves the linearity. So, for the purpose of passing the TX Presets test, you may use VOD = 0 or 1.  However, for the best system performance and maximum eye opening, VOD = 3 should be used.

Regards,

Davor

If I set the VOD = 3.

I will fail with TX Preshoot Preset #9

Hello-

As mentioned before, the DS160PR410 linearity is not ideal as it is true for any linear redriver.  It slightly attenuates PCIe TX pre-sets (that is, it reduces TX equalization).  However, it also adds significant RX equalization (up to 18 dB), so the benefits of a redriver a more than justified.

With VOD=3, you fail P9 by less than 0.05 dB.  You have the following options:

Option 1: Reduce VOD to 1 or 0.  With reduced VOD, the redriver linearity improves, so TX equalization improves by a fraction of a dB, however, you reduce the signal amplitude by -3.5 dB or - 6 dB, making it difficult to pass other TX/RX equalization tests, plus your overall system noise margin is reduced.  This option is not recommended.

Option 2:  See if there is a way to slightly increase TX Equalization of your PCIe device by 0.1-0.2 dB.

Option 3:  Waive the failure.  The 0.05 dB of TX equalization is insignificant compared to the RX equalization added by the re-driver.

Regards,
Davor

Hello~

About waive failed, you say the 0.05dB value, how to get it, please confirm.

If Preset # 9 fails, what problems will occur (such as PCIe certification, Host Link...etc)

Does TI have experience on Preset # 9 failed and how to get certified.

Is it recommended to fix VOD = 3 and adjust CTLE to increase the current -0.8dB (Index 0) to + 1.3B (Index 1).

If we cannot adjust the device TX, is it recommended to remove the Redrive IC on this path

BR

Edward

Hello Edward,

The P9 pre-shoot spec is 2.5 dB to 4.5 dB.  You reported 2.4581 dB.  This is less than 0.05 dB below the min spec of 2.5 dB.  From the system performance point of view, this is insignificant.

The DS160PR410 passes all PCI-SIG compliance tests, however, for the TX preset tests, we had to select an endpoint that passes the TX preset tests by itself with good margin.  For example, if a PCIe TX has a 2.7 dB P9 preshoot, the AIC would likely fail the test with the redriver in the data path. A PCIe TX with a 3 to 3.5 dB P9 preshoot would be ok with the redriver in the data path.

Changing CTLE setting does not significantly impact redriver linearity.  So, with VOD=3, increasing CTLE will likely not help with the pre-set test.

Regards,

Davor

Hello Davor

please help below

1. If we fix VOD = 1(-3.5dB), this setting was passed the Preset 9 item, and adjust CTLE to increase more and more (the current -0.8dB), this is efficacious for far end eye diagram improve.

2. I would like attach test report for you reference, could you provide the mail address.

3. About the PLL test, I have some question we fix the Re driver CTLE to index 0 and with below setting

      a. PCBA1 set the VOD to L3 Pass and test with P7

      b. PCBA1 re set the VOD to L1 Pass and test with P5

      c. PCBA 2 set the VOD to L0 Pass and test with P3

      d. PCBA 3 set the VOD to L3 Fail can't not be luck wave, try use P0,P4,P7, both fail. 

    Why PLL result will not the same, with the difference VOD setting?

    It always difference result with every time test the PLL?

    How to fix PLL issues?

4. P9 test is necessary for PCIe SIG  compliance tests. We need description the P9 item fail is insignificant. could you provide the more description and detail document.

5. We have put 0.22uF CAP on TI Re driver IC input and output two sides, this will effect PLL and TX test?

BR

Edward

Hello Edward,

1.  Yes, it is possible that low VOD (Index=1) and higher CTLE allow you to pass far end eye diagram test.

2.  Could you please forward the test report to your  TI FAE?  The FAE should be able to forward the report to me for reference.

3.  Are you testing multiple boards or just one board and performing multiple attempts?  It appears to me that there is a setup issue or excessive deterministic jitter.   Can you try increasing  CTLE Index for this test?  If using P4 (0 dB of TX EQ), the CTLE Index (RX EQ) should match the channel loss.  If using P0 (~ 6 dB of TX EQ) or P7 (~9 dB of TX EQ), the TX EQ + CTLE Index (RX EQ) should match the channel loss.

