The USB 2.0 specs requires a differential impedance of 90 Ohms and a common mode impedance of 30 Ohms (with some tolerance). This, by definition, gives a target of an odd mode impedance of 45 Ohms and an even mode of 60 Ohms.

I have already some experience on single ended controlled impedance but this is my first differential design.

For cost reasons, I can't do an impedance controlled board (at least this time I got a 4 layer board!) and I'm stuck with the 'house' stackup:


Notice that everything is 10% and epsilon is stated at 1 MHz, though I haven't found the specific datasheet of that material.

With the 4.5 nominal prepreg value I got a plausible starting configuration: 0.45 mm width and 0.25 mm spacing (easy to manufacture, too).

However, I tried with different calculators and they gave me quite different answers:

EEweb coupled microstrip: Zodd = 52.68, Zeven = 48.79

Wcalc: Zodd = 45.45, Zeven = 59.03

kicad calculator: Zodd = 44.63, Zeven = 59.65

atlc: Zodd = 41.84, Zeven = 55.26 (using 3 as epsilon for the mask)

I know that all of these use different approximations (atlc is a field solver) but it seems a little too wide. The odd mode of 41.84 to 52.68 gives a Zdiff of 83.68 to 105.36 which is outside the mandated value of 76.5 to 103.5 (OK, I could try lowering all of them a little). And this is before the 10% material tolerance and etching tolerance are applied.

Is this a common situation?

I'll make the lines as short as possible (and probably the connectors and components will make a noticeable mismatch) but in practice how much it affects the workability of the design?

  • 2
    \$\begingroup\$ It depends on a lot of factors, like how long are the PCB tracks and cable before it. With short tracks the effect may not be that much, but the real problem is that if the values are not within USB specs then it's not an USB device and can't be sold as such. \$\endgroup\$
    – Justme
    Jan 20, 2022 at 8:30
  • \$\begingroup\$ In practice, not hugely unless you're routinely using it on the end of a max-length 5m cable. I've done USB over FFC before, which definitely wasn't controlled impedance but nor was it 5m long. \$\endgroup\$
    – pjc50
    Jan 20, 2022 at 9:56
  • 1
    \$\begingroup\$ @Justme when people ask you for the cheapest possible board do you expect they would pay for USB conformance testing? :D I want at least that on paper it's sound and, of course, that it works on the field \$\endgroup\$ Jan 21, 2022 at 7:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.