If a trace is designed to be 50 ohms, how much of an effect would 10% or 20% mismatch between line impedance and termination values ?

For instance, a 50 ohm line with 45 ohm or 55 ohm series termination ? or 50 ohm line with 40 ohm or 60 ohm series termination ?

How much wiggle typically available before you run into potential problems ?

  • 1
    \$\begingroup\$ I have many multi gigabit traces speced as +/- 10% if that helps. That's normal procedure for me. \$\endgroup\$ – Some Hardware Guy Apr 19 '15 at 23:58
  • \$\begingroup\$ @SomeHardwareGuy it does. That means I can probably increase my trace width so that maybe its 55 or 60ohms without too much worry then. \$\endgroup\$ – efox29 Apr 20 '15 at 0:03
  • \$\begingroup\$ @efox29, increasing trace width will decrease \$Z_0\$. \$\endgroup\$ – The Photon Apr 20 '15 at 0:28
  • \$\begingroup\$ @ThePhoton oops \$\endgroup\$ – efox29 Apr 20 '15 at 1:13

You can calculate the reflection coefficient from the well-known formula:

$$\Gamma = \frac{Z_L-Z_0}{Z_L+Z_0}$$

With this you can work out, for example, that a 10% error in the termination impedance gives about a 5% reflection.

So really, the question is, how much reflection can you tolerate and still achieve error-free data transfer. And that depends a lot on what other discontinuities exist in your signal path. In source-terminated schemes, for example, we allow near 100% reflections at the load end of the trace, knowing that the termination at the source end will prevent these reflections degrading the signal seen at the receiver.

To get a good idea how reflections affect your particular situation, you could do a simulation like what's proposed in Rolf Ostergaard's answer to a recent question about SPI.

Generally, just one discontinuity is not likely to cause problems. Multiple discontinuities are more likely to cause problems the farther apart they are on the line because they'll interact at lower frequencies.

a 50 ohm line with 45 ohm or 55 ohm series termination

Remember that without special attention (and added cost), your board shop will probably only promise a 10% tolerance on the trace characteristic impedance. So if your design calls for a 50 ohm trace, the actual \$Z_0\$ might end up anywhere between 45 and 55 ohms. So a termination that's 10% off of the nominal \$Z_0\$ might end up almost 20% off from the actual \$Z_0\$ (For example a 55 ohm termination would be about 20% mis-matched if the actual trace impedance comes out as 45 ohms.

At modest data rates (say, up to 1 GHz), it's generally easy to ensure that the trace impedance uncertainty is a bigger error than the load mismatch. For example, a 49.9 ohm +/- 1% resistor terminating a 50 ohm +/- 10% microstrip line.


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.