Electronics Design AUScope showing 200+ mV spikes on my 3.3V rail — is this real or a probe problem?
Asked by fresh_grad_fern ·
Probing the 3.3V output of a switching regulator on a new board and I'm seeing large spikes on the scope that don't make sense to me.
The waveform shows the DC level sitting around 3.27V but with sharp spikes riding on top — maybe 200–300 mV peak-to-peak at what looks like the switching frequency, plus some higher-frequency content I can't clearly identify. It looks alarming.
Here's the confusing part: I also measured the same rail with my multimeter and it reads 3.27V, completely steady. No variation.
I know the multimeter is too slow to show high-frequency noise so that doesn't rule anything out. But I'm also not confident my scope setup is right. I'm using a 10× probe, ground clip attached to the alligator clip on my bench supply's chassis ground terminal, probing a via near the regulator output cap.
Scope is a Rigol DS1054Z. I've read the oscilloscope basics article and understand AC/DC coupling, but I genuinely don't know whether this is real switcher noise I need to fix or whether my probe setup is introducing it. How do you tell the difference?
3 Replies
The alligator clip on the bench supply chassis is your problem. This is the most common cause of exactly the symptoms you're describing.
What's actually happening
The ground lead on a standard 10× probe is typically 15–20 cm long. That lead, combined with the probe tip connection, forms a loop. That loop acts as an antenna — it picks up switching transients from the regulator, radiated fields from nearby traces, and any ambient RF in the environment. At switching frequencies and their harmonics, a long ground loop can inject 100 mV or more of noise into a scope capture that has nothing to do with what's on your actual rail. Routing it back to a chassis ground rather than a local PCB ground point makes this worse, not better.
How to separate measurement artefact from real noise
Shorten the ground connection. Most Rigol probes include a small spring-tip ground accessory in the accessories bag — it's a clip that fits over the probe barrel and makes a short-spring contact within a few millimetres of the probe tip. Dig it out and use that. If you can't find it, improvise: locate a GND via within a centimetre of the point you're probing, wrap the ground lead around the probe barrel and touch its end to that via. The goal is a ground loop smaller than your thumbnail.
If the spikes drop significantly, you've confirmed it was a measurement artefact. If they stay, you have a real noise problem to investigate.
What to expect on a real switcher output
Some ripple is normal. A well-designed buck converter's 3.3V output typically shows a few millivolts to a few tens of millivolts of switching ripple under load (exact figures depend on the switching frequency, inductor value, output capacitance, and ESR — check the regulator's datasheet for typical output ripple at your load current). Anything much above 50–100 mV after proper probing is worth investigating. If you're still seeing 200+ mV with a short ground connection, look at the output capacitor placement and value, and whether the layout has a tight input bypass capacitor right at the regulator.
Short the ground lead first. That one change resolves this specific symptom most of the time.
Spot on diagnosis. One additional technique for when the spring tip isn't enough or the geometry makes it hard to get close: the coaxial stub measurement.
Remove the probe from the BNC, connect a short length of 50Ω coaxial cable directly to the BNC. At the other end, strip back the outer braid about 10 mm to expose the centre conductor. Solder or clip the centre conductor to your measurement point and the braid to the nearest GND pad. You're now measuring with the coaxial return path, which has orders-of-magnitude lower loop inductance than a flying ground lead. This isn't terminated at 50Ω — at the low-MHz frequencies of most switchers, cable reflections in a short piece of coax are negligible, and you're using it purely for the ground inductance benefit.
The measurement improvement is often striking. On precision analog boards I've seen apparent "noise" of 100–200 mV vanish completely when going from a 15 cm clip lead to a 1 cm coaxial connection. That's how much the probe is adding, not the circuit.
If you're doing this regularly — power rail measurements on new designs — it's worth investing in a purpose-built power rail probe (several major scope makers offer them). They work on the same low-ground-inductance principle but in a convenient probe package. For one-off diagnosis, the coax technique costs nothing.
This exact scenario cost me about two hours on a board bring-up last year. Saw large spikes on a switcher rail, assumed it was real, spent the afternoon swapping capacitors and tweaking component values on the layout. Nothing moved the needle.
Eventually a colleague walked past, looked at my probe setup, bridged the ground clip to a GND via on the board about 3 mm from the probe tip with her finger. Noise dropped to nothing. Two hours wasted.
Now it's the first thing I check — before I assume anything about the circuit is wrong, I confirm my measurement is set up correctly. Shorten the ground lead. Look at the waveform again. Then decide if there's a real problem.
For what it's worth: the PCB bring-up checklist covers power rail checks as an early step and suggests scope + multimeter verification. The implicit assumption there is that you have a valid scope measurement — which requires the probe technique to be right first. Learned that the hard way.
Related Discussions
Decoupling caps placed right next to the IC but still seeing power rail noise
Fast digital board, MCU running a few hundred MHz core clock plus a couple of high-speed peripherals. Put 100 nF caps right next to every po
STM32 USB not detected by Windows after jumping to bootloader mode
Working on a custom STM32F411 board, trying to jump into the built-in USB DFU bootloader from application code instead of holding BOOT0 on p