8
\$\begingroup\$

USB3 offer a much faster bandwidth than USB2 but, technically, at the price of more wires that results in a thicker cable. USB3 is build with 9 wires, 4 are standard USB2 and 5 are the USB3 serial (2 differential transmit, 2 differential receive and one GND).

In an effort to make a thin & flexible cable, would it be possible to make a 5 wire cable, just with the USB3 wires? Assuming that both sides of the cable are USB3?

\$\endgroup\$
  • 5
    \$\begingroup\$ Not if you actually want to implement USB 3.0... \$\endgroup\$ – Ignacio Vazquez-Abrams Apr 23 '14 at 14:06
  • 4
    \$\begingroup\$ Even with USB3 devices, isn't power, enumeration, etc. still handled on the original 4 wires? The additional pairs are for LVDS data only. \$\endgroup\$ – Dave Tweed Apr 23 '14 at 14:22
  • 3
    \$\begingroup\$ Isn't that basically what OP is asking? I too have no idea if usb3 requires usb2 d+/d- to work, or just to meet the standard. Good question actually. \$\endgroup\$ – Passerby Apr 23 '14 at 17:40
  • 1
    \$\begingroup\$ Oh, and since USB3 allows for higher current sourcing than the USB2 500mA "standard", you might see thicker VUSB and GND cables, which would negate any gain from non-usb2 compatible usb3 cables. \$\endgroup\$ – Passerby Apr 23 '14 at 23:08
8
\$\begingroup\$

I could not find any information regarding initial synchronization for USB3. The best idea I could come up with is insulating the UTP pair from the old USB and trying it out.

So I covered the two center data pins on my USB3 hard drive cable with some kapton tape and plugged it in. The computer negotiates with the drive fine. No longer initial sync than normal.

It is likely that a 6 wire USB3 cable would work. 4 for LVDS pairs and Power and Ground.

However, most cables have a separate ground for each LVDS pair. This is to give return current paths other than the pair. This is helpful for signal integrity, just like how return current actually flows in the ground plane under LVDS pairs, not in the other pair, on a PCB.

So you are actually loosing only 2 of 10 wires.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Taking one for the team :D +1 \$\endgroup\$ – Passerby Apr 23 '14 at 23:08
  • \$\begingroup\$ Weird. USB 3.1 spec describes a cable with no VBUS D+ or D- though, in section 5.5.2: USB 3.1 Standard-A to USB 3.1 Standard-A Cable Assembly \$\endgroup\$ – endolith Sep 26 '16 at 16:45
7
\$\begingroup\$

The nine wires used in a USB 3.0 cable are shown below.

1   VBUS                        Power
2   D−                          USB 2.0 differential pair
3   D+
4   GND                         Ground for power return
5   StdA_SSRX− / StdB_SSTX−     SuperSpeed transmitter differential pair
6   StdA_SSRX+ / StdB_SSTX+
7   GND_DRAIN                   Ground for signal return
8   StdA_SSTX− / StdB_SSRX−     SuperSpeed receiver differential pair
9   StdA_SSTX+ / StdB_SSRX+

Leads 1 through 4 are used for USB 2.0 connections. Leads 5 through 9 were added for USB 3.0.

VBUS and GND are power wires, and can be used to power a device, supplying either 100 mA or 500 mA at 5V.

Note that GND_DRAIN is not the same as GND. The two differential pairs are each wrapped with a ground shield, and GND_DRAIN is connected to this shield. GND_DRAIN is not internally connected to GND inside the cable. I assume that GND_DRAIN is connected externally to GND only at one end of the cable to avoid ground loops.

The differential pairs are outputs from one end, and inputs to the other, therefore the dual labels.

According to this article under the heading Bus Speed, the transmitter first tries to detect the termination of the differential pair on the receiver side. If none is found, then the host drops back to USB 2.0 and uses the D-/D+ pair for communication. From this I assume you might be able to get by without the D-/D+ leads 2 and 3.

If you are acting as a host, you must supply 5v on the VBUS line. If you are a device, you can ignore this line if you are self-powered.

So as a minimum, you might be able to get by with the six leads 4-9.

\$\endgroup\$
3
\$\begingroup\$

I bought a USB 3.0 extension cable and check whether D+/D- can be ignored by 3.0.

  1. Extension cable works good with 3.0 and 2.0.
  2. I cut D+/D- 2.0 data line from the USB 3.0 cable.
  3. there was no USB 3.0 connection. no 2.0 of course.
  4. I re-bind D+/D- 2.0 line, connection OK.

From my experiment, I think D+/D- cannot be ignored even for 3.0. I test this with USB 3.0 memory and USB 3.0 hub. Both case no connection without D+/D-.

\$\endgroup\$
2
\$\begingroup\$

The report from Doojin is wrong, something else was not right. There is no reason for USB3.0 not to work if D+/D- wires are cut off. I just did the same experiment, and a USB3 pen drive works just fine.

The USB3 connect protocol is as follows:

  1. USB3 device gets plugged in. Rx channel gets terminated with 45 Ohms to ground, but D+ is not pulled up yet.

  2. The USB3 host sends a positive pulse on both Tx lines, and watches for RC response. The host does this periodically at about 100ms intervals.

  3. If USB3 termination is present, the line charge timing is defined by Rsense and 0.1uF decoupling cap, which is connected to ground via the Rx termination. The special circuitry in host transmitter detects this, see section 6.11 of USB3.1 specifications, Figure 6-36. Then the host starts Polling.LFPS signaling, following by a handshake from device, and link training protocol begins. USB 3.0 communication gets established. NOTE: The D+/D- are not involved.

  4. If USB3 Rx termination is not present, the line charge is much faster since it is determined only by bulk cable capacitance. The host detects this and does not proceed with Polling.LFPS and USB3 connection is not established.

  5. If the above USB3 connection handshake fails, only then the device tries to connect as USB2 device, by pulling D+ high. If D+/- are not present, no USB connection will be established.

Conclusion: yes, it is possible to remove D+/D- twisted pair of wires, and have the USB3 link working. However, does the elimination of 10% of overall wires (measured as overall AWG) really saves anything in mechanical stiffness of cables?

\$\endgroup\$
  • \$\begingroup\$ less wires is better when, for instance, using a USB over a slip-ring. Less wires is a cheaper and smaller slip-ring. \$\endgroup\$ – Gilad Nov 2 '16 at 17:59
  • \$\begingroup\$ @Gilad, slip-ring maybe, but I am not sure if the ring can maintain proper signal integrity at 5Gb/s USB 3.0 signaling rate. \$\endgroup\$ – Ale..chenski Nov 2 '16 at 18:34
  • \$\begingroup\$ As of today, there are no slip-rings that can handle 5Ghz. The common solution today is to convert USB3 to optical and use an optical slip-ring. Hopefully slip-ring manufacturers will solve this issue soon. \$\endgroup\$ – Gilad Nov 2 '16 at 18:57
  • \$\begingroup\$ @Gilad, I appreciate your remarks, led me to some learning. It appears that the modern technology uses only two fibers to get both USB3.0 and USB2.0 over the optical cable, see corning.com/microsites/coc/ocbc/Documents/CNT-075-AEN.pdf So there is no need to reduce any wires and cut USB2. \$\endgroup\$ – Ale..chenski Nov 2 '16 at 19:40

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.