8

I have a PCB with tracks of no controlled impedance. The longest track is shorter than 1/5000 of a wavelength. Does the impedance of the track even matter?

If not, then at what length would I need to start thinking about matching the track impedance to the source and load impedances?

Nick Alexeev
  • 38,181
  • 17
  • 100
  • 237
  • 8
    Be sure you're using the correct wavelength. If a digital signal, it should be the wavelength corresponding to the rise time of an edge, not the clock period or data symbol rate. – pericynthion Jun 29 '17 at 22:10
  • The obvious answer is the one regarding transmission lines like all the answers below, but what about controlling trace impedance for precision DC circuits? Does anyone do that? I'd guess the series resistance is not a factor when compared to realistic input offset voltages. – DavidG25 Jun 30 '17 at 00:09
  • What are your actual risetimes? 1:5000 track:wavelength seems strange to me; even TTL has wavelengths of ~3m. – uint128_t Jun 30 '17 at 00:11
  • My actual risetimes are very slow. 2uS or so, and the signals that concern me are generally sinusoidal. –  Jun 30 '17 at 14:49
  • Thanks for the replies all! It looks like I'm on solid ground then (or at least a mostly unblemished plane layer). –  Jun 30 '17 at 14:52
  • The characteristic impedance isn't relevant at such low frequencies, but the capacitance and inductance of the traces might matter. – Evan Jul 07 '17 at 05:02

4 Answers4

10

I have a PCB with tracks of no controlled impedance. The longest track is shorter than 1/5000 of a wavelength. Does the impedance of the track even matter?

No it won't matter.

It starts to matter (as a rule of thumb) when the track (or wire) length becomes about one tenth of the wavelength of the highest frequency signal of importance.

If not, then at what length would I need to start thinking about matching the track impedance to the source and load impedances?

Well, not all scenarios like this require matching - for instance if you are designing a quarter wave impedance transformer you don't match on purpose. If, on the other hand, you are transmitting data then it makes complete sense to match the impedances to avoid reflections and the possibility of data corruptions.

Andy aka
  • 456,226
  • 28
  • 367
  • 807
  • It sounds like my understanding was correct. I've used those rules at much higher frequencies, but I was thinking that sometimes when you get far enough away from "normal" that other factors come into play. –  Jun 30 '17 at 14:50
  • So if I leave a transmitter open (no cable, no antenna) and transmit, then everything will be fine because the connector itself could be seen as a mismatched trace but it would be so short compared to the wavelength that it wouldn't matter? – pgibbons May 28 '20 at 14:28
  • It all depends on the transmitter of course - it may be designed to work into a load. – Andy aka May 28 '20 at 14:30
6

General rule of thumb: the impedance of the track begins to matter when its length is greater than 1/10th the wavelength.

The reason it doesn't matter when the track is short: although there will be reflections due to impedance mismatches at each end, those reflections can propagate across the line so fast relative to the rate at which the signal is changing that equilibrium is reached "instantly", or at least fast enough it's negligible in most cases.

Phil Frost
  • 57,310
  • 18
  • 145
  • 267
  • When you think about it, its kind of funny that this also applies to "DC" because it isn't really DC when you switch it on. – PlasmaHH Jun 30 '17 at 09:37
  • Definitely! :) Digital is a special case of analog. –  Jun 30 '17 at 14:50
3

A controlled impedance trace is the same as a transmission line.

A quote from electronicdesign.com:

A cable becomes a transmission line when it has a length greater than λ/8 at the operating frequency

So at 1/5000 of the wavelength you're still very far away from that point so a controlled impedance track would not even behave as a proper transmission line (at that frequency).

You don't have to impedance match, even if you are working with a transmission line. It depends on the behavior that you want. If you want good power transfer and little signal reflections then yes you need impedance matching.

Bimpelrekkie
  • 80,812
  • 2
  • 94
  • 185
0

Back in the old world odor, I was configuring 1553 data busses. The rule of thumb there is that anything less than 1/5 the pulse rise time is treated as a lumped impedance. Our stuff flew just fine.

richard1941
  • 670
  • 4
  • 7