8

I am trying to source a FFC/FPC cable for USB3.0 (+5gbps). This brought me to the question of signal transmission. I'm somewhat of a beginner in this topic. I know that you should match impedance on your PCB with your connector/cable to minimize reflections.

I was wondering how to tell how fast of a signal you can transmit over a wire. Specifically what kind of cable parameters affect transmission speed? Any help is appreciated.

Nick
  • 93
  • 1
  • 5
  • 1
    Quick question: are you more interested in the physical properties of the wires or are you more interested in the particularities of fast transmission protocols. – Simon Marcoux Sep 10 '18 at 16:35
  • physical properties – Nick Sep 10 '18 at 16:39
  • 2
    don't forget to accept an answer once all your questions are answered. – Simon Marcoux Sep 11 '18 at 00:57
  • When you increase the frequency the em wave tends to free propagation (the same of the aerials). This is the reason of using wave guides or coaxial wires (that are the same thing) for microvawaves propagation. The external conductor constrains the wave to follow the path – Gianluca Conte Sep 11 '18 at 12:51
  • Some people seem to imply that USB 3.0 signals need pre-emphasis and equalization after traveling over a FFC cable: https://patents.google.com/patent/WO2017101329A1/en – davidcary Oct 23 '20 at 22:58

4 Answers4

22

The maximum frequency is mostly related to the frequency-dependent loss characteristics of the cable. Eventually you get to a frequency where you simply don't get enough signal at the other end to use.

  • Resistive losses in the conductors (including skin effect)
  • Dielectric losses in the insulating materials
  • Radiation losses if the cable is not fully shielded

All of these tend to increase with frequency.

This is why we generally switch to other technologies at very high frequencies: waveguides for microwave radio equipment and optical fibers for high-speed data.

Dave Tweed
  • 168,369
  • 17
  • 228
  • 393
  • I would upvote but I do not have enough rep. Thanks for your response! – Nick Sep 10 '18 at 16:41
  • Is the signal loss due to resistance in the wire? – Nick Sep 10 '18 at 16:41
  • 1
    Partly resistance in the conductors, and also dielectric losses. There are also radiation losses if the cable is not fully shielded. All of these tend to increase with frequency. – Dave Tweed Sep 10 '18 at 16:43
  • @Nick, upvoted for you – jsotola Sep 10 '18 at 16:44
  • How do the dielectric losses come into play? – Nick Sep 10 '18 at 16:45
  • 2
    The signal imposes a time-varying electric field across the dielectrics used in the cables. Charges within those dielectrics move around in response to those fields, and sometimes they don't conserve all of the energy used in that motion. For example, if the material is at all piezoelectric, some of the energy goes into distorting its physical shape, which eventually turns into randomized heat. – Dave Tweed Sep 10 '18 at 16:48
  • 2
    Don't forget about dielectric dispersion (different phase velocities at different signal frequencies). This technically isn't loss but causes distortion. – Captainj2001 Sep 10 '18 at 20:04
  • @Captainj2001: True, but that kind of distortion can be equalized out. – Dave Tweed Sep 10 '18 at 20:09
7

You can't just "source" FFC/FPC cables for USB 3.x. These cables (and corresponding connectors) are not qualified for USB 3.x channels. The cables for USB 3.0 have to meet many more requirements than just a wire parameters and not just having certain differential impedance.

To use non-standard (not within USB defined configuration) cables you will need to run all USB cable qualification tests on your own, ensure limits on insertion loss, NEXT/FEXT crosstalk, differential impedance across mated connectors, etc. etc., if you want your product to work with any reasonable degree of reliability.

To run your own qualification you will need at least a 8-16GHz oscilloscope and a 20-GHz TDR (Time-domain reflectometer) instrument, plus make a dedicated break-out fixture to access signals in correct way. The list of electrical requirements for USB 3.0 transmission lines is given in the following USB-IF document. Although the document is mostly for qualification of standard cables and mating connectors, appendix to the document shows the general electrical requirements to meet.

Harry Svensson
  • 8,139
  • 3
  • 33
  • 53
Ale..chenski
  • 38,845
  • 3
  • 38
  • 103
  • You will need to learn to use SMA connectors. – analogsystemsrf Sep 11 '18 at 03:22
  • @analogsystemsrf, didn't you mean "OP will need to learn SMA connectors"? And not forget to buy a 5/16 properly pre-set torque wrench for them... just $216.26 from Pasternack for example... :-) – Ale..chenski Sep 11 '18 at 04:33
  • 1
    @analogsystemsrf Have you seen a USB 3 connector on a motherboard? They're pin headers. – user71659 Sep 11 '18 at 16:55
  • @user71659, Have you seen any USB 3.0 test fixture? http://www.usb.org/developers/estoreinfo/SuperSpeedTestTopologies.pdf – Ale..chenski Sep 11 '18 at 17:22
  • 1
    @Ale..chenski That's because you want the fixture to be easily de-embedable. It's not necessary for the normal operation of USB 3.0. – user71659 Sep 11 '18 at 17:24
  • @user71659, we are talking about "qualification" of USB 3.0 interconnect link, not "normal operation". To do so, you need reliable and repeatable connection to instruments, BERT generators, gigahertz scopes and TDR, all of them use RF type connectors. So yes, you want connections that can be de-embeddable with confidence and repeatability, preferably without any de-embedding and re-calibration. – Ale..chenski Sep 11 '18 at 17:39
4

The frequency that can be used inside a wire is highly dependant on the skin effect. Simply put, the bigger the wire, the lower the frequency it can carry without having signal loss caused by an increase of its impedance.

At low frequency, the signal will be equally distributed through most of the wire, with a higher frequency, the signal will be predominantly distributed around the perimeter of the wire (the ''skin'').

The wires that allow the best characteristics will always be really small and with multiple conductors to reduce the skin effect. Despite this, the higher you go, the more loss you will get. Then the protocol steps in and will increase the voltage, use twisted differential pairs and push the boundaries to the maximum until you need to switch to a different transfer technology altogether.

Simon Marcoux
  • 1,277
  • 7
  • 19
0

I was wondering how to tell how fast of a signal you can transmit over a wire. Specifically what kind of cable parameters affect transmission speed? Any help is appreciated.

The Comcast XB6 Cable Modem will do over 1.5 Gbps using your standard cablevision coax. The speed is limited to your last-mile speed, otherwise it would be higher.

PCIe 5.0 does ~4GB/s (or x16 @ ~128GB/s). A x1 connection, the smallest PCIe connection, has one lane made up of four wires. It carries one bit per cycle in each direction.

So 2 pieces of wire can do ~2GB/s in practice, in theory you could squeeze some more out of it. For plain wire coax cable is the fastest because it's shielded. Along with shielding length is the next most important factor, with shortest distances (inches) being the best.

Rob
  • 259
  • 1
  • 5
  • 12