4.  Per item 1 ( VOD=1 and increased CTLE), you pass P9?  For PCI-SIG certification purposes, I'd suggest to focus on passing the P9.

5.  There should be one ac-coupling capacitor between PCIe TX and DS160PR410 RX and one ac-coupling capacitor between DS160PR410 TX and PCIe RX. Extra ac-coupling capacitors may affect PLL testing.

Regards,

Davor

Hello Davor

I correct the test result as below

And based on the results of these days of discussions, next plan as below.

Hello Edward,

During the TX preset testing, the TX preshoot (P9) is going through the redriver U2. Correct?  So, we are reducing VOD on U2 to Index =1 (-3.5 dB) to improve the linearity of the redriver and help pass the P9 pre-shoot test.  The VOD on the redriver U1 should always stay at Index 3 (0 dB).

I'd suggest the following options: 

VOD (U1_TX)= 0 dB

VOD (U2_TX)=-3.5 dB

EQ (U1_RX) = 13.6 dB may or may not be optimal setting.  Try also 10.7 dB and 7.1 dB.

EQ (U2_RX) = -0.8 dB may or may not be optimal setting.  Try also 1.3 dB and 5.7 dB.

If necessary, please arrange on-site help from your local TI FAE.

Regards,

Davor

Hello Davor

Yes, TX preshoot (P9) is going through the redriver U2 (RX path) of AIC.

You can get our schematic design from Taipei TI FAE : Ben Kuo.

DS160PR410 is quite new product, no much experience and related test environment at Taipei after confirmed with local FAE.

It seems local FAE couldn't help to confirm the setting result due to no test equipment here.

So I want to confirm with you again, is it possible to send or bring our DUT to TI USA for issues setting fine-tuning?

But same item need correct

VOD (U1_TX)

VOD (U2_RX)

EQ (U1_TX)

EQ (U2_RX) 

And about you suggestion we update the test plan as below

BR

Edward

Hello Edward,

Your results so far allow us to conclude that we should use the following VOD settings:

VOD (U1_TX)= 0 dB

VOD (U2_TX)=-3.5 dB

As mentioned before, VOD=0 dB is a recommended setting and provides the highest eye opening.  We only adjust it down if we see TX Preset failures.

The remaining items are to see if the  EQ settings could be optimized further.  In general, EQ settings should be optimized based on the channel loss and the expected PCIe TX preset.  For example, if the channel loss between root complex PCIe TX  and redriver RX (U1) is 20 dB, assuming PCIe TX would normally operate with P8 (~7dB TX EQ), the redriver RX EQ should be set to ~13 dB.  This would be a good starting point for fine tuning.  From here, you should sweep the redriver RX EQ up and down to see what gives you the best eye opening for that specific channel.

Similar approach should be done on the channel between the endpoint PCIE TX  and the redriver RX (U2).

Also, let's take the discussion about sending your board to TI USA offline.  Send an email to Ben Kuo (TI FAE) and ask him to loop me in.

Best Regards,

Davor

Hello Davor

Our local FAE contact window is Ben Kuo, and he's mail as below..
benkuo@ti.com

We need confirm our customer, If we can send the samples I will ask Ben

BR

Thanks a lot

Edward

Hi Davor

Will retest below setting at our vender lab on 1/13.

Did your have any suggestion about the setting.

Hello Davor

Thanks for your advice

But we don't have much time to try to reduce the CTLE of U1 from 13.6 to 10.7 and 7.1dB.

U1 CTLE changes will effect the RX test results. No abnormality was found on the RX side for now 13.6dB setting.

So please tell me in detail.

Why U1 is recommended to be changed.

Is it because of the PLL test approve or something else.

Thanks a lot

BR

Edward

Hello Davor
Thank you for your reply
I cannot perform an optimized U1 CTLE test for three reasons
1. My client does not agree to send the test sample to TI for test
2. The laboratory I can find currently costs $ 500 an hour
3. TI Taiwan has not been assisted in verification. I can only cooperate with my suppliers, but the time is limited

So would like to ask
What is the result of the 13.6dB setting, such as the end product will not be easily recognized? Or the PC host will be affected? Or is it related to the PLL and TXEQ test?

Could you description to more and detail? not the say "If time allows, try optimizing the U1 CTLE further"

BR

Edward

Hi Davor

Good day

Could working DS160PR410 used on the PCIe Gen 2 environment?