0
\$\begingroup\$

The title says the question.... I don't quite understand the need for a single-end impedance if I'm routing a differential pair. What I mean is, if I need to route a differential pair on a circuit board for a high speed signal (like USB), do I need to worry about the single-end impedance?

Let's take USB 2 as an example. In the USB 2 spec, there is a 90 Ohm differential requirement, but I don't see a single end requirement. Do I even need to worry about a single-end impedance for each track in the pair, or does the receiver only see the differential?

I ask because, in a previous question I asked on split termination with differential pairs, it seems like only the single-end impedance matters for termination, not the differential. The differential impedance requirement sets some spacing. But obviously, I can route two 45 Ohm traces as a 90 Ohm differential without very close spacing, or I could route two 50 Ohm traces as a 90 Ohm pair as long as the spacing is close enough. The USB 2 spec seems to imply that the single-ended impedance doesn't matter. Is this correct?

\$\endgroup\$

2 Answers 2

0
\$\begingroup\$

No, the single-ended impedance provided in differential line calculators does not matter.

Still, as you correctly analyzed it is much better to think of the individual lines in a differential pair as two length-matched single-ended lines, both with the half impedance.

USB 2 e.g. has two length-matched 45 Ohm lines. When you bring them close together, the impedance of a single line drops to below 45 Ohm, because the nearby conductor becomes an additional return path that is ignored in the single-ended microstrip figure. As a result, you need to narrow the trace to arrive at 45 Ohm actual single-ended impedance again. Basically the single-ended microstrip impedance in differential pair calculators is a fake value, and half the differential pair value would be the actual single-ended impedance.

Another consequence is the following: If you densely route several single-ended lines, I suggest you throw the values into a differential line calculator. Does the differential impedance drop to well under twice the single-ended impedance ? In that case, you have to increase the spacing to guarantee the single-ended impedance, because due to the close proximity of the other traces, the actual impedance of your line has dropped to half the displayed differential impedance.

\$\endgroup\$
7
  • \$\begingroup\$ It seems to me that those calculators which give a single-ended value are giving an accurate value for the single-ended impedance, but are you saying that number they give is only the impedance if that trace were totally isolated from any other trace? \$\endgroup\$
    – sparaps
    Oct 25, 2021 at 7:09
  • \$\begingroup\$ @sparaps Exactly. If you enter your geometry specs into a single-ended microstrip calculator you will get some value. Now if you enter the same values into the differential microstrip calculator, the obtained single-ended value will be identical to the former value, i.e. it assumes that the second trace were not present at all. \$\endgroup\$
    – tobalt
    Oct 25, 2021 at 7:11
  • \$\begingroup\$ I guess then this means the neighboring trace modifies its single-end impedance by acting as an additional return path as you say, which would make sense... However, it can't possibly be that the single-end impedance of just one of the pairs is exactly half the differential, or is that really the case? \$\endgroup\$
    – sparaps
    Oct 25, 2021 at 7:12
  • \$\begingroup\$ @sparaps Yes it is the case. As you assumed in your question, one can route USB2 with two widely distributed single-ended 45 Ohm traces if you follow length matching. The differential pair is slightly more space efficient than two random single-ended lines, though, because it makes sense to take eachother as return path, because no interference can be expected given the signals are mirror images of eachother. \$\endgroup\$
    – tobalt
    Oct 25, 2021 at 7:16
  • \$\begingroup\$ Ok thank you. I guess the takeaway is this: In a thick dielectric, the single ended width might be really big, so you have no choice but to do controlled impedance differential pairs to get to the smaller width you need to save space. But otherwise, if you could get to very thin single-end traces with thin dielectric, you could essentially route as either single-end or differential because the nearby ground causes the two trace widths to be very similar, and at that point the isolated single-end trace width ends up being about the same as the trace width when it is part of a pair... \$\endgroup\$
    – sparaps
    Oct 25, 2021 at 7:24
0
\$\begingroup\$

I'ts true that the common mode impedance is not defined for the devices, but for the cable it is, 30 ohms +/- 30%.

This is why many USB application notes from different manufacturers say that the common mode impedance should be 30 ohms (21 to 39) which makes the single ended impedance 60 ohms (42 to 78 ohms). Also some of these application notes say that to create a differential pair with 90 ohms differential impedance, the single ended impedance should be larger than 45 ohms.

\$\endgroup\$
1
  • \$\begingroup\$ wouldn't the common mode impedance also be more than half the single-ended impedance for a closely spaces conductor pair ? \$\endgroup\$
    – tobalt
    Oct 25, 2021 at 7:32

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.