TUSB4020BI: Downstream port failures and will not connect to devices

Attached is the picture I thought was inserted in the the original post.

This is a scope capture of a "good' connection on my HW. Yellow = DP, Blue = DM and Red is the differential DP-DM. Note the good K-J chirps after the long device K chirp.

And this is a failed connection. Note the lack of K-J chirps on the blue DM signal after the long device K chirp. DP K-J chirps look OK.

Hi Paul,

The DM line looks damaged.  Can you try connecting a low speed device to the port (USB mouse) and capture what happens on the bus?  Also can you ohm the DM line to ground on a good port and on a bad port - without any downstream devices connected.

Regards,

JMMN

Hi JMMN,

Thanks for the quick response. I connected a Low Speed (LS) Mouse and it connects and works on both good and "bad" downstream (DS) ports. But there are obvious differences.

Here is a LS mouse connection on a good DS port. Again Yellow = DP, Blue = DM and Red is the differential DP-DM. On mouse connect (first transition) the mouse pulls DM high to signal a LS device. DM goes to about 3.3V here. Hub then drives DP/DM low to start a reset and proceeds form there to connect and enumerate OK. Both the DP and DM signals can drive to full 3.3V.

Here is the same connection on the "bad" DS port. The signalling and timing look the same, but note that the DM signal can only drive to about 2.3V instead of 3.3V. So definitely looks like extra loading on the bad port DM line. But I guess it is still good enough for the LS logic to work and allows the mouse to operate at LS on the bad port.

I also measured the the DP and DM resistance to ground as follows on two PCBAs, both with one good port and one bad port. DM lines on the bad ports are definitely lower than the good ports. Readings that look different to me are highlighted in yellow.

To summarize:

1) Low Speed device works on both good an bad ports but DM voltage levels on the bad port are low and seem to have extra loading on them.

2) High speed device will not connect on bad port. The extra DM line loading seems to be too much for HS to work.

3) Resistance readings on the DM line of bad ports are low.

So question now is what could cause this? And how to prevent it in the field?

Thanks,

Paul

JMMN,

one other probably unrelated item, but never hurts to double check. I noticed that the original designer left the TEST pin not connected so it is floating. In going through the TUSB4020BI datasheet and check lists, I noted that it should be pulled to ground. Will this cause a problem or in any way related to this issue?

Thanks,

Paul

Hi Paul,

Yes, the TEST should be pulled to ground, but there is an internal pulldown on it and if the hub was entering test mode, the entire hub would not be functional.

It is unusual to see port damage across multiple units in the field unless there is a common denominator like a bad cable or device that is causing electrical over stress events.  For example a frayed cable that is occasionally shorting power to the DM line,  Is there anything in common that got attached to all the bad units?

Regards,

JMMN

Hi JMMN,

This is a very unique application. On the DS port side there aren't any USB cables or connectors involved. It is all mechanical. On the data recorder node there are just 4 metal pads on the outside of the enclosure that when placed in the rack they just mate to mechanical spring clips that are mounted directly on the PCBA with the hub chip. Think old style 900MHz cordless phones that are placed in the charging cradle when not in use. So no other USB devices could ever be attached, just nodes. Now they are inserted and removed quiet frequently. The four pads are Gnd, V+ (for charging the internal node battery), DP and DM.

So the difference with a standard USB application would be that we have no back shell shield that mates first as you have when you plug a USB cable in that is connect to a chassis ground. Also as the design currently is, we cannot guarantee what signal makes connection first when inserting a node. We have looked at ways to make sure ground mates first but haven't worked that out.

But having said all that, this has been in production for two years with hundreds of units in the field. We have seen some isolated issues in the past but are now suddenly seeing a sharp increase. But it is in a dessert  & dry environment where ESD could bemore of a problem. But with the ESD protection of the hub chip itself and the added TI TPD2E2U06DRLR ESD protection IC, we assumed it would still be OK.

Any idea what magnitude of an ESD or over voltage event would be required to cause this type of damage? Assuming you concur that the DM pin has been damaged this way?

Thanks,

Paul

JMMN - should also note that the way the node is designed, it's internal USB device controller (a FTDI FT232HL) is not powered up when the node is inserted. The nodes first detects the presence of the external V+ then after a delay to allow any bouncing to die out, it then turns on the FTDI IC to start the USB connection to the HUB.

The node's USB will be powered and communicating via USB when it can be suddenly removed from rack without warning. But that is no different than any other USB device.

Paul

Hi Paul,

Yes, it seems like some sort of EOS event, but it would be hard to determine what voltage or duration.  Can you check a few of the other bad units to see if the damage is consistent to DM on other boards?  That would indicate that there may be one particular node that is causing issues rather than it being randomly occurring.

Regards,

JMMN

Hi JMMN,

I checked a third board and this one has a low ohm reading on the DP signal (about 10K instead of 20K as the other signals). So the issue does not seem to be confined to just the DM signal. Looks like the damage can randomly be either USB data line. Thus doubtful that a particular node caused this. That would be very low probability anyway the way the nodes are used in inserted into the racks.

A bit more background on our system. This is how the USB is connected:

PC (Linux OS)  <====> 24 port powered USB HUB <====> 24 two port cradles (TI HUB IC here) <=====> 2 nodes/cradle (x24, 48 total nodes per rack)

The two port cradles are powered from a separate 12V supply but are enabled and powered up only when a good 5V is sensed from the 24 port powered hub via a standard USB cable to each cradle. On the cradle PCB we have two battery chargers (one for each connected node) that generates 8.6V charging voltage to the nodes when connected. The TI  TUSB4020BI then connects the USB to each of the two nodes.

We apparently have the ability to turn off the 24 port powered USB HUB via SW from the Linux PC. This in turn would turn off the USB 5V to the cradles that would then also power down the cradle. Therefore there would not be any active voltages or signals on the cradle to node interface when nodes are inserted. The idea of powering down the USB hub and cradles while nodes are inserted/removed has been proposed.  I think this might help if the issue is with some over voltage or glitch/spike due to live power being applied as a node is inserted/removed. But I am not sure it would help or hurt if the issue is purely ESD discharge. I don't have a feel for if the PCB and ICs are more or less susceptible to ESD damage with power off. Thoughts?

Also assuming it is ESD/EOS, can you or TI recommend additional ESD/EOS protection/filtering components or circuits to further harden the USB signals?

Regards,

Paul

Hi Paul,

It is difficult to recommend protection without knowing what's causing the failure.  Have you removed any of the units from the boards?

Please accept my friend request and let me know if you can send me a bad unit.

Regards,

JMMN

Hi JMMN,

Sent you a private message.

Paul

ok, responded there