Do you have tips for how to get rid of mains hum EMI in hi impedance preamp

[Balthazar]

Hi

I’m a hobbyist who just finished making an electric cello, for reference I asked an earlier question about the project here. It’s based on polymer film piezo pickups, which are known for flat frequency response, and susceptibility to noise, compared to ceramics. And noise I got. The noise is completely dominated by 50-cycle hum (I live in Europe), so much so that when I digitally remove the 50 Hz frequency from a recorded signal, there is no audible noise at all left. I get about 20 dB SNR with the hum, which is far too low.

Polymer film piezo sensors consist of two metal plates, in my case 1.5 square cm and 28 um apart, with piezoelectric material in between. They can be viewed as a voltage source in series with a 480 pF capacitor.

The preamp is in a cavity lined with copper shielding tape, and experiments show that almost all the noise is picked up by the sensor, which isn’t yet shielded. So, I reckon I have two paths forward:

1) Sandwich the sensor in copper tape,
2) Make a differential preamp

Number one is simple, but it might affect the audio quality as well. Besides, I thought this might be an opportunity to learn an understand EMI a little better. So, do you think this might be a good candidate for a differential preamp?

At first I had the clueless idea I would make a "humbucker", which we from the guitar community are familiar with, where two pickups receive signal in phase, and noise out of phase by 180 degrees. So, by wiring two piezo transducers with opposite polarity, I thought the noise would be cancelled out. Since they were placed on opposite sides of a bridge rocking back and forth, the signal would still be in phase. It looks like I was correct that the signal was in phase, as two sensors gave a signal about 6 dB stronger than one, but the noise didn’t cancel, and the sound was a bit more focused with only one sensor on one side of the bridge, so I went for a single sensor.

So it seems I must change my mental model about how this works. If the noise was a voltage generated across the two metal plates of a transducer, as I imagined it, then surely the polarity of the wiring would affect the polarity of the noise picked up. So maybe I should instead view the two plates as two independent antennas, capable of picking up potentially the same noise? If they pick up an identical, or near identical noise signal, then surely a differential preamp might be the way to go.

The current wire from sensor to preamp is a single shielded conductor, with ground connected to one plate, and hot connected to the other. I would then replace it with a twisted shielded pair, with each plate connected to a wire in the pair, and would have to make an instrumentation amplifier like the one at the bottom here to keep input impedance high. Do you think that might be a good idea to try?

[Andrew_K]

1) Hard to say. Since they pick up physical vibrations in the instrument, attaching more material absolutely could change the response.

Do you have pictures of the piezo pickups in your instrument?

Is the connection balanced or unbalanced? 2 wire or 3 wire?

Have schematics of the humbucker you tried? (And what it looks like now?)

2) Might work. But just like the humbucker, it requires equal amounts of noise to be picked up to properly cancel. CMRR only works when the signal is truly common.

If you go the differential pre-amp route, use an off-the-shelf instrumentation amp. The matching and input impedance will be much better.

[PartialDischarge]

Put a resistor (100k, not more) in parallel with the piezo. If that doesn’t work you’ll have to shield it

[magic]

There is a fractional-pF capacitor from mains wiring around you to each cable in your circuit with 230V RMS at the mains side.

If you ground one side of the piezo, a tiny current flows through that capacitor to ground as it charges and discharges and basically nothing interesting happens. If the other side of the piezo is only grounded by 480pF and the preamp input is effectively open circuit / some megaohms, then a capacitive divider forms and some fraction of mains voltage appears at this side of the piezo.

You could shield the high impedance wire (or just use coax, like guitar cables) and maybe put a grounded shield around the piezo. It doesn't need to be very close to it, just surrounding it from all sides. Simply putting your hand over this area while touching a mains ground prong may bring improvement. Note that shielding adds capacitance to ground, which makes a capacitive divider for the signal. (Also note that every two conductor cable has capacitance anyway).

If you un-ground the piezo and connect to a preamp with high-Z differential input using two identical wires close together, both sides will float equally and CMRR should improve things.

[Balthazar]

1) Hard to say. Since they pick up physical vibrations in the instrument, attaching more material absolutely could change the response.

Exactly. The pickups are embedded in a thin plastic film to protect it, but I worry that if I stick copper tape on both sides of it, it may be hard to get that tape off without damaging the pickup.

Do you have pictures of the piezo pickups in your instrument?

