SN75LVPE5412: IBIS-AMI Simulation Questions

Hi Kian,

1. I am not aware of such materials. I believe I have seen this kind of information in customer-specific presentations before, but not in the form of publicly available guides.
2. It is hard to say why their simulation is not passing from here. It is possible that their transmitter TX/RX simulation characteristics are different from the model that we use. Also, what criteria is the customer using to define a "pass"?
3. The EQ boost value represents the CTLE functionality and is the primary variable that we tune in simulations to improve signal quality. It tunes the frequency-selective boosts the device applies, and the idea is that higher values can compensate for more channel-induced loss. However it should not be confused with the DC gain of the device, which is another variable that boosts all frequencies equally.
4. Let me check with the team. Next week there will be holidays on Thursday and Friday. If we are able to schedule a Webex, it would probably be earlier in the week.

If your customer can tell us their pass criteria, I can run IBIS-AMI simulations with our models to see if I can get 54 dB link loss to work. We can then prepare a presentation from the results that you can send to them. Let us know if you have any other questions.

Best,

Evan Su

Hi Evan,

Thanks for prompt reply.

Customer said they did change our IBIS-AMI model and eye was closed in pre23dB+post23dB case simulation results. The pass criteria is eye height > 70mv and eye width > 0.3UI. Could you kindly give a simulation report to convince customer this device can pass this criteria? If we can pass, could you also send the IBIS-AMI Model used in your simulation? Thanks.

They also show another questions:

1. How to set different redriver gain in the active simulation?
2. They didn't change the simulation model but change the S spec and set the eq_bst value of redriver and RX as 8 , then pre18+post28 and pre28+post18 case both passed. Could you explain this to them?

BTW, could you kindly help open access to my security of this device?

Thanks!

BRs

Kian

Hi Kian,

The pass criteria eye height > 70mv and eye width > 0.3UI looks not right?

May ask customer which test spec. they are talking about?

I list the PCI-SIG Gen5 test spec. for you reference

Thanks

YT

Hi YT,

I don't know why this criteria looks not right. I think this is reasonable. In PSIE5 protocol, VTXA minimum value is 22mV, so why eye height>70mV looks not right.

If I don't give enough specs, pls kindly let me know and I will align with customer.

Thanks!

BRs

Kian

Hi team,

After discussion with customer, I think they show some questions about the accuracy of this new device -- SN75LVPE5412 IBIS-AMI Model.

• Our pre23db+post23db case simulation results can't pass the criteria eye height > 70mv and eye width > 0.3UI. Ps: they don't change any spec of our model. So could you kindly give a simulation result(w/ specs configuration) of this case to them?
• Numberofbits is set to 10e4, why do we set as this? they usually us 10e6. Ps: Because this new device model don't have a user guide so they show many questions about spec configuration.
• 
• The impact of redriver's front and back link loss distribution on the performance of redriver, is there a trend?
• Can you give a simulation result that can pass height > 70mv and eye width > 0.3UI when the total loss is 54db? So we convince customer this spec is right?

The most important issue of customer is they need post loss>28db@16GHz and the bigger the better(their total loss is 46db@16GHz), so which is the largest post lost the redriver can support? Can we show a simulation result and spec configuration to them?

Thanks!

BRs

Kian

Hi Kian,

I am working on the simulation. In the meantime I will try to answer some of the customer's questions:

• In ADS 2021, you can change the redriver's EQ boost and DC gain by going to the redriver's AMI page and changing the "eq_bst" and "drv_sel_gain" parameters. These are parameters specified by the IBIS-AMI model and they should be tuned to achieve the desired signal quality because the default values apply very little equalization. It sounds like the customer figured this out.
  • "drv_sel_gain" is encoded in binary. As I understand, setting of 5 = 0 dB of DC gain, setting of 7 = +2 dB of DC gain, setting of 3 = -2 dB of DC gain, setting of 1 = -4 dB of DC gain, setting of 0 = -6 dB of DC gain. I will double check this as I continue to run experiments.
• I think what YT thought was unusual about the customer's pass requirements is that > 70 mV and > 0.3 UI is much stricter than the PCIe Gen 5 specifications, which are the basis for our testing. I am working on finding a configuration that satisfies the customer requirements but it is more difficult.

I am not sure what exactly the customer means by pre-loss and post-loss. I initially tested a simulation with 23 dB of trace loss before the device and 23 dB of trace loss after the device, however with our transmitter and receiver models the total loss in the system was 65 dB; at that level, I believe I met the customer's eye width requirement by applying maximum equalization and DC gain, but not their eye height requirement. I am now testing a model with a total pre-device loss of 23 dB (transmitter + trace) and a total post-device loss of 23 dB (trace + receiver), which should be easier. If I succeed with that I will try 54 dB total loss with 28 dB post-loss.

I will get the results to you along with answers to the other questions as soon as possible before the holidays.

Best,

Evan Su

Hi Evan,

Thanks for your reply. Update the customer's simulation progress below:

Using our simulation link and change S Spec, Redriver eq_bst=8 and RX eq_bst=8, S spec is attcahed via link(https://tidrive.itg.ti.com/a/N7Eij4gxIrMlsSK2/9a19f552-9dfc-4dc8-aa95-1f73bb03df42?l). They pass two simulation:

• PRE 28dB+REDRIVER+POST 18dB@16GHz  height 76mv，width1.03UI  

• PRE 18dB+REDRIVER+POST 28dB@16GHz  height 81mv，width1.06UI

And the customer may want bigger post loss than 28dB, 28dB is the minimum required value. So could you give a simulation result about how big the post loss redriver can support?

Thanks

BRs,

Kian

Hi Kian,

Please check your email for my simulation results so far and some other remarks.

Best,

Evan Su