FEDEVEL
Platform forum

MLc , 10-20-2023, 04:25 AM

Hello everyone,
I want to ask you for help what is wrong. Why this happen.
I designed board with 100W PD(MAX25432B), USB Hub based on USB5744 chip. I don't use aby custom configuration through SMBUS for Hub driver.
USB2.0 communication/data transfer worked without problem.
I have problem with hub in USB3.0 data communication. When I connect USB3.0 flash drive, it was connect only for a while then disconnect, again connect and this happens all the time. One time USB3.0 was connected longer without falling out of communication so I managed to try copying the data. It worked. So, I am thinking about some EMC/EMI noise?? Something disturbance tx/rx lines?
I use ESD protection: TPD4E02B04DQAR for tx/rx lines and ESD401DPYR for data+- and CC lines.
My stackup is: 1L-power + signals (critical signals), L2-GND plane, L3-signals, L4-power+gnd pour

Any advices?



USB-C wiring on board




Thank you
Martin

qdrives , 10-20-2023, 03:13 PM

You might find some help using Wireshark to report messages about the USB connection.
Is it your hub that is disconnected or the device?

MLc , 10-21-2023, 02:44 AM

Hi, this is good question... thank you.
I use USB3.2 C-C cable. I checked it and I think that hub is disconnected in loop sequence. I opened device manager and I saw when my HUB is connected to laptop it still refreshing/searching for hardware changes. Few times I saw that Generic SuperSpeed USB Hub was disconnected and only Generic USB Hub was connected. After few refresh/scanning for HW changes SuperSpeed USB Hub was found again. Also, I saw higher % CPU usage for Windows Explorer process with USB3.2 C-C cable.
My USB3.2 C-C cable is with E-mark chip.


I did another test. I have USB2.0 C-C cable. I use that. I saw in device manager only Generic USB Hub . That's logical. With USB2.0 cable I don't have problems with refreshing device manager in loop. USB2.0 C-C cable work without problems.

So, in my opinion there is some problem on SuperSpeed communication lines. I did schematic wiring according to EVAL board. One different between EVAL board schematic and my is that: I swapped RX lines polarity and they swap TX lines polarity.

I will try the WireShark too. But need to learn how to use it.

Martin

qdrives , 10-21-2023, 05:31 AM

What if you connect devices in USB 2 to the hub and the hub in USB 3?

Other tools:

MLc , 10-21-2023, 06:55 AM

If I connect usb2.0 device (mouse, usb flash) to the hub which is connected in 3.2, the device work.

MLc , 10-22-2023, 07:35 AM

UPDATE:
If for some reason the USB3 flash remains connected, it is not connected through SuperSpeed communication...

qdrives , 10-22-2023, 12:49 PM

Are all solder connections good?
Did you check under a microscope or did x-ray?

MLc , 10-22-2023, 01:04 PM

I checked it under magnifying glass only. But I will try to resolder hub driver and ask at school if they have microscope.
Also, look...
I don't know how, but it stay for a while in SS connection. But the connection failed after while.

qdrives , 10-23-2023, 12:53 PM

Is your stack-up the same as the evaluation board?
Is the actual board stack-up as you designed it? (cross section the board)
Are the components correct? ie. are the termination resistors the correct value (physical).

MLc , 10-23-2023, 01:10 PM

Hi,
I don't know if my stackup is the same as microchip because they share in pdf only top and bottom layer(they have 4Layer too). I think the have in both inner layers GND/PWR planes. I have only GND plane in L2.
How I checked if the actual board stack-up is as I designed it?
Components are correct. I measure it.

But, maybe I found the problem.... I think.... because USB3.0 in SS sometimes worked for a while.
Microchip said in AN26.21: The length of a SuperSpeed USB differential pair traces must be within 0.13mm of each other. I didn't do a length matching is SuperSpeed lines ... So, the way how to find out if this makes the connection looses is -> order the new redesign board with correction :|

qdrives , 10-23-2023, 01:31 PM

How much is your length difference?
0.13mm is very small. What is the rise time for USB3?
That distance would be for ~1ps skew.
Did you design the stack-up to have the correct characteristic impedance?

MLc , 10-23-2023, 01:50 PM

For example:
UsbC RX lines - 71.103 vs 72.421mm
UsbC TX lines - difference 1.152mm

Downstream port1 - RX difference 0.401mm, TX difference 0.019mm
Downstream port2 - RX difference 0.175mm, TX difference 0.019mm
Downstream port3 - RX difference 0.595mm, TX difference 0.594mm
Downstream port4 - RX difference 0.594mm, TX difference 0.594mm

Mostly for my tests I use the port1 or port2.

What do you mean with "rise time for USB3"?
I design the stack-up according my layout and easy routing and use the impedance setting in Altium to set Diff pair impedance matching 90ohm.

robertferanec , 10-24-2023, 09:59 AM

Did you try different USB cables? Possibly, did you try the same cable with development boards?

MLc , 10-24-2023, 10:17 AM

I tried it only with usb2.0 C-C cable and usb3.0 C-C cable. But will try it with another usb3.0 C-C cable.

I don't try it with development board. I don't have it. This is my first circuit and testing USB5744.

robertferanec , 10-24-2023, 10:39 AM

many USB cables are bad, so I would recommend to buy number of them and test them .... just to be sure the cable is not the issue.

MLc , 10-25-2023, 08:10 AM

Hi, I bought another C-C usb3.0 cable. But the working is same.
USB hub is connected in superspeed until there is no USB3 device connected to him. After I connect device the issue with disconnecting start.
So cable is not a problem, probably..
Also, my usb3.0 flash drive die. Stop working today. I think the hub kill him..

qdrives , 10-25-2023, 02:35 PM

USB signals either 2 or 3 cannot damage a product even if the D+ and D- are reversed.
Your device is now defective, than the problem should be more in the Gnd, power or shield.

MLc , 10-26-2023, 02:09 AM

Hi folks,

I have plan... I re-design my board
- I will apply the length matching to data and RX/TX pairs.
This is the only thing that has not been tried yet. And microchip telling about it in AN26.21 and AN26.2.
Each member of a High-Speed differential pair should be no more than 1.25 mm (50 mil) longer than the other member. Each member of a SuperSpeed differential pair should be no more than 0.05 mm (2 mil) longer than the other member. The transmit differential pair (TX+ and TX-) does not have to be the same length as the receive differential pair (RX+ and RX-)
and
The length of a SuperSpeed USB differential pair traces must be within 0.13mm of each other.

- And also I will try to change the stack-up to -> L1 Signal, L2 Plane GND, L3 Plane PWR/GND, L4 Signal.

Your opinion on it?

EDIT:
How to make the length matching on the lines from HUB IC to USB-C connector? Because the pins are in USB-C connector doubled. So, do the length matching with final length together with the short connection in connector?
With light blue lines or only match the red lines?

qdrives , 10-26-2023, 02:31 PM

I would match the length via to via or pad to via.