1
\$\begingroup\$

USB 2.0 Full Speed (12Mbit/s) and USB 2.0 High Speed (480Mbit/s) are both specified as having a maximum cable length of 5m.

Why does Full Speed not allow a longer cable length, given its much lower bit rate? (I'm assuming that it's a voltage drop problem, rather than waveform issues, but I do not know.)

How can you extend USB 2.0 Full Speed to 20m (cat5 cable) using active adapters at each end?

i.e. What do devices such as https://cpc.farnell.com/pro-signal/usb-boosta/usb-booster/dp/CS16464 contain?

\$\endgroup\$
  • 2
    \$\begingroup\$ A complete guess: Full Speed is an older standard that's less strict about cable properties, even though any cable you buy today will almost certainly conform to the High Speed requirements at a minimum (and thus would probably, but is not guaranteed to, work over longer distances). \$\endgroup\$ – Hearth May 13 at 13:20
  • \$\begingroup\$ @Hearth you are correct that Full speed is an older standard and any 5m or less cable you can buy today will handle High Speed requirements you should not be able to buy Full/High speed cables longer than 5m. Only low speed devices, such as mice, can have longer cables and these must not be disconnect-able at both ends. \$\endgroup\$ – Warren Hill May 13 at 13:37
  • \$\begingroup\$ Reading: superuser.com/a/1105099/142433 One of the main issues is reflection times from a long disconnected cable, this makes sense regarding your comment about mice. \$\endgroup\$ – fadedbee May 13 at 13:43
  • 1
    \$\begingroup\$ My suspicion is that because USB1.1 is half duplex and they use cat 5 cables the protocol is similar to Ethernet between the boxes. \$\endgroup\$ – Warren Hill May 13 at 13:43
  • 1
    \$\begingroup\$ It can be extended by for example wrapping the USB packets to some other high speed interface. That's how HDBaseT can extend USB 2.0 interface for 100m (among other interfaces). \$\endgroup\$ – Justme May 13 at 15:37
1
\$\begingroup\$

Limit on USB cable length is dictated by overall USB architecture of having 5 hubs, and artificially-imposed time-out of device responses (due to half-duplex nature of USB 2.0 communication protocol). The worst-case connectivity model in USB 2.0 is:

enter image description here

Overall, per specifications, a USB host will wait 18 UI (18 * 83.33ns) = 1500 ns (in FS mode) before detecting a failure to respond, and will re-try the failing packet in hope of self-recovery.

The 1500 ns is the time of signal flight in both directions. Assuming the device worst case response of 400 ns, this leaves 1100 ns for signal flight, or 550 ns in one direction. Again, assuming typical dielectric qualities of USB cables, or 5.2 ns/m, this gives a working distance of 105 m, or up to 300 ft. So, theoretically speaking, a single FS device should be working over a 300-ft bare cable.

In practice the USB over CAT5/6 extenders use some up-conversion into truly differential Ethernet-like signaling, and use some silicon ICs to do so, which incurs digital propagation delays. As result, practically-achievable distance comes down to about 150 ft: enter image description here

Please note the words "single USB device" in the product description. So be cautious: if you use this extender behind even a single external USB hub, the link will likely fail.

Also be aware that USB host designs are listening to the bus and checking if the bus is in idle state before making a decision to transmit next packet. A 50-100 ft bare cable in FS mode will produce massive reflections beyond the transmitted End-Of-Packet (FS bus is not terminated with 45-Ohms unlike in HS mode). Some overly-constrained USB host designs might construe these reflections as unsolicited bus activity, and the host port might shut itself down.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ You mention "truly differential Ethernet-like signaling", implying that USB doesn't have that. USB 2.0 FS has D+ and D- signals. Are these not a differential pair? \$\endgroup\$ – fadedbee May 14 at 21:07
  • 1
    \$\begingroup\$ @fadedbee, no, USB 2.0 signals are not fully differential. The FS bus state has extremely important End-Of-Packet state which is single-ended Zero, every packet has it. HS bus also has the idle state which is SE0. But this is not the most important. I meant to say that the intermediate link has likely impedance-matching terminations on both ends, which eliminates reflections. \$\endgroup\$ – Ale..chenski May 15 at 0:00
2
\$\begingroup\$

Section 7.1.19 of the USB 2.0 Specification discusses delays. There you will find that signals in a cable must travel from connector to connector within 26ns (5.2ns/m max). This is why standard detachable cables are limited to 5 meters.

https://www.usb.org/sites/default/files/Propagation_Delay_between_Host_Transcievers_and_Downstream_Ports.pdf

| improve this answer | |
\$\endgroup\$

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.