Bruce Carsten's 'EMI Sniffer Probe'

If you're looking for EMI emitters, why shield the coil?

I sometimes hang a short piece of wire on the end of a coax, into a spectrum analyzer. For mag fields, a cheap unshielded drum core inductor works great.

Friday, trying to find my jitter problem, just waving a scope probe around was all I needed. I have two dc/dc converters on my board, and therorized that one or both were getting into my nearby timing ramps. Parking the probe above each converter showed its switching frequency.

Once I figured out the frequency of each converter, I varied the trigger rate into the box. At some cardinal rates, I could see the trigger rate heterodyning the jitter pattern. Neither dc/dc switch rate corresponded. The thing that did correspond was the oscillation frequency of an LDO.

Sometimes a good fet probe doesn't even need to touch a signal. Just get close.

Electronics presents us with riddles to solve.

To block the electrical field--you're hunting currents, not capacitively-coupled voltages.

I do that too. Also, for mixed hash, trigger on the magic wand probe, while viewing the hash on another channel. That freezes the various hash components selectively, one-by-one.

Yep. I'm intrigued with air-core power converters at the moment, hence the appeal of B-field probes.

Cheers, James

This is my mag probe. At the end of the semi-rigid belongs a 0402 or 0603 pickup coil which is broken off. The brass tube shields electrostatically. It has a slot at the tip.

The alu box has nothing to do with it, it's just that the stuff does not roll away. It's for my new 12 GHz trigger prescaler in statu nascendi for the sampling scope.

I don't remember where I got the idea from. Probably from EDN back in time when men were men and EDN was a thumb thick bundle of interesting things.

Fun. Maybe make a transformer from really thin kapton flex.

HP once sold a "modulation domain analyzer." They had a cool 3d diagram in an ad that showed a scope on one axis - voltage vs time - and a spectrum analyzer on another axis - voltage vs frequency - and proposed a new instrument on the 3rd axis that measured frequency vs time, or maybe it was time vs time. They sold some but I don't think it was popular.

I often see, say, time or period jitter and wonder what the source is. If the time error vs time showed a 60 Hz component, I'd know where to look. Imagine a spectrum analysis of the discriminated time error. The telecom people care about this.

A digital oscilloscope should be able to display time error vs time, but I don't know of one that does.

I'm seeing maybe 10 ps RMS jitter on my 25 ns range, just about an inch from the two dc/dc converters. But that jitter includes my DDG and my old Tek sampling scope, so it's actually less. One rough r.o.t. is that a really good analog ramp has 1 part in 20,000 RMS jitter.

If I can commandeer some muscle, and clear off a square yard of bench space, I'll drag up the monster LeCroy scope and get a better jitter measurement.

Nice probe! I feel a bit silly I've gone all this time without making a nice one.

Since posting, I wound two coils, ten close-wound turns of #32 wire each. One was formed on a 0.058" (1,5 mm) mandrel, the other on a 0.082" (2,1 mm) diameter mandrel.

They look very promising. I think they'll offer much better spacial resolution than anything I've done before, plus add useful directivity.

A ferrite rod, if I can manage to machine one that small, could improve the resolution and let me get the same signal from fewer turns, raising the self-resonant frequency in the bargain.

Cheers, James Arthur

You can build it for 10 USD as shown on EEVblog

I just bought a 5 set version for 30 USD at Aliexpress, since I am lazy:

I can beat that--I was too lazy to order. I wound my coils on-the-spot from buss wire for $0.00 previously, faster than you could click "buy."

But my ad hoc coils never had anything close to the spacial-resolution or directivity of Carsten's 10-turn x 1,5mm-diameter shielded coil.

I watched Lasse's EEVblog link, where Dave makes a single-turn big-loop shielded pick up coil from semi-rigid coax, then goes hunting for gross noise-making areas with a spectrum analyzer.

That's a fine 'finger-in-the-wind,' but that's not Carsten's idea-- Carsten's pencil-lead-diameter 'microscope' allows zooming in on individual component leads and PCB traces, and qualitatively non-contact probing the time-domain fields and currents.

(When I posted I was considering building a Carsten Sniffer from semi-rigid coax, I meant 'locating the coil inside the coax shield,' not 'making coax-loops.') .-------. .----| | =+====================================| | | | .------------. #######| | | '-|10-turn coil|---[50 ohms]---------| | BNC | '------------' #######| | | ======================================| | | \ '----| | slit shield '-------'

And, Carsten's ten turns produce nearly ten times the signal of a single turn, +20dB with no amplifier needed.

Cheers, James Arthur

Thanks James... I need to re-read the app note to get all the details. I've made much cruder things with a few turns of stiff magnet wire. (say 20 AWG) and my x10 'scope probe.

George H.

These guys put 35W through a 10mm coreless PCB-printed transformer:

That's cool. Cheers, James