Strange interference pickup

[solarstandard]

Hey,

I have a problem with a device that I believe picks up wireless interference at levels up to 100 mV P-P. I can reproduce the interference signal by simply connecting a BNC cable with a small antenna (Fig. 1, see attachment) loop to an oscilloscope (Tektronix DPO7054, 500 MHz bandwidth, 50 Ohms termination). The signal is a periodic train of pulses at ~93 kHz (Fig. 2) and each pulse has a duration of ~200 ns (Fig. 3). I don't think I'm measuring the true form of the pulse because the oscillation pattern and frequency depends on the length of the cable (compare Fig. 3 with 14" BNC cable and Fig. 4 with 71" BNC cable, same antenna loop). The most puzzling fact is that I see the signal even in different buildings, few hundred meters apart and with different oscilloscopes. Has anyone seen anything like this before? What causes this interference and how do I get rid of it?

Now I know you can see all sorts of things by waving an unshielded antenna loop in the air, but the amplitude of the interference is quite big and I can't get a piece of equipment work because of it. I now this is not powerline noise or a ground loop (ran the test from an isolated UPS). I have tried putting the device in a shielded enclosure and I'm working on an active low-pas filter to suppress the noise at the input and everything seems to help bit by bit. Currently I'm down to 40 mV P-P, but I need to bring the interference amplitude down to <1 mV. And I would love to understand what's going on.

[T3sl4co1l]

Looks like a converter running at 93kHz with a poorly shielded 110MHz spur -- hunt around and try ferrite beads on any adapters you can find?  (Could be neighbors too; turn off everything and check again.)

Also seems to be a ~300kHz converter there, not quite as strong or nearby.

Real question is, why are you trying to probe 500MHz bandwidth signals with unshielded cables / probes / loops of wire?

If your signals are not 500MHz, turn down the freaking bandwidth!  Obviously, you'll need it under 100MHz (usually 20MHz is a selectable option on the scope itself) to avoid what's shown here.

If your signals aren't on loops of wire, don't use loops of wire!  Use coaxial probing techniques.  Among other resources, Linear Tech's AN-47 is brief yet comprehensive.

You can pull the ground clip and grabber tip off a common 10x probe and use it coaxially.  Make a spring contact for it, and solder that into the circuit.  Or, use the "no probe 10x" method (i.e., 50 ohm cable plus a 450 ohm series resistor to the node under test).  Both are better options than a beefy (terminated!) 50 ohm cable, which is quite a heavy load indeed.  Even the "no probe 10x" is a lot of load for many circuits.  For very sensitive circuits, you have little choice but to design buffers directly into the circuit itself -- or get a JFET probe ($$$).

Tim

[solarstandard]

Thanks for the detailed response, Tim. I will try reducing the bandwidth of the oscilloscope to test if that will suppress the signal, that's a great tip!

Tim, could you also clarify a few things:

a) When you say "converter running at 93 kHz" you mean a switched mode PSU? Would you expect these to generate so much EMI? Or would this indicate a faulty PSU?

b) Could you explain in more detail what is a "poorly shielded spur"? And why is it at 100 MHz? As you can see from the Figs. 3 and 4 the spur frequency depends on the cable length.

c) In general how do you what sort of signal this is? Is there a "Interference Waveforms Handbook" out there or is it all through experience only?

I will expand on a few points:

1. This is all happening in a research lab on university campus, so we do get "neighbors" doing all kinds of unimaginable stuff that can cause interference, but I have went to a different building and could still see the same signal, so I don't think it is a location issue (at least not on a small scale).

2. I am using the antenna loop to show that the interference is aerial, i.e. it is not somehow unique to the device that is giving me trouble. I get pretty much the same signal when I connect a shielded BNC cable to the device, which is a presumably well-working commercial board in a home-made metal enclosure plus the few runs of 5 cm wire to connect the PCB to the front panel terminals. All connections are grounded and the signal path and the board are shielded.

3. The enclosure does have quite a few perforated vent holes and also the panels don't mate that well, so I'm working on a new enclosure with EMI immunity in mind.

4. I am also working on an active low pass filter to cut down the acquisition bandwidth to a few hundred kHz, I use the 500 MHz scope to capture the interference signal to show here on the forum.

5. The device inputs and outputs are 15 kOhms differential and I believe that makes them quite a bit more susceptible to interference and they get overloaded easily (especially by 50 Ohm loads). I think an active filter can also act as a buffer between the unit and 50 Ohm input impedance ADC.

6. I hope a new enclosure and a filter will do the trick, but in general I would like to understand the origin of the interference. As I have mentioned I can go to a different building and still see the signal, even when using a different oscilloscope altogether!

Thanks again for your help!

[T3sl4co1l]

a) When you say "converter running at 93 kHz" you mean a switched mode PSU? Would you expect these to generate so much EMI? Or would this indicate a faulty PSU?

It may or may not be meeting FCC Part 15 (Unintentional Radiator) regulations, but yes, a switching supply normally makes a lot of crap.  The rising edge is around 300V tall, so if you're only seeing a fraction of a volt, that's still a lot of attenuation (80dB+).  That most things are as quiet as they are is fairly remarkable.