They aren't visible when mounted. I have a picture of one not mounted here:

Is the connection balanced or unbalanced? 2 wire or 3 wire?

The connection from pickup to preamp is unbalanced. One conductor with shielding, 5-6 cm. The shielding is soldered to one terminal at the pickup and grounded. The conductor is connected to the other terminal. This is what I consider changing to a balanced connection, with a shielded twisted pair.

The connection from preamp to audio equipment (audio interface, amplifier etc...) is unbalanced with a 1/4" jack. I have also been considering changing it to a balanced XLR cable. Then I would only need to amplify a balanced signal

Have schematics of the humbucker you tried? (And what it looks like now?)

Right here:



I drew the pickups in dashed rectangles, with equivalent ideal-component circuit inside. The sound quality was improved simply by removing one of the pickups. They are wired with opposite polarity, but because they were on each side of a cello bridge rocking back and forth with the vibrating string, one would be compressed when the other was expanded. Therefore the (wanted) signals were in phase and added up, and experiments showed that part to work. However the noise didn't cancel, so clearly reversing the polarity of the wiring didn't reverse the polarity of hum picked up.

The 33M resistor is a little higher than it needs to be, but if I drop it too much bass response will be lacking. I had gone for the higher value because counterintuitively it reduces thermal noise (thermal noise is shunted by the pickups), but then of course I didn't know that noise would be completely dominated by mains hum.

If you go the differential pre-amp route, use an off-the-shelf instrumentation amp. The matching and input impedance will be much better.

Thanks for the tip. Unfortunately I have made a few too many orders from my electronics supplier where I order parts for like 2 euros and then pay 14 euros for shipping, so I was hoping to make it work with the opamps I already have.

[Balthazar]

There is a fractional-pF capacitor from mains wiring around you to each cable in your circuit with 230V RMS at the mains side.

If you ground one side of the piezo, a tiny current flows through that capacitor to ground as it charges and discharges and basically nothing interesting happens. If the other side of the piezo is only grounded by 480pF and the preamp input is effectively open circuit / some megaohms, then a capacitive divider forms and some fraction of mains voltage appears at this side of the piezo.

Thank you. That was exactly the kind of insight I was hoping to get from asking here. I will start experimenting with shielding the piezos and possibly a differential preamp now.

[magic]

By the way, I hope you made sure that the hum go away if you disconnect input cable from the preamp (and maybe replace it with a capacitor if Johnson noise of the biasing resistor is too much)

[Balthazar]

By the way, I hope you made sure that the hum go away if you disconnect input cable from the preamp

Yes, if I disconnect the piezos the hum is gone.

[Balthazar]

Hi

I would like to bring an update to what I did and how it went. Due to the extra complexities of a differential amplifier, and I didn't have an instrumentation amp IC, I first tried to shield the pickups to see how much I could buck the hum that way. It turned out a great success.

The first thing I tried was to lay grounded copper tape on the cello body under the pickup (it's a solid body, not a classical cello), and let the top plate of the pickup be connected to ground. So I didn't need to stick copper tape directly on the pickup, and the plate carrying the signal would still be sandwiched between grounded metal sheets. At first I got no sound at all! It turns out, I had the "negative" side of the pickup facing down, but connected to signal and not ground (the other side connected to ground). What I didn't know was that if you press this side against conductive material you make contact. So I had shorted the pickup. Once that was resolved, the extra shielding improved SNR from 20 to about 50 dB. What surprised me was that the output from the pickup also became much stronger, which is another reason for improved SNR. I can only guess at the reasons. Could it be that the softness of the tape and adhesive actually distributed the mechanical force on the pickup more evenly? I have no idea, but the signal was significantly stronger.

50 dB SNR is ok-ish, in fact it's about what you had on cassette tapes in the old days, while well recorded LP records had about 60 dB. Anyway I wanted to improve it further, and found that even with the pickups disconnected from the preamp, I had the same noise level. So I improved the shielding of the cavity housing the circuit, there were some holes around the controls (volume control, switch etc...), and after that I had -78 dB noise level compared to typical signal level when playing, and for the first time not dominated by 50 Hz hum. So, I went ahead and plugged the piezo back in, but now I still had about -50 dB noise. So, I put a sheet of copper tape on the underside of the bridge, to ensure not only the pickup itself, but also the soldered connections were enclosed, still without putting copper tape directly on the pickup. After this, I had improved SNR to about 78 dB. I only have a 16 bit (CD quality) recording, so I can't even inspect the noise to see what it consists of (-78 dB signal level on a 16 bit recording leaves only a few bits of precision left).

