I just tried to connect a (relatively) cheap USB3 hub (GL3520) from eBay to a brand new USB3 controller (based on Fresco Logic FL1100, replaced a VL800 based one).
The USB3 hub is marked as "bus powered" on its package, but has a 5V supply connector and the seller added a 5V/2A supply.
So while everything worked fine at first, I noticed the power LED of my PC wouldn't turn go out when I turned off the PC.
I noticed that the new USB3 hub obviously supplied (parts of) the 5V line of my PC and its supply got pretty hot.
I opened the hub and there doesn't seem to be any kind of protection or transistor switch that would separate the VBus of the USB connector from the 5V supply.
Indeed I'm pretty sure that they are directly connected even though I was too lazy to really measure it.
I then looked inside an older USB2 hub of the same manufacturer. In this case it's clearly visible that the 5V DC input is directly connected to VBus on the USB connector.
However this hub doesn't have the same effect on my PC. The USB2 devices are connected to my mainboard's USB2 connector (Intel chipset), so I assume that in this case, the mainboard has a reverse current protection.
In a nutshell: is the hub crap or the controller or both? Isn't there any general agreement that either a (powered) hub or the controller or both should avoid current flowing from the hubs/devices into the controller and then into the PC?
To use the hub (always) powered, I guess it should suffice to cut the 5V line going to the Vbus pin on the USB connector. I think I already spotted a 0Ohm resistor that could do the trick.
Then again I wonder what would be the proper solution. Would it be enough to use a P-FET as in a normal reverse current protection scheme?
Like gate is connected to ground, drain to Vbus on the connector, source to 5V supply. So the channel is deactivated if Vgs is below the threshold and the body diode blocks current from the 5V line back into Vbus.