TUSB4041I: Full-Speed Downstream Signal Quality Testing: Eye Diagram and EOP Failure

Did you daisy chain couple of hubs?

Best

Brian

Hi Brian,

yes, that's the case. I have 2 Hub chips daisy-chained on the pcb. Plus the daisy chain of several external hubs according to the official full-speed signal quality measurement (i reduced this number so that i have the maximum of seven layers in my usb tree...).

Is that a problem?

Best Regards,

Gregor

Hi Brian,

i made a quick test and removed the daisy-chain on my pcb. The eye diagram improved (no violations anymore, but quite close), but the EOP failure is still present. Do you have any idea what could be the root cause for the EOP failure?

Please see the pictures below.

Thanks and best regards,

Gregor

ok, let me check in our lab.

Best

Brian

Hi Brian,

thanks for the help so far. Have you found any information on how to influence this EOP timing yet?

Thanks and best regards,

Gregor

Hello Gregor:

    since I'm out of office, my colleague Nicholaus will help.

Best

Brian

Hi Gregor,

Brian is out of office, and so I will be performing the test.  I have the compliance setup ready in the lab I will report back to you as soon as it's done.

Regards,

Nicholaus

Hi Gregor,

I finished the upstream test with all passing results.  Moving to downstream testing.

Regards,

Nicholaus

Hi Gregor,

The downstream test also passed everything.  EOP was measured as 162.42ns.  What test fixture, cables, and oscilloscope are you using?

Regards,

Nicholaus

As you can see in my test the EOP was measured 154.22ns. I used the Test fixture from LeCroy, and used 5m USB cables where needed as described. Do you know what could influence this? 

Hi Gregor,

EOP width is shortened by a delay of the J and K states immediately before the EOP.  This delay happens at each hub, so it is a function of the hubs being used.  You can try swapping the hubs used.

One thing that I noticed is you are using all 7 possible hubs, but that would be including 2 possible root hubs.  You are using only 5 non-root hubs correct?

Try just using 1 non-root hub (The TUSB4040I) and see what the results show.  Then we can compare results directly.

Regards,

Nicholaus

Hi Nicholaus,

I made the tests as you described:

- When i use no extra Hub, so only root hub and DUT, it's ok. EOP = 166.71ns

- When i insert 1x Full-Speed Hub, it's still ok, but EOP already shorter. EOP = 160.58ns

- When i insert 1x Full-Speed Hub + 1x High-Speed Hub, it's not ok. EOP = 158.16ns

So it seams that EOP gets shorter with more hubs.

The problem now is that, according to the test description i have, you have to insert the maximum number of Hubs (1x Full-Speed Hub plus several High-Speed Hubs) for this test. Please see attached the picture of the setup description that i have gotten together with the LeCroy Test Fixture etc.

So if i do the test according to this description (which seems to be according to the official USB spec), the test fails.

Did you do your test according to this setup?

If not, why?

Thanks and best regards,

Gregor

Hi Gregor,

Thank you for this test; breaking down the EOP timing for each hub is very helpful. Two bit periods (166.6ns) is an ideal SE0 width for the EOP of a hub. 

Your EOP width drops by ~6ns after the first hub, which is significant as it must stay between 160ns and 175ns. Relevant information is found in section 7.1.13.2.1 of the USB 2.0 specification.

In regard to the test methods, yes, the LeCroy document you sent showing 5 hubs is correct for official FS compliance. My setup includes all of the same elements, but the number of hubs may be reduced for debugging purposes, as you have done.

Here are the numbers I had for each test.

1 DUT - 166.35 ns

2 HUB - 162.37 ns

3 HUB - 160.58 ns

4 HUB - 160.42 ns

5 HUB - 160.36 ns

My advice is to change your hubs, especially the ones that add a long increase in EOP width.  

Regards,

Nicholaus

Hi Nicholaus,

thanks again for checking. I did some further tests, too:

I tried to change some of the hubs, but i keep getting the same results. I only have two different full-speed hubs available and both show the same increase in EOP of around 6ns, where as high-speed hubs show much less increase.

So i can build up the required setup with all external hubs being high-speed hubs and also pass the test. But as soon as i insert one full-speed hub (as described in the specs) i get an increase of 6ns and with 2 or more hubs in total, the test fails.