So I am more than happy, and have not noticed any degradation in audio quality from the shielding. I may bring pictures tomorrow to show what I did, but right now I'm too tired.

Also, the circuit only draws about 0.2 mA, so the battery will last for a long time. I'm using a single opa196.

[bob91343]

The question is, does the hum come from pickup or from a ground loop?  If you run your amplifier from batteries, is it still there?  If so, it's pickup and shielding is an answer.  If not, look at power supply issues.

[magic]

It may be possible to ground the cases of volume pots and switches for perhaps one last dB of improvement
Also, if such things are left floating, they may cause hum when somebody manages to touch some metal part of them.

As you have seen, high-Z circuits really hate electric fields in them.

[mclute0]

if it is power noise at 50hz I would expect 100Hz and 150Hz harmonics and more.

try looking at something like this
http://www.fhi-berlin.mpg.de/elab/pub/Standardgeraete/Bandsperre_e.html

not saying what exactly is your problem, but as a side note... one of the things I learned the hard way in playing with some cheap Chinese SDR clones, sometime the RF signal you think is coming for outside the device is really coming from an internal component that was either badly designed or badly implemented. I found that that the aluminum tape used here in the US on HVAC systems costs less than copper tape and makes great shielding.

[Balthazar]

Thank you all for great contributions. As was mentioned in my long winded post above, there are no hum issues anymore. I managed to increase SNR from about 20 dB to 78 dB during testing, and 78 dB SNR is more than you could possibly care about for this use case. I did it all with shielding, and don't think there is a reason for the extra complexity of a differential amp.

I mentioned I would post pictures of what I did. So, here goes.

I wanted to test shielding without sticking copper tape directly on the pickups, to avoid damaging if I needed to remove it. It turned out to be a good idea, because one of the sides of the pickups turned out to be non insulated. First I tried to only shield only the underside, and let the grounded plate be on top. This increased SNR substantially, to about 50 dB. I did it by sticking copper tape only on the cello underneath where the vibrating bridge would be placed, and placing the grounded plate of the piezo on top. Afterwards, I stuck copper tape on the underside of the bridge as well, which further improved the SNR to about 78 dB, like this:



The bass side foot of the bridge presses down on the pickup, whereas the other foot makes contact with the grounded shielding below (I decided to go with only one pickup on one side). This is just a temp bridge I use to experiment. At first I had a problem with the pickup being shorted, because the underside, the one without the text, turned out to not be insulated:



The one on the left has the non-insulated side facing up. I resolved it with insulation tape between the copper and the metal plate, it can also be resolved by connecting this side to ground, which is probably what you're supposed to do.

During my testing I noticed that the SNR even when no pickups where connected was only 50 dB, which prompted me to improve shielding of the cavity where the electronics where placed. If you look closely at this picture, the area under the volume pot and the switch had bare wood under it. By lining this area with copper tape I removed the last bits of hum for when the pickups where disconnected:



In this picture you can see that there is still a little bit of bare wood under the output jack (upper right corner), but the shielding is good enough now, so I decided to let it be.

When I was done experimenting, I glued the piezo down with super glue. This improved the audio quality even a bit more, and the SNR ratio as well, so it is now probably well above 78 dB which is more than enough anyway.

[Balthazar]

Sorry, I could not make the inline images work. I have also made some recordings and screenshots of the spectrum of noise before I started insulating and after. I don't have recordings of playing it, but the noise is recorded at such a level that when playing, the volume will be at a typical level for streaming services, Youtube etc.

I made the spectrums in Audacity, which made the graphics a little hard to decipher, but you get the idea.

[magic]

Audacity FFT function is weird.
IIRC, I found an option to change the limits of y scale, buried somewhere in preferences

A more convenient software with WAV file spectrum analysis function is RMAA5, an old free version of RightMark Audio Analyzer.

[Sengcid]

Have you considered a charge amplifier? It is a circuit often used with piezo accelerometers. The output is accurately proportional to the charge generated regardless of stray capacitance. Herewith some practical tips. The opa196 should do nicely (not the best but, with an extra gain stage, will provide better SNR than your recorder). 100% screening still essential.
Good luck, Chris