TPS65983B: Two port TB3 device doesn't want to do PD negotiation on second port.

Lyubomir Tomov

Part Number: TPS65983B

Hello,

I am doing a Thudnderbolt 3, two - port design, Tapex Creek based.

The port controllers are tps65983b - one for each port, they are interconnected with i2c.
The Thunderbolt controller is Titan ridge.
The design is self- powered and we are fixed source 5V/3A on both ports.
The issue is that when chaining devices we don't do Power Negotioation with a specific display. (Display port over type C).
However with other displays we tested the chaining and it works fine.
With PD protocol analyzer we see that in the non - working scenario we don't respond to the display at all.

The behaviour of both our ports is the same.
The problematic scenario is:

1. I connect our device to the host - and it gets enumerated, PD, thunderbolt all is fine. At this point our device acts as power source to the host.

2. I connect the display to our device - the display advertises itself as a source periodically but we don't respond.

The working scenario is:

1. I connect the display to our device - power negotiation takes place. We become the source.

2.I connect the device to the host - both display and device get enumerated by the host and start working. Again we are source for the host.

So the question is why doesn't the port controller doesn't even try to negotiate the PD with the display?

Thanks,

Lyubo

over 3 years ago

0 Jacob Ontiveros over 3 years ago

TI__Expert 6280 points

Hi Lyubo,

Could you please send the PD logs that you captured to determine this? We need to determine where the PD communication is failing,

I have attached the Tapex Creek template for reference that only sources 5V. Please try that just in case there are some mistakes in your project file.

Tapex_Creek_TPS65983B_5V_PA_PB.pjt

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Hello Jacob,

The project file settings are similar to what we've got. Here is the PD log. The test case is:

1.Dut - Host connected
2.Display - DUT connected
it looks like conneciton is restarting
3.plug cycle the host <-> DUT connection.
4. Display is working.

The .csv file is exported from the ccgx3 file. The PD analyzer is CY4500 EZ PD protocol analyzer utility from Cypress.

log3.zip

Thanks

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

H Lyubo,

I was unable to open your PD trace, from the xls it looks like the connection is cycling. Could you capture VBUS and the CC lines?

Does your design work with other systems?

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Hi Jacob,

Many thanks for you fast replies.

Yes, the connection looks like cycling without any messages going out from the DUT side.

Yes, the DUT works with other platforms, with another usb-c dipslay, and with a Thunderbolt display.

Unfortunately I can't reproduce the failing setup in-house which makes the debugging a lot more difficult.

I am uploading a few snapshots from the PD trace.

1. Colour legend for the signals.

2. The graphics is at line 1 - This is before cable plug I think.

3. The graphics is for line 2 to 22 - This is the bus cycling. This goes like this until I the host connection on the other port is unplugged and plugged again.

4. The graphics is for line 543 to line 563 - Here the host is disconnected and connected again. And power negotiation starts and the system starts working in the end.

PD_snapshots.zip

For the PD trace file I am using :

Cypress Semiconductor Corporation
CY4500 EZ-PD Protocol Analyzer Utility
Version : 3.1.0
Build : 31
Copyright © Cypress Semiconductor 2020

Lyubo

0 Hari Patel1 over 3 years ago in reply to Lyubomir Tomov

TI__Genius 12700 points

Hi Lyubo,

Jacob is currently out of office and will provide you with feedback when he returns.

Thank you,

Hari

0 Lyubomir Tomov over 3 years ago in reply to Hari Patel1

Prodigy 65 points

Hi Hari,

Is Jacob returning any time soon? I would appreciate some help.

Thank you,

Lyubo

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

Hi Lyubo,

I am back today. It's a little difficult to see what's happening from the captures. I see that we have a PD contract but I'm wondering if we are seeing issues during a Power Role Swap. From the VBUS voltage it looks like we are attempting this. Can you please send the log and the download link to the tool to confirm my thoughts?

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Where do you see the PD contract happening? On which capture?

The link for the software is:

https://www.cypress.com/documentation/development-kitsboards/cy4500-ez-pd-protocol-analyzer

4578.log3.zip

And this is the file you should open in the cypress tool. The description is in my second post.

Thank you, and tell me if you need anything else.

Lyubo

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

Hi Lyubo,

Sorry for the delay, I was having a difficult time getting the latest Cypress tool.

From looking at the log, we do end up in DP alternate mode and the HPD messages do go through. There is a bunch of VBUS toggling during the initial connection, which could be caused by the monitor.

The next thing we would have to verify would be the I2C communication between the TBT controller and the PD controller to make sure all of the data is reported correctly.

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Hi Jacob,

I need to emphasize that if you just plug the Host on port A and then the Display on port B, you stay with vbus toggling forever.

And no PD messages pass. How can I determine whether the cause is in our side or in the display?

Thanks,

Lyubo

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

Hi Lyubo,

From the PD log you attached the connection does resolve into Display Port and the Attention message for HPD is sent from the monitor. From this we know that the Display Port connection is successful from the PD communication perspective. The next step is to verify the I2C communication.

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Hi Jacob,

What needs to be verified on the I2C? Am I right that before I reconnect the host connection the DUT doesn't respond at all to the Display?

Lyubo

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

Hi Lyubo,

I2C is the interface used to communicate the connection between the PD Controller and Thunderbolt Controller. This will tell us what connection the Thunderbolt Controller is detecting.

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

I know, Jacob, but what exactly should be checked? Is there a parser for the communication? Besides, doesn't this communication start after some PD contract is reached? At least in terms of Sink and Source?

Lyubo

0 Jacob Ontiveros over 3 years ago in reply to Lyubomir Tomov

TI__Expert 6280 points

Hi Lyubo,

Please refer to the Titan Ridge Datasheet:

8.12.3 Titan Ridge Data Status Register

This register is read from the PD Controller, you can use and I2 C analyzer for this. You have to look at all the traffic during the connection.

Jacob

0 Lyubomir Tomov over 3 years ago in reply to Jacob Ontiveros

Prodigy 65 points

Hi Jacob,

As I wrote you, we don't have the setup in-house. Isn't there a way to fix this with the PD FW settings ot the NVM Titan settings?

After all, we are using reference design with FW provided by TI and Intel.

Shouldn't it be "working"?

Besides Intel advised that we talk to TI about this issue.

Lyubo

Interface

Interface forum

TPS65983B: Two port TB3 device doesn't want to do PD negotiation on second port.