So here are the questions then:

- did you also use a full-speed hub in your setup?

- can you tell me the exact model of the full-speed hub you use (and maybe also the model of the high-speed hubs while we're at it)? I intend to buy the exact same one and re-try.

Thanks and best regards,

Gregor

Hi Gregor,

Yes, I used a full-speed hub. 

All the high-speed hubs are TUSB804x or TUSB802x, the full-speed hub is the Belkin FSU101.  Belkin ExpressBus (F5U101) 4-Ports External Hub for sale online | eBay

If it helps raise your confidence in the TUSB8044, our group has members in the USB-IF organization and has helped defined USB specs, we perform USB compliance tests for all of our devices before release, and so it is rare for them to truly fail compliance.  I took a look at the testing for the TUSB8044; during its release it passed HS compliance across different ports, units, multiple voltages, and temperatures (300 passing HS compliance tests).  It also passed FS compliance 75 times.

If further assistance is needed, then you may want to reach out to a 3rd party test vendor.

Regards,

Nicholaus

Hi Nicholas,

thanks for the info, unfortunately this Belkin USB Hub is not available here in Switzerland or anywhere i can easily order.

I'm also not doubting the performance of the TUS8044 chip, it's pretty obviously related to the setup (and the High-Speed tests were all pass). I was just checking my own pcb design and i want to make absolutely sure that it's ok. Plus i wanted to understand why my tests failed.

But now I don't really know how to proceed since it's quite difficult to get any full-speed-only Hubs on the market :-). I wonder what you should do / buy if you want to build up a new testing setup...

Anyway, i guess i'll have to find some.

I'll write you back if i have further questions, until then, thanks a lot for your help.

Best regards,

Gregor

Hi Gregor,

Yes, it is a difficult issue and hard to find full-speed hubs.  Good luck on your testing.  I will close the ticket for now and you can reopen it if needed.

Thanks,

Nicholaus

Hi Nicholaus,

the following thing occured to me:

In the description of the setup, you can see that the first hub after the Host should be the full-speed hub. In my setup however, the host is a laptop, which has already a high-speed capable hub inside (after the root hub), the ports of the actual root hub are not accessible from the outside of the laptop. So the order of the hubs is different:

host (root) -> hs hub -> fs hub -> hs hub -> hs hub -> DUT

Do you think this can be a problem?

But i guess this too is the same way in your setup, right? Or do you have access to the root hub ports of the PC?

Thanks again and best regards,

Gregor

Hi Gregor,

Unfortunately, I am sick today so I cannot go into the lab to check our setup.  I believe what we did was plug in a PCIe USB host controller into the PC and thus are given access to a root hub.  I would go into the lab on Monday and double-check.

Regards,

Nicholaus

Hi Nicholaus,

in the meantime i made the following test:

With a W&T USB Server i created USB Ports via Ethernet. This also creates a new Root Hub where the ports are accessible. But the measurement of EOP was exactly the same. So i guess this is also not the root cause of the problem...

Best regards,

Gregor

Hi Gregor,

I see.  I was able to confirm that we do add an external xHCL Compliant host controller to our test system, which allows access to a root hub.

There is one interesting option I see in the official USBET20 v2.2 compliance software.  It allows you to select the number of hubs in your test.

I have never used this option and always followed the "5 hub" rule, but I have e-mailed USB-IF to see why this option is available.

Regards,

Nicholaus

Hi Nicholaus,

to be honest, i don't know what that setting with the number of hubs is, either. In the LeCroy Testing Software it's called "Tier". I played around with it but it didn't have any impact on this EOP measurement. But i would wonder what it means if you get an answer from USB-IF.

It's really strange, in the mean time i've tested 3 different Full-Speed Hubs but i get the same EOP timing results with all of them. Also (as described) with the Ethernet USB bridge where i have access to the ports of the root hub. It's just always a bit too short no matter what i do...

Best regards,

Gregor

Hi Gregor,

Do you see any difference with a change in host?  

Regards,

Nicholaus

you mean with the change to the other (external) root hub? No, there's no change in EOP timinig measurement with that.

Hi Gregor,

I see.  I'm also stumped.  Is there any other way I can assist with the TUSB4040I?

Regards,

Nicholaus