I'm designing a PCB where I need to have two possible USB 2.0 connectors (and hence 2 sets of diff pairs) connect to a single USB 2.0 port on an MCU. Only a single connector would be used at any given time. What can I do to improve the signal quality? I've heard of redrivers/retimers but can these be used in this scenario to get around the effects of reflections due to the stub I would be introducing? If so, how many would I need and where would I add them? Diagram


This would always be implemented using a switch IC rather than just a junction; with that, there'd be no stub line, no need for redriving.

Many silicon manufacturers do offer such ICs, many even explicitly specified for USB2 signaling. Check e.g. TI.com's selection of USB 2.0 switches. Or Maxim's. Or NXP's. Or ON Semi's (show "ALL" instead of 50, search for USB). Or Analog Device's. Or intersil's (now Renesas). Or ROHM's.

| improve this answer | |
  • 2
    \$\begingroup\$ It is not "probably be made easy", it is a MUST to use either electronics switch, or electromechanical (RF-grade relay) switch to accomplish the task. It is plain impossible to fork high-speed communication wires without introducing horrible reflections and demolishing all signal integrity. \$\endgroup\$ – Ale..chenski Jan 25 '18 at 0:36
  • \$\begingroup\$ @AliChen Better, now? \$\endgroup\$ – Marcus Müller Jan 25 '18 at 0:45
  • \$\begingroup\$ @AliChen by the way, what you said about impossibility of high-speed comm lines isn't strictly true; for unidirectional wires, a simple matched 3dB split would do, for bidirectional you'd need something isolated (like a Wilkinson divider). Problem with those is that it's hard to make them work matchedly over a large bandwidth (large relative to the center frequency). And since USB isn't exactly "spectrally nice" in that aspect (or generally, tbh), you're right, practically feasible is only switching. (also, it avoids the at least 3dB of power loss.) \$\endgroup\$ – Marcus Müller Jan 25 '18 at 0:58
  • \$\begingroup\$ Yes, this is better. 3dB loss will surely kill all USB signal template (eye diagram) requirements. Also, since the USB 2.0 is bi-directional, "re-drivers" are notoriously difficult to implement and do not exist. Instead, USB hubs with "repeater path" are used. Also, a fork will create logical difficulties for USB protocol if both devices are attached. \$\endgroup\$ – Ale..chenski Jan 25 '18 at 1:28
  • 1
    \$\begingroup\$ Ok, I posted my first question ever on EE, electronics.stackexchange.com/q/351917/117785. \$\endgroup\$ – Ale..chenski Jan 25 '18 at 2:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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