WEBENCH® Tools/TUSB501: The webench simualtion shows bad jitter performances worse than Datasheet

Hi,

Could you describe the simulation setup that was run? Please see image below for setup and result from Webench I was able to generate. Pay attention tot he highlights in the image below and your test setup. Is TUSB551 not available in webench? There is a publicly available Hspice model for use if necessary.  

http://www.ti.com/product/TUSB551/toolssoftware

Hi Malik,

thanks for your answer.

The data rate is 5Gbps. The channel_1 (S-parameter) are in the zip file I've enclosed in my previous post.

It seems that the TUSB501 is not able to recover a signal passing in that channel.

The TUSB501 in the specific application has to redrive a signal which passes through a channel described by the S parameters (zip file).

regards,

Domenico

Hi Domenico,

Could you describe what your application is? What is the s4p file supposed to represent? The S4P file you provide shows a large insertion loss at 2.5 GHz. Does your USB Host TX have some pre-emphasis that should be included? There are also many ripples in your s4p file which usually indicates there are impedance mismatch in your system.

Hi Malik,

the signal path represented by the SP4 file is the following one:

USB 3.0 Connector of  a commercial PC - cable (1m) - USB type C connector - PCB - Flex cable - Board to Board connector - PCB - Flex cable .

At the end of the path there is a SSD.

The source of the signal is a standard USB 3.0 port of a PC, there isn't any pre-emphasis.

The model (SP4 file) has been generated considering the model of each element in the path.

In my simulation with Webench I was not able to have an open eye at the output of the TUSB501.

If you think that a different device could do the job (TUSB551, TUSB1002A) please let me know.

regards,

Domenico

Hi Domencio,

I recommend taking the s4p file with the SSD in the path, generally only transmission media is included in the s4p file. A large amount of loss should be expected form this setup. If the amount of loss is greater than the equalization at the re-driver receiver  then the eye will not be able to be opened. You may consider changing the placement of TUSB501. Right now the TUSB501 is place at the end of transmission media in the simulation (after last flex cable) and nothing on the post channel.  I was able to obtain an eye opening with around 10.76 dB loss at 2.5 GHz on the pre-channel and 5 dB loss at 2.5 GHz on the post-channel with high reflections. 

Hi Malik,

thanks for the suggestion. Right now the TUSB cannot be placed in onother place in the path.

Do you confirm that the TUSB501 cannot open the eye if the path is modeled according to the S parameters I attached in the previous post ?

regards,

Domenico

Hi Domenico,

With your current model the answer is no however the model does have some issues that can cause the simulation to look worse than it could actually.