What causes these artifacts in my electrophysiology recording? Large power spikes at many frequencies

Question

I'm a biologist who does a fair amount of computational work, but has little electrical engineering experience, so please excuse my ignorance.

I'm recording electromyography (EMG) signals from muscles in moving rats for an experiment. The electric signal at the muscle is picked up by silver wires (which are insulated but not perfectly so). The signal is passed through a cable to an amplifier (which low-pass filters <1000 Hz I believe), and then to a digitizer and computer to be recorded at a sampling rate of 512 Hz. The Fourier transform of a 10-second segment of signal looks like this. As you can see, the signal has relatively even power at all frequencies.

However, there will occasionally be a great amount of noise in the signal. In the Fourier transform, you see large spikes of power at many frequencies: 50 Hz, 100 Hz, 150 Hz, 200 Hz, 250 Hz, but also smaller spikes at other frequencies.

And here is the same noisy signal, zoomed in on the Y-axis:

I figure that the multiples of 50 Hz must have something to do with the utility frequency (I'm in Europe), and maybe the other frequencies are due to interference of overtones or something. But I don't understand why the artifact comes and goes. Most animals have no problem at all, in some animals it appears and disappears, while in others it's present almost all the time. Browsing StackExchange, I've read something about voltage resonance, but I don't really understand it well or know if it's relevant.

Any ideas about the source of this problem? And about how I can prevent it?

Answer 1

Your guess of it being mains related is a good one.

It looks to me like the fundamental frequency is 150Hz - something like a mains powered motor would be my first guess.

If it's appearing & disappearing, it's probably something on a fairly low frequency control cycle - like temperature control.

Try turning your air-con off while taking a measurement, or look around the lab for a chiller unit, air pump, something of that sort.

A good tip is to stick your measuring wires on something inanimate, that can be moved around - see if those frequencies are still there, and if they get stronger/weaker as they move around the room. At that point it's just playing hot & cold to find the culprit.

Once you have found the source, either ensuring it is disabled while you are doing your measurements, or shielding it is possible.

Good luck!

Answer 2

If you're in Europe, I'd suggest noise from the mains, along with harmonics. You might also see rectified noise, form things like flourescent lights.

If you have dimmers on your flourescents, get rid of them, or better yet, lose all flourescents.

Note that your filter corner seems too high for your sample rate, so your noise might be aliased down to a lower frequency.

Answer 3

The nerve and muscle communication is done by chemical daisy-chain by axons (electrical conductors) insulated with a sheath for each then the bundle.

Steady muscle contractions are modulated by the intensity and repetition rate of these pulses. Thus what you see is the FFT spectrum of each pulse and the aggregate of variable pulse rates for signals.

Then for line harmonic noise, due to poor shielding and poor common mode rejection ratio CMRR for both conducted and radiated. Possibly you are seeing the effects of a power supply that pulse charges a bridge cap every cycle crossing and generating harmonics so you see 50,100,150 etc. This is normally not radiated by the grid or motors which prefer to generate currents with odd harmonics.

The nerve or muscle pulse looks surprisingly just like the heart muscle sympathetic pulse except much faster resulting in for each pulse as shown below, the mean centre is 400 Hz with a low Q bandwidth of 200 Hz.

The repetition rate will be variable with activity.

But yours appears to be dominant at 10 Hz. Any other spikes could be false readings.
We can't tell if the 40Hz is the IMD result of 50 Hz noise and 10 Hz rep rate, (40=50-10) or if the Rat muscle is working so hard, the axons are firing at 40 Hz rate maximum triggered by some stimulation.

e,g. If outside the CM range or from spike saturation, it could induce frequency mixing = intermodulation distortion (IMD) creating new harmonics of sum and difference values. The higher frequencies are also broadly spread from 50 to 300 Hz due the spectral shape of each pulse.. There are also possible IMD effects from the ADC not having a good anti-alias pre-filter. (aka Nyquist filter)

I have recently looked at this so I can share the spectrum of each pulse centre frequency and BW.

Now you can imagine variable pulse rates will spread out the spectrum at the low end and if the shape of the pulse changes it can spread out the spectrum of each pulse at the high end due to changes in rise time.

If you have specific goals in mind, please share exactly what you have done so we can point you in the right direction.

Like use EEG best capture practices for EMG. (these have to capture uV level pulses)

The cyclic line harmonic interference could be any device that is cyclic temperature controlled but due to your high unbalanced impedance there is mutual coupling (and thus poor CMRR).

Try Earth grounding a shield or sharing your photos of layout and schematics.