I need some help..

I developed a test platform for the hardware tests of an electronic board panelized in sets of 10. The boards are placed on the test platform in their panelized form. As shown in the schematic below, the platform powers only one board at a time, allowing communication between the board's firmware and the test software via USB. The problem is: when I place a single board on the platform from any slot, communication via USB works fine. However, when I place the entire panel (i.e., 10 boards at the same time), even though only one board is powered at a time, the USB communication with the powered board fails.

I think this is due to impedance mismatch on D+ and D- lines. Is there a way to fix this?


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ At what speed are you running your USB? HS? FS? \$\endgroup\$ Commented 17 hours ago

2 Answers 2


It likely is not due to impedance mismatch.

A more likely explanation is that either your boards or the chip where the USB lines go to have input protection, which clamps the voltages on D+/D- wires to supply voltage.

As unpowered boards have no supply voltage, they try to clamp the USB data wires and power up via them, loading the data wires down to a level it simply does not work.

So basically, it is not even related to USB, you cannot have (in general, exceptions apply) unpowered chips on same wires where you try to communicate.

And when it is related to USB, you cannot have multiple boards hanging on a single bus like that, for impedance matching and signal reflection reasons.

You will solve it by muxing the USB data to each board as well, with suitably high speed analog muxes or USB hub, or simply testing one board at a time.

  • \$\begingroup\$ Thanks for your comment. I use nRF52840 chip for the boards. I couldn't find a note there was be any protection for the USBD pins. \$\endgroup\$
    – OzkulA
    Commented Jul 10 at 10:29
  • 5
    \$\begingroup\$ @OzkulA Absolute maximum ratings for IO is VDD+0.3V. If VDD is 0V, then max you can have on IO pin is 0.3V without damage and it means you can't expect USB communication to work if unpowered chips are hanging on the bus. \$\endgroup\$
    – Justme
    Commented Jul 10 at 10:42
  • \$\begingroup\$ I faced something interesting.. When I placed only 2 boards to the platform, the board located near the USB connector establishes a USB connection, but the board located farther away cannot establish a connection. When three or more boards are connected, none of them can establish a connection. \$\endgroup\$
    – OzkulA
    Commented Jul 10 at 11:16
  • \$\begingroup\$ @OzkulA well, reading Justme's answer and their comments, how is that surprising? More clamping diodes clamp stronger. \$\endgroup\$ Commented Jul 10 at 12:16
  • \$\begingroup\$ @OzkulA It is surprising. I expected it not to work at all, it really shouldn't. Even if it works with two boards it is still incorrect way to do it and a bad idea to continue using it and the design error should be fixed one way or another. \$\endgroup\$
    – Justme
    Commented Jul 10 at 12:21

You should put a USB Hub chip on the test assembly. That will do a fine job of isolating all the DUT USB lines from each other. It is actually easier to implement than dealing with muxes you'd have to control and worry about. USB hub chips are fairly robust and designed for disconnects. So that would be what I'd recomment.

Yes, you'll need to redo that test assembly.

  • 1
    \$\begingroup\$ A couple of problems though - USB relies on mechanical connector sequencing the GND/power and data pins, what a custom test platform (with e.g. pogo pins) does not do unless designed so. Also 10-port USB hub chips are unlikely to exist, so a solution could be to chain two 7-port chips or three 4-port chips. \$\endgroup\$
    – Justme
    Commented Jul 10 at 15:20
  • \$\begingroup\$ Good point about the pin sequencing. Easy to do with pogo pins since they are easily replaceable as wear items - they just slide into their sleeves. So getting a pogo pin with more stroke and/or a bit longer uncompressed length is the easy solution. \$\endgroup\$ Commented Jul 11 at 13:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.