7
\$\begingroup\$

I understand that USB data transmission is done over various speeds where transmission line effects could come into play(max 480Mbps which means a max frequency that is much higher considering the square wave harmonics/edge rate, perhaps 5th harmonic for 2.4GHz max significant signal frequency content) as a differential pair with controlled 90 ohm differential impedance or 45 ohms to ground for each individual wire in the pair.

My question is if it's true that ANY USB cable capable of transmitting data must have D+/D- wires manufactured in a precise way as to maintain this impedance along the length of the cable. If so how is this done? Surely this must require some complex calculation involving inductances/capacitances and very specific spacing between the wires and surrounding ground shield that must maintain constant over the length of the cable, right? Do all cable manufacturers really take this into account?

Or am I wrong and the cables do not need to achieve this characteristic impedance at all? Does it depend on the length of the cable such that normal cable lengths do not need to worry about this?

Taking a random USB cable picture off google it doesn't seem like the data wires get any special treatment...

enter image description here

\$\endgroup\$
  • 2
    \$\begingroup\$ Judging from the poor quality of the ~30cm long pigtails that connect box-mounted USB sockets to the mainboard (often not even twisted) I'd say that the standard is pretty tolerant. \$\endgroup\$ – Agent_L Aug 25 '17 at 9:39
13
\$\begingroup\$

USB cables do require some precision engineering. There are stringent requirements on value of differential impedance, quality of interconnects, and amount of losses per cable. The high-speed part of USB cable, even at USB 2.0 480 Mbps data rate, is made of a twisted pair of wires, all wrapped into a shield. This makes it a "bi-axial" cable. The cable is done as uniform as possible, by exercising good manufacturing discipline, which reduces wave reflection/scattering as signal transitions/edges propagate along the channel.

However, all "complex calculations" were done fifty-seventy years ago, and the net result is rather simple.

The characteristic impedance of a twisted wire (or planar waveguide, or coaxial cable) is a function of conductor geometry and dielectric constant of the insulator between the conductors. To make it 90 ohms differential two wires of certain core diameter and certain thickness of insulator are twisted together (at certain number of twists per meter). The twisting ensures that the distance between wires stays constant, because it is a major parameter that defines uniformity of the transmission line. Insulation thickness should be also well controlled, for the same reason. The number of twists per unit lenght is also important, but to a lesser degree.

Now, the insulation diameter defines the distance between two signal conductors, and electrical properties of this geometry were calculated long time ago. So a manufacturer of USB cable (I mean bulk cable) knows what diameter to pick, which kind of insulation to use, and how to twist. The final selection of wire parameters goes through experimentation, where the cable quality and parametrics is measured by special equipment. After which the raw cable goes into production, and sold to cable makers, who solder chunks of it to a pair of USB connectors, and encapsulate the ends by pressure molding.

So, USB signal wires do have a special treatment.

\$\endgroup\$
  • \$\begingroup\$ Thank you for this detailed answer. I only have two questions. I know what you mean when you talk about signal reflections but what do you mean by "scattering" I am not as familiar with that term. Also, what about the 45 ohm impedance to ground? In this case so long as the 90 ohm differential impedance is maintained is this single ended 45 to ground unnecessary/irrelevant? If not how is the impedance to ground of each wire held constant, I don't believe that's addressed by your words. Thanks. \$\endgroup\$ – scuba Aug 25 '17 at 4:34
  • 1
    \$\begingroup\$ @scuba, scattering means multitude of small reflection from line imperfections. At any given time there might be several different bits of data in-flight along the cable, each one will produce small reflections, which interpose. It is also called "Inter-symbol interference", and causes elevated jitter of signal edges at receiver point. The differentilal impedance is a vector sum of single-ended impedances (one wire to ground), and coupling between two wires. Two isolated 45-Ohm coax would form a 90-Ohm impedance, but one can have two tightly-coupled wires with high impedance to ground. \$\endgroup\$ – Ale..chenski Aug 25 '17 at 4:46
3
\$\begingroup\$

Impedance controlled Cables are designed to meet the data rates, impedance and tolerance. Impedance mismatch tolerance is measured in Return Loss, RL [dB] which is also called s11 , one of four scattering parameters for two-port devices.

RF Impedance for coax. is defined by the ratio of outer conductor to inner conductor factored by the relative dielectric constant. Similar complex ratios are used for twisted pair, microstrip and stripline. These ratios are fairly consistent and include the skin effects of conductors.

All cable suppliers must adhere to these geometrical design criteria and it's design is tested by return loss. It doesn't mean every manufacturer of every cable gets tested to these requirements in production or that every cable was engineered and verified with sophisticated methods, but good quality cables most verify the design and tolerances for manufacturability.

As data rates get higher, the flexible cable may get more rigid for better consistency.

As conductors get thinner or spaced further apart with more air gap between insulation, the impedance rises.

The differential signal impedance is usually the sum of each single ended impedance

This impedance term used here is called the "characteristic impedance" where the length cable length is approximately >=5% of the highest frequency wavelength in the signal spectrum.

Below this frequency, the cable inductance or capacitance is more significant for transient currents and high impedance circuits respectively and of course DC insulation resistance.

\$\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.