0
\$\begingroup\$

I've been reading about ringing and overshooting and all sorts of things, but all without getting a real "feeling" for how likely these things are to affect my circuit.

I've got a chip generating an SPI signal, traveling about 5cm to a 3.3->5V level shifter, and then traveling ~35cm (yes, it's a very long PCB) to the chip that receives it. From what I understand, I shouldn't have issues with this: my traveled distance isn't even 1% of the wavelength at 1Mhz, which gives me hope.

What considerations SHOULD I have in mind when laying out the PCB to avoid issues with this signal?

\$\endgroup\$
  • 1
    \$\begingroup\$ In general, it's not the wavelength that matters as much as it is the rise and fall times. You could have a low frequency signal with very fast rise/fall times which could still result in ringing. With a lower frequency, you have more time for the signal to settle, but the ringing and overshoot can cause damage (such as in motor driver gate drives circuits where the frequencies are lower, but the fast rise/fall times can cause large enough ringing and overshoot/undershoot that damages things. \$\endgroup\$ – DKNguyen Mar 10 '19 at 22:59
  • \$\begingroup\$ With long SPI runs, remember that the clock must travel from and back to the master in at least as much time it takes for the data from the slave to reach the master. This limits SPI distance. \$\endgroup\$ – DKNguyen Mar 10 '19 at 23:02
  • \$\begingroup\$ @Toor Thanks for clarifying! Is there any way to estimate how short a rise/fall time would be potentially dangerous with the specs I've listed? The datasheet I'm looking at for my level shifter is mentioning propagation delay, but the only spec related to rise/fall time I've found is 1ns for 10%-90% rise or fall. Additionally, this is all one-directional, so I'm not worried about the signal getting back. Thanks again! \$\endgroup\$ – Helpful Mar 10 '19 at 23:11
  • \$\begingroup\$ TBH, I wouldn't even worry in your case. If you're worried leave room for series source termination resistors and just zero ohm jumper them. If you find things aren't reliable then scope it and if ther is an issue them put the resistors in. But you won't need them. \$\endgroup\$ – DKNguyen Mar 10 '19 at 23:24
  • 1
    \$\begingroup\$ SPI is 3-wire, and one of those is CLOCK. You don't want much ringing on the CLOCK, because you'll experience double or triple clocking. \$\endgroup\$ – analogsystemsrf Mar 11 '19 at 4:15
2
\$\begingroup\$

At 1MHz you have nothing to worry about. I've committed atrocities such as passing 8MHz SPI over 30cm ribbon cables with impunity (maybe I got lucky!)

When you go higher, say, 30MHz, you have to start worrying about controlled-impedance traces, matching lengths, not taking sharp turns or having vias, and so on, but you are nowhere near that operating regime.

\$\endgroup\$

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.