Brad Karp wrote:
> So in full, the picture is:
>
> PoE 12V
> port0 ------------> DWL-P50 ----> net4826
> / |
> HP PoE switch | USB cable
> \ |
> port1 ------------> DWL-P50 ----> DUB-H7 USB hub
> PoE 5V
Generally the USB hub expects to be powered either from the
device it is hanging on, or externally where the power
supply is galvanically isolated from the data lines.
Reason: There is no guarantee that a signal ground
of the data lines is the same as the power ground
of the power supply. So either they have to be the same
(internally powered), or isolated so that they can
float to the same potential.
Secondly, I don't know about PoE, but I don't think that
it guarantees much beyond that the device gets its power.
The signal grounds between the two devices on a PoE might
be the same, or might be 48V apart.
Depending on the devices used what you are trying to do
might or might not work, but is surely unsupported at best
and can kill the devices used at worst.
Generally, loops with power involved are bad. Very bad.
Note: ethernet connection itself is galvanically isolated.
> We find that while the net4826 can talk to the USB hub when the USB
> hub is *not* externally powered (i.e., unplug the 5V cable in the
> above picture), the net4826 fails to talk to the USB hub when the USB
> hub *is* externally powered.
If the signal and data ground differ on the net4826 or the
ground potentials of the DWL-P50s differ, the hub gets one
idea of ground via the net4286 and the other one via power
supply. Bad.
> Moreover, we seem to have *permanently* killed the internal USB
> hub on at least one of our net4826 boxes by connecting an externally
> powered USB hub in this way.
Well possible.
> Before you tell me that we're powering the hub incorrectly, consider
> that when we power the USB hub in exactly this way over PoE, it works
> flawlessly when connected to *any other* type of desktop PC we've
> tried.
Depends on the devices used.
> Even more strangely, when we power the USB hub not using PoE, but
> using the AC adapter supplied with the USB hub, the USB hub works
> *fine* when connected to the net4826. But the AC adapter and PoE
> splitter output are both extremely close to 5V.
Because the AC adapter is galvanically isolated from the rest
of the world.
> We're mystified. Has anyone seen behavior anything like this
> elsewhere, or might anyone have a hypothesis as to why PoE behaves so
> strangely with this combination of devices?
In the presence of the loop in the power supply connections
this behaviour is completely logical.
> We're trying to find a work around as quickly as possible, as we need
> to purchase dozens more small PCs ASAP. We'd like to use more
> net4826es, but the problem with USB and PoE is a showstopper for us,
> as the main role of these boxes in our lab is to control USB devices.
You need to galvanically break the power loop. Either power one of the
DWLs via an AC adapter instead of via PoE, or the Soekris,
or the USB hub.
If this is not possible, there are galvanically separated DC/DC
converters that can help, but I have no idea how much cost
they add to your setup.
One final note: an externally powered hub can be quite demanding -
500 mA per port. The net4826 will also draw a good 1A. Are you
sure the DWL-P50 can handle it without melting?
Regards
--
Stano
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
