1
\$\begingroup\$

Is there any basic methods to calculate crosstalk between different layers in a multilayer PCB. I am not after anything exact just a way to grasp the magnitude of the issue.

The situation I am trying to solve is in a 4 layer space constrained PCB with a stackup of Signal-Ground-Power-Signal, on my bottom layer I have a number of differential +/- 10V Signals max freq 1KHz. On the power layer above these traces I have a power trace with a current varying between 0-200mA with a minimum rise time of approximately 0.1ms, the trace for the return current is directly adjacent.

Assuming there is no other way I can route my PCB how can I go about getting a feel for what kind of affect a step in my current can affect my signals.

Other data

Signal Output impedance <100 ohms

  • Signal Input impedance approx 10k ohms
  • Signal trace width 12mil, trace separation 12mil
  • Power trace width 40mil, separation 20mill
  • Copper Weight 1oz
  • Outer plane separation 0.2mm
  • Inner plane separation 1.2mm
  • Length of parallel run 75mm
\$\endgroup\$
6
  • \$\begingroup\$ check out the PCB Toolkit by Saturn PCB Design Inc. \$\endgroup\$
    – Techydude
    Mar 1, 2016 at 6:38
  • \$\begingroup\$ Why have I never run across this tool before, it looks pretty cool, any recommendations how I can get it to work for my stackup which is effectively two signals above a ground plane. \$\endgroup\$
    – Hugoagogo
    Mar 1, 2016 at 6:50
  • \$\begingroup\$ So are you saying you've got a current-carrying pair on the 'power'-inner layer, and a signal diff-pair on the bottom layer, and they're both running parallel to each other for 75mm? But what's the lateral separation between them? \$\endgroup\$
    – Techydude
    Mar 1, 2016 at 7:46
  • \$\begingroup\$ On top of each other I know it's bad that's why I am trying to evaluate this vs running wires. \$\endgroup\$
    – Hugoagogo
    Mar 1, 2016 at 11:36
  • \$\begingroup\$ I recommend Hyperlynx SI, though it is very expensive software there is a trial version. mentor.com/pcb/hyperlynx/signal-integrity \$\endgroup\$
    – DerStrom8
    Mar 1, 2016 at 13:22

1 Answer 1

1
\$\begingroup\$

As @derstrom8 suggested, there are better tools for doing this, but tend to be expensive & aimed at pros. I vaguely remember there is a FOSS field-solver tool out there somewhere, but can't remember. But if I interpret your PCB stack-up figures correctly, you may be able to approximate your situation with the Saturn tool I mentioned:

Saturn PCB Toolkit

If you set it to 'stripline' mode where you're calculating crosstalk of 2 tracks sandwiched within 2 planes, and then set the distance of the bottom plane effectively to infinity (500mm) because that doesn't exist in this scenario, then set H1 to the gap between your 2 inner planes (where the lower plane is your power track on the inner-bottom, right?), and H2 to the gap between bottom-layer & bottom-plane (what you describe sounds like a typical 4-layer stack-up), I see -84dB, or 1.25mV.

Perhaps with that rough idea you can use your input-impedance & knowledge of the broader system to work out how significant this may or may not be.

\$\endgroup\$
8
  • \$\begingroup\$ How can I relate this coupled voltage to an input impedance, what would the impedance of the coupling better. I assume the higher the input impedance the more effect it will have. \$\endgroup\$
    – Hugoagogo
    Mar 1, 2016 at 21:47
  • \$\begingroup\$ hang on - my answer isn't right at all! i mistook your +/-10V to be what the PSU is doing, but that's just your signal swing. So your PSU is switching 0-200mA, but what's its voltage? \$\endgroup\$
    – Techydude
    Mar 1, 2016 at 21:52
  • \$\begingroup\$ This particular traces is out of a current regulator, that will drive a signal somewhere between 0-200mA to a maximum of 15V. While not technically a power rail I though it should be separated out from my signal layers. \$\endgroup\$
    – Hugoagogo
    Mar 1, 2016 at 22:40
  • \$\begingroup\$ 15V - ok, so that's not much different to the 20V I put in based on your +/-10V anyway, so the result won't be hugely different, 0.94mV. \$\endgroup\$
    – Techydude
    Mar 2, 2016 at 2:04
  • \$\begingroup\$ So is crosstalk always a certain induced voltage, for some reason I thought of it as injecting a certain amount of power into the adjacent line, which then translates into certain voltages via the impedance. \$\endgroup\$
    – Hugoagogo
    Mar 2, 2016 at 2:06

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.