3
\$\begingroup\$

I'm trying to make a connector for a box of devices that have an Micro USB 3.0 input. The box is just a few physical connectors, no USB hub is planned.

But no matter how I try to connect the devices, it detects them as USB 2.0.

enter image description here

If I use a direct USB-A to Micro 3.0 cable, it detects them as proper USB 3.0.

I cannot use an USB hub because of bandwidth, I need separate USB channels, all of them go to a dedicated root hub on the motherboard.


I tried two approaches: first cut of the A plug from a USB 3.0 A-Micro cable and solder a female USB B 3.0 to that end:

usb A-micro

Did not connect one pin as it was an OTG detect pin that did not even have a wire in the cable.

This detects as USB 2.0


Second, I tried to make a board with 3.0 female-female A-B, so I can connect a USB A-B cable from one end and a USB A-Micro cable from the other end.

usb A-B

This unfortunately also detects as USB 2.0.


From what I'm guessing, I have to exchange the SSTX+/SSTX-/SSRX+/SSRX- wires somehow. I tried to change some pairs, but wasn't able to find the proper combination.


This is what I tried, the two above methods on one board. Not too perfect solder, but electrically it was OK.

soldered

\$\endgroup\$
  • 4
    \$\begingroup\$ USB 3.0 has much tighter tolerances on how the cable is made. It's possible that your devices are trying to achieve USB 3.0 speeds, but when the connection is not stable they give up and revert to 2.0 speeds. \$\endgroup\$ – vini_i Sep 29 '16 at 16:04
  • 3
    \$\begingroup\$ Could you please provide a picture of your wiring? USB 3.0 signaling requires tightly controlled differential impedance of Tx and Rx pairs to work. \$\endgroup\$ – Ale..chenski Sep 29 '16 at 16:26
  • 1
    \$\begingroup\$ The easiest solution may be to use a USB 3.0 Hub. This would allow you to use premade cables avoiding any wiring problems. \$\endgroup\$ – vini_i Sep 29 '16 at 16:30
  • 1
    \$\begingroup\$ This cannot be a USB hub - I need almost full bandwidth from each device. Each cable is connected to a dedicated root hub in the PC. \$\endgroup\$ – Krzysztof Bociurko Sep 29 '16 at 16:53
  • 1
    \$\begingroup\$ Thanks for the picture. This wiring will NEVER work. Do you really have three independent xHCI controllers in your host? \$\endgroup\$ – Ale..chenski Sep 29 '16 at 17:03
2
\$\begingroup\$

Use a USB3.0 hub, it has precisely the purpose. You will save on two long cables. With wiring like yours, you will not have any performance gains since you likely use the same root hub in your host system.

Use of passive adapters and extenders is not encouraged by USB specifications. Solder-down wire adapters are 99% doomed. Of course you can go full-blast and develop a junction box with carefully configured USB 3.0 repeater ICs (lookup Pericom or TI or else), but it will cost you a lot of efforts.

\$\endgroup\$
  • 1
    \$\begingroup\$ No, I have dedicated root hub per cable - I do need the full bandwidth here. But why can't I extend the cable? There are for example female A to B passive USB 3.0 boxes that are supposed to work. Is my soldering too imprecise? \$\endgroup\$ – Krzysztof Bociurko Sep 29 '16 at 17:06
  • 1
    \$\begingroup\$ Your soldering uses long stretches of wires. Even if the wires are paired and coupled, differential impedance of your pairs is likely in 150-200 Ohms, which creates strong inhomogeneity along the channel, even in Case1. At 2.5GHz this wiring will create strong reflections causing intersymbol interference and likely intolerable amount of bad transactions. Also, using 1.8m cables already stretches the signal integrity to maximum, especially when uAB connectors are good only with 1m-long cables. \$\endgroup\$ – Ale..chenski Sep 29 '16 at 17:26
  • 1
    \$\begingroup\$ Distance: but I've got a perfectly working 5m USB 3.0 A-Micro cable. Impedance: is there any way for me to be precise enough for this to work? \$\endgroup\$ – Krzysztof Bociurko Sep 29 '16 at 17:33
  • 2
    \$\begingroup\$ Yes, you need to use something called TDR - time domain reflectometry - to evaluate quality of differential impedance (see Tektornix 82000 equipment). Vector network analyzers can do the job too. The best way is to evaluate signal eye diagram at the end of your channel (uA end) for compliance to USB 3.0 specifications. You will need a special adapter, a 8-12-GHz scope, and proper processing software. The level of signal jitter would be of major concern here. \$\endgroup\$ – Ale..chenski Sep 29 '16 at 17:39
  • 2
    \$\begingroup\$ Regarding the female-A to B passive adapters - they almost never work, for the same reason. Their internals are made of loose wires with no regard to impedance matching requirement. This kind of adapters barely worked at USB2 speeds (240MHz), but are a total disaster at 10x frequency. \$\endgroup\$ – Ale..chenski Sep 29 '16 at 17:45

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.