DS125BR820: AMI model

Hi Aaron,

What is the amplitude of the signal coming from the ASIC TX? What VOD setting did the customer use? And what was the reason for connecting P and N with a 100 ohm resistor?

I ran the model in ADS. I see the amplitude from the ASIC TX is ~1 Vpp (+/-500 mV) and the VOD setting is 6. Across all 5 BST values, the amplitude was in the range of +/- 0.4 - 0.5 V.

Best,

Lucas

The amplitude from the source is +/-0.5V, so ~1Vpp. 

I placed the 100 ohm resistor in response to prior traffic with TI in the past about this topic. But at that time the Mentor tool didn't have a means to actually simulate "through" the redriver so I wasn't able to really make use of the model. I am really only able to get a good eye when using this with a generic model from Mentor instead of the IC vendors model. 

I have left the VOD at the default, suggested value of 6 in all cases. I am confused though by your statement "Across all 5 BST values". The DS125BR820 only supports BST = 0, 1, 2, 3. So there are only 4 levels, not 5. 

Hi Scott,

Scott Gerhold said:

I am confused though by your statement "Across all 5 BST values".

Sorry that was a typo. I meant "Across all 4 BST values."

I'm looking into this and will get back to you by EOD tomorrow.

Best,

Lucas

Hi Scott,

I have some questions about your setup.

1. Where specifically are you measuring waveform amplitude? Is it after the ReDriver_AMI1 block, after the Repeater_Tx_PKG block, or somewhere else?
2. Are you looking at an eye diagram or specifically measuring only amplitude of the waveform?
3. What data rate and test pattern are you using for simulations?
4. Where specifically are you placing the 100 ohm resistor connecting P and N traces? Is it right before the Repeater_Rx_PKG1 block, or somewhere else?
5. Can you try removing the 100 ohm resistor and see if this changes the amplitude at all?
6. Can you share schematic and supporting files so I can try using your schematic with Hyperlynx? If this isn't possible, can you share an image of your schematic?

Note that at TI, we typically use Keysight ADS with our IBIS-AMI models. While I can get access to Hyperlynx, it may take me longer learn how to use this application and I can't guarantee I'll be able to find a solution as I'm not familiar with Hyperlynx.

Also I wanted to clarify, while it is expected that there will be some differences in amplitude between EQ BST settings, this is a side effect and not the intended purpose of EQ BST. This setting is used to boost the high frequencies of the signal to open an input eye, and amplitude can slightly vary due to over-equalization. The main parameter used to control output amplitude is VOD.

Best,

Lucas

1. I think you may be identifying part of the problem. With Hyperlynx using the Linesim, I have a model. It only allows me to measure the signals at the input pins to the model I believe. I will check with the FAE for the tool though. With Linesim you can't connect the output of the RX block directly to the input of the TX block, you need a special method the links the output of the RX block (which controls BST) to the TX block (which controls VOD). 

2. My question was really just addressing the amplitude of the waveform in general. The eye diagram changes, but not necessarily in a manner that is proportional to BST setting. I understand the gain of the DS125BR820 increases at higher frequencies, but even at 1.5GHz it varies from 2.1 to 6.8dB as BST increases from 0 to 3. 

3. I am simulating at PCIe Gen 3 (8Gbps) rates. The Stimulus is currently set to  NRZ, 128b130b with 1million symbols. I have it set to Time-domain analysis (not statistical). I am not including any crosstalk. 

4. The 100 ohm is immediately before the repeater block. The Linesim schematic is simply a TX model, 5 s-parameter models representing the plug-in cards and the backplane traces and connectors then the repeater block. 

5. I will test it both with and without the resistor.

6. I can share them if you have a location I can send them to. I don't really want to attach them directly to this ticket. I can also provide the instructions on how I am using Hyperlynx, but won't guarantee you that is the correct method.

Hi Scott,

Thank you for answering my questions and providing more information about your setup. You can share your simulation files with me via E2E private message. It would be much appreciated if you can provide instructions on how you are using Hyperlynx. I will look into this more once you have shared your files.

For your reference, this is what my schematic looks like in ADS. I am using the example project included in the IBIS-AMI zip folder.

Best,

Lucas

Hi, I have backed up a little bit. Decided to try ADS as it has direct support for the redriver. Also realized that there were both terminator and package s-parameters in your schematic. So hold for the moment. I will get back to you on Monday.

Hi Scott,

Sounds good, let me know how the ADS simulation goes.

Best,

Lucas

Hi Scott,

I wanted to check in, did you have better success using ADS with the IBIS-AMI model?

Best,

Lucas