4
\$\begingroup\$

I am chasing the cause of some weird microcontroller resetting behaviour in a prototype robot. As part of the debugging, I decided to look at the stability of the power supply for the MCU.

I soldered some micro coax across one of the 3.3v capacitors on the MCU's PCB, and connected another single wire to the grounded chassis of the robot. I connected the ground wire to the ground of the differential probe on our MSO-X3104T oscilloscope, and connected the centre and shield of the coax to the + and - sides of the probe.

When the robot is powered up and sitting still, everything looks good on the scope. But when the robot starts to move, we see occasional events like this:

Weird oscilloscope behaviour

It's hard to see from the photo, but the horizontal scale is 50ns/div, which means that the rise and fall time of this signal is in the order of 5ns. The voltage seems to touch 0v at a couple of points.

I wanted to know if we were seeing a real phenomenon, or if it was something to do with the way we were taking the measurement, so next I soldered both sides of the micro coax to the 3.3v power rail. Now I really should see a flat line on the scope. However, this is what we got:

Weird oscilloscope behaviour

Again, we can see the same fast edges, but now it looks like quite a pure ring of about 100MHz. These events happen randomly every few seconds.

Questions:

  1. How can the differential probe be picking up a signal if the ends are shorted together? I have been careful to use coaxial cable to make sure there's no loop to pick up any radiated energy.
  2. If there was a strong common mode signal, would the differential probe see something?
  3. Is there a better way to measure the power supply stability?
  4. What could be causing these signals? Static? EMI?

Update: I've changed the probe to a single ended one, with the shield connected to Gnd, and the core connected to 3.3v. Now there are a lot fewer events, and I suspect these are more real.

\$\endgroup\$
6
  • 4
    \$\begingroup\$ Could be common-mode, yeah. Could also be the body of the differential probe picking up noise (that is, not the wires and stuff, but the amplifier inside the probe). Motors do tend to generate a lot of noise. Try shielding the probe body with some foil and see if that changes anything? (careful not to short the shield of the coax to the probe output) \$\endgroup\$
    – Hearth
    Dec 19, 2022 at 13:35
  • 1
    \$\begingroup\$ Also, what probe? \$\endgroup\$
    – hobbs
    Dec 19, 2022 at 13:36
  • \$\begingroup\$ Does the probe still show this noise if you short the coax, turn on the robot, but leave the probe end floating relative to the robot? \$\endgroup\$
    – Hearth
    Dec 19, 2022 at 13:37
  • \$\begingroup\$ What is the value and type of capacitor you probed? I=C dv/dt, so for example, neglecting ESR, 100nF*3V/5ns implies a current of 60 Amps which is not going to happen on a microcontroller's power supply, and should give a hint that it's not measuring what you think it's measuring. If it's a high-ESR cap then maybe. In any case something is making a lot of noise. \$\endgroup\$
    – bobflux
    Dec 19, 2022 at 13:39
  • \$\begingroup\$ If you're interested in the 3V3 supply you can solder a coax to the cap, shield to ground, 50R SMD resistor in series with +3V3. MELF resistors are best for this because they don't break when doing this type of stunts like chip resistors do. You can also use a thru-hole resistor. Then you will get a better signal. If you do not enable 50R termination in the scope it will be inaccurate, but if you enable it it will draw power from your 3V3 and you MUST check the maximum DC voltage allowed in the scope manual for 50R termination to avoid damaging the scope. \$\endgroup\$
    – bobflux
    Dec 19, 2022 at 13:42

2 Answers 2

5
\$\begingroup\$

I soldered some micro coax

This could be the source of your problem. Although oscilloscope probes may look like they connect to the oscilloscope with standard coax, in fact the coax is very non-standard. This article by Doug Ford goes into the details, but in short, the coax used with oscilloscope probes is designed to dampen reflections in the cable. If you substitute regular coax and feed it into the probe connector, you will probably get such reflections or ringing. You won't get them if you connect 50\$\Omega\$ coax into a scope's 50\$\Omega\$ input.

I don't know if this is causing your problem, but it is something to evaluate.

\$\endgroup\$
2
\$\begingroup\$

Why am I seeing a signal on a shorted differential probe?

The signal then is common-mode, by definition. You need to use an isolated battery-powered scope to get rid of that, or an actual isolated differential probe. Those cost real money for a reason: they are very useful tools that save a lot of debug time, and they are not very cheap to make, although that's a bit of a chicken-and-egg problem. Too few people use them for the prices to plummet :)

Also, if the "probe" is not terminated in the charasteristic impedance of the cable on both ends, it's mostly garbage in - garbage out situation, where it's hard to tell artifacts from events you're after.

\$\endgroup\$

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.