2
\$\begingroup\$

I want to sniff USB3 and Ethernet diff lines without affecting communication on my PCB. But I couldn't figure out how to do that with a small form factor.

\$\endgroup\$
7
  • 2
    \$\begingroup\$ are you able to do it with large form factor? \$\endgroup\$
    – jsotola
    Commented Jul 21, 2022 at 15:20
  • 1
    \$\begingroup\$ Can you sniff USB3 and Ethernet signals in general? At what speeds, and at what level must you sniff them? The level gives an approximate ballpark if you need equiment that costs approximately as much as a computer, a car, or a house. \$\endgroup\$
    – Justme
    Commented Jul 21, 2022 at 15:23
  • 1
    \$\begingroup\$ do you already have the necessary high-speed signal / logic analyzer, and this is about probing (if so, what do you have?), or do you intend to build the sniffer yourself? \$\endgroup\$ Commented Jul 21, 2022 at 15:30
  • \$\begingroup\$ @jsotola I mean equiments \$\endgroup\$
    – devscg
    Commented Aug 16, 2022 at 21:07
  • \$\begingroup\$ @Justme 1gbit ethernet and 5gbps USB signals. \$\endgroup\$
    – devscg
    Commented Aug 16, 2022 at 21:09

1 Answer 1

1
\$\begingroup\$

Inductive pickup is one way to do it. It may introduce less of a discontinuity than a differential amplifier input would. Back when I was doing it, I did need the galvanically isolated pickup. You may not need it though. The signal degradation due to a galvanically connected high speed diff-amp may well be acceptable with modern parts. Decades ago, in my application, it wasn’t the case.

So, you could try with a pair of pogo pins touching exposed traces, and a fast diff—amp.

I’ve made such things in the past - like 30 years ago - and it was quite a bit of work to get usable signals. Maybe I was going the hard way about it, but in the end it worked.

I used a tiny inductive pickup – a solenoid with its axis perpendicular to the differential pair run. It was mounted very symmetrically on top of the gap between the differential traces. This was done under a microscope with an XY stage to position it prior to gluing.

The coils needed were tiny - grain-of-sand sizes. IIRC would wind 50AWG wire on a 40AWG “core”. That takes some custom tooling as the wire just rips if you try to handle it manually. It was like watchmaking on steroids – not very practical but work it did.

Right next to the solenoid was a high speed amplifier/buffer. I’d end up needing custom emphasis/“cleanup” after the preamp, to bring the signal back to a form that a PHY was able to deal with.

These days I’d try with a magnetoresistive head from a hard drive as a pickup. They have the feature sizes that should work well for this, and are made to handle high bandwidth signals. They also have ready-to-use analog preamplifiers close by. Those may be reusable as-is. The head gap would be oriented along the differential pair, centered between the traces. For highest signal there may be a need for slight shimming – or de-shimming.

If you are laying out the PCB to facilitate such probing, make a window in the soldermask at the probe points. The shim thickness may need to be smaller than the soldermask thickness. Non-conductive shims are probably best, although you’d need to experiment. Thin Mylar films would be probably a good shim material. Tweak the shim thickness for maximum signal amplitude.

Another way to shim that is to leverage the smooth surface of the head and put a couple of very small non-plated through holes centered between the differential pair conductors. Those would make a slight discontinuity, but would let you push air at adjustable pressure/flow from the opposite side of the board, and use the head float as a shim adjustment.

\$\endgroup\$
1
  • \$\begingroup\$ Thank you so much for your answer. Do you think an RF splitter would achieve this? \$\endgroup\$
    – devscg
    Commented Aug 16, 2022 at 21:11

Your Answer

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

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