The pulse is also heavily filtered: you see how it's ringing, like a wavelet, rising and falling gradually?  It's not just a single bump, from some switching edge coupling into the air.  Some of that ringing may be due to the way the scope is reading it (many traces, superimposed together, coincident only because of the trigger -- easy way to tell, compare to a single trigger capture), but it's quite possible the ringing is due to a particularly unlucky length of cable.  Or series of cables, even (accidental antennas, and tuning, oh my!).

At 110MHz, the electrical length is around 8 feet (= 2.2 feet for a 1/4 wave whip, or 6.7 feet for a 3/4 wave), quite reasonable for charger cords and the like.  Something like that could be resonating -- and radiating.

b) Could you explain in more detail what is a "poorly shielded spur"? And why is it at 100 MHz? As you can see from the Figs. 3 and 4 the spur frequency depends on the cable length.

"Spur" refers to "spur"ious spikes seen on the frequency spectrum (try an FFT of that waveform).  Or, I suppose clusters of spikes might be called 'spurs' by rough analogy to the eponymous star-shaped wheels.

Of course, a time waveform isn't a spectrum, so I'm inferring the spectral content from what I see (once you learn Fourier transforms, it's easier to see the duality between rise and fall, ringing, bandwidth and spectrum).

Anyway, harmonics and crap are supposed to be kept inside the switcher.  If they didn't use enough shielding or filtering (because hey, 20 cents counts, right? right?), it leaks out, and you see stuff like this.

c) In general how do you what sort of signal this is? Is there a "Interference Waveforms Handbook" out there or is it all through experience only?

Hmm, no idea.  Maybe there should be?

I know from experience of course, but one could make some rough guesses on the origin, based on theory.  If nothing else, it's something pulsed.  It could well be exactly what it looks like (pulsed RF), but that's where you need experience (or a handbook?) to tell.

I will expand on a few points:

1. This is all happening in a research lab on university campus, so we do get "neighbors" doing all kinds of unimaginable stuff that can cause interference, but I have went to a different building and could still see the same signal, so I don't think it is a location issue (at least not on a small scale).

Ok, so it could be lab equipment, lighting, computers, students' counterfeit "Apple" phone chargers, etc....

2. I am using the antenna loop to show that the interference is aerial, i.e. it is not somehow unique to the device that is giving me trouble. I get pretty much the same signal when I connect a shielded BNC cable to the device, which is a presumably well-working commercial board in a home-made metal enclosure plus the few runs of 5 cm wire to connect the PCB to the front panel terminals. All connections are grounded and the signal path and the board are shielded.

6. I hope a new enclosure and a filter will do the trick, but in general I would like to understand the origin of the interference. As I have mentioned I can go to a different building and still see the signal, even when using a different oscilloscope altogether!

No idea what else is in the box, but... not necessarily a guarantee that outside crap stays out.

Could be something systematic, like the lights.  Those are likely the same across campus.

Or VFDs (motor drives) in the ME labs (if applicable) or physical plant in the building(s).  The noise from which could potentially propagate a long ways along on-campus wiring, though I'd be surprised if varying power line conduits and cabling was all that great at 100MHz over that kind of distance.  The repeat rate also seems high for a VFD (they're usually in the low 10s of kHz, since motors are slow like that).

Could also be the scope.  Tek wouldn't be the first to have products with endogenous noise problems (PSU, LCD display, etc.).

It seems unlikely to me that such an intermittent, periodic and spiky signal would well and truely be a long distance propagating signal (surely it would be interfering with radios by now, and generating complaints).  But I will add this:

Where I went to school, the campus radio station is on top of one dorm building.  FM band (88-108MHz), I think 100W.  A block away in the EECS lab, we had only unterminated banana jack leads for the scopes (yeeech!).  Measuring anything the least bit sensitive was impossible: even on 20MHz BW, the attenuation at 100MHz wasn't strong enough to eliminate that stupid radio station -- every trace was hairy to the tune of 10s of mV.  Even when I requested proper 10x probes from the help desk, it didn't help much.

It wasn't just CW either -- due to reflections and interference (multipath), the different frequencies in the FM carrier interfere differently in space, and therefore the amplitude (envelope) varies with the audio.

There were more than a few students who brought powered speakers for their computers / game consoles / whatever, which demodulated and amplified the station, loud and clear.  My own amplifier had the same problem, which I solved easily with some ferrite beads and ceramic caps (reducing bandwidth for the win).  Don't think anyone else bothered to (or thought of it, or had the parts to do it?).

Oh yeah, I forgot to mention my attachment... when I'm doing low level measurements, besides using terminated coax (see the tee -- although my scope has an internal termination option, so I don't actually need that), it can help to put a nice ferrite bead on the cables, which reduces ground currents that effectively get added to the signal, and hence dirty up your trace.

I have these nice big ferrite cores handy, so I can put a couple turns on there (every time it passes through the center is a turn, so that's 4 turns) and get a nice wad of impedance, reducing the ground currents.

Tim

[solarstandard]

Thanks again for your comments. It gave me some thoughts and solutions to try out. At least now I have some sort of reassurance that interference signals like these aren't just black magic.

[T3sl4co1l]

Thanks again for your comments. It gave me some thoughts and solutions to try out. At least now I have some sort of reassurance that interference signals like these aren't just black magic.

Hey, I never said they aren't but it is tractable, at least.

Tim