I've been working on a project wherein I'm amplifying a very high impedance signal. Typically it will be a piezo pickup, but could also be a ribbon microphone. I decided to use the AMP02 from Analog Electronics, which is expensive but looks pretty good. The problem is, I have an unacceptable level of hum coming from the finished board.
I get pretty good sound coming from the balanced piezo pickup I'm using, but there's a hum that's way too loud to make the circuit useful. It goes away when I turn the volume down, so it's in the input stage. I don't have an oscilloscope to see, but I noticed that of all the places in the circuit, if I put my finger near the input HPF, especially the high value resistors, that the same hum gets much louder. I've triple checked all my grounding wires/cables.
I'm wondering, if the input filter is so sensitive, maybe this is noise coming from my voltage regulators nearby, picked up by the high value resistors?
Is there any way to fix this? I'm willing to try my layout again. This was basically practice at figuring out how to lay out a PCB.
Images:
Schematic
Board Top Layout
Board Bottom Layout
Thanks all! I'm still learning, especially when it comes to this high impedance stuff. I've done chipamps before and of course had no problem, but this is a different beast.
I get pretty good sound coming from the balanced piezo pickup I'm using, but there's a hum that's way too loud to make the circuit useful. It goes away when I turn the volume down, so it's in the input stage. I don't have an oscilloscope to see, but I noticed that of all the places in the circuit, if I put my finger near the input HPF, especially the high value resistors, that the same hum gets much louder. I've triple checked all my grounding wires/cables.
I'm wondering, if the input filter is so sensitive, maybe this is noise coming from my voltage regulators nearby, picked up by the high value resistors?
Is there any way to fix this? I'm willing to try my layout again. This was basically practice at figuring out how to lay out a PCB.
Images:
Schematic
Board Top Layout
Board Bottom Layout
Thanks all! I'm still learning, especially when it comes to this high impedance stuff. I've done chipamps before and of course had no problem, but this is a different beast.
If you also have a problem with noise in the technical sense of the word, a random signal that sounds like hiss or like waves on a shore:
1. Increase the input coupling capacitors to reduce the voltage drop that the equivalent input noise current of the instrumentation amplifier and input resistors cause across the capacitors.
2. For really high impedance sources, use an instrumentation amplifier with a much smaller equivalent input noise current, preferably something with a JFET input stage.
3. Consider making a separate input for the ribbon microphone. A ribbon microphone has a very low impedance, so you need very low equivalent input noise voltage while equivalent input noise current doesn't matter much.
1. Increase the input coupling capacitors to reduce the voltage drop that the equivalent input noise current of the instrumentation amplifier and input resistors cause across the capacitors.
2. For really high impedance sources, use an instrumentation amplifier with a much smaller equivalent input noise current, preferably something with a JFET input stage.
3. Consider making a separate input for the ribbon microphone. A ribbon microphone has a very low impedance, so you need very low equivalent input noise voltage while equivalent input noise current doesn't matter much.
Yes this is in an aluminum enclosure connected to ground.
1. So if I increase C7 & C9, this should decrease the noise, but also lower the frequency of the band passing the filter?
2. So the instrumentation amplifier I have chosen has a relatively high input noise current? That's kind of what I was afraid of.
3. The ribbon microphone use case was unlikely, I figured this should work under both circumstances. What I really want is for it to work well with a piezo pickup that is balanced.
However, all this being said, it isn't noise like a hiss. This is a hum. I could measure it to determine its frequency, but it's definitely hum. Sounds higher than 60hz but lower than ~200.
I'm fairly certain the problem with hum is originating from my HPF at the beginning of the circuit.
1. So if I increase C7 & C9, this should decrease the noise, but also lower the frequency of the band passing the filter?
2. So the instrumentation amplifier I have chosen has a relatively high input noise current? That's kind of what I was afraid of.
3. The ribbon microphone use case was unlikely, I figured this should work under both circumstances. What I really want is for it to work well with a piezo pickup that is balanced.
However, all this being said, it isn't noise like a hiss. This is a hum. I could measure it to determine its frequency, but it's definitely hum. Sounds higher than 60hz but lower than ~200.
I'm fairly certain the problem with hum is originating from my HPF at the beginning of the circuit.
Good, but see below.
Yes, it reduces the cut-off frequency of your high-pass.
Yes, although many are worse. With 9 nV/sqrt(Hz) and 0.4 pA/sqrt(Hz), it is optimized for source impedances of the order of 22.5 kohm (say 5 kohm to 100 kohm).
If it were mains hum, it could be that the contact to your aluminium case is not good. It is easy to get bad contacts with aluminium because of the insulating aluminium oxide layer that forms on its surface when it comes in contact with oxygen.
If it is a hum that has nothing to do with the mains frequency or its harmonics, then I wonder if the switched-mode DC-DC converter has something to do with it. Its typical switching frequency is 300 kHz at nominal input voltage and full load, but does it go into some sort of hickup mode at lower loads? If you converted the output with an ADC, could it be an alias of a switching harmonic?
Last edited:
MarcelvdG, this gives me a lot to think about.
Since you seem to know more than myself, would you say that an INA111 would be a better choice here? I was initially going to use it but it was out of stock pretty much anywhere when I was getting parts.
Will I even get DC on my piezo input? I could try bypassing the input capacitors and see if that does anything.
I'm pretty certain I have a good connection to my chassis. I used a bolt with double nuts and a ring terminal sandwiched between.
If all else fails, I will try popping in a different instrumentation amp.
Anyway, I hope that if I have to change instrumentation amps, that I can desolder this one. I have a heat gun so that should help, but I really don't want to have to completely restart the project.
I should probably analyze that hum with a spectrograph or something. I will connect to my computer and see what I can do.
I know this was biting off more than I can chew, but I've really learned a lot so far.
Since you seem to know more than myself, would you say that an INA111 would be a better choice here? I was initially going to use it but it was out of stock pretty much anywhere when I was getting parts.
Will I even get DC on my piezo input? I could try bypassing the input capacitors and see if that does anything.
I'm pretty certain I have a good connection to my chassis. I used a bolt with double nuts and a ring terminal sandwiched between.
If all else fails, I will try popping in a different instrumentation amp.
Anyway, I hope that if I have to change instrumentation amps, that I can desolder this one. I have a heat gun so that should help, but I really don't want to have to completely restart the project.
I should probably analyze that hum with a spectrograph or something. I will connect to my computer and see what I can do.
I know this was biting off more than I can chew, but I've really learned a lot so far.
Last edited:
I tried really hard to do my grounding properly, but I'm also kinda newbish. All audio chips reference to a common ground that is incorporated into ground planes on the top and bottom of the board. Input/Output grounds, and the chassis, connect to this ground plane via individual wires. Common ground is connected to "0v", referenced by the voltage regulators, by a 1nF capacitor.
I tried connecting chassis to 0v instead but that made it slightly worse.
I tried disconnecting my indicator just in case.
I have not tried connecting XLR input ground to chassis instead of board, but I can't imagine that improving things.
Reply to post #7: as your main issue is not noise but hum, I'd suggest trying to fix that first. Maybe the noise performance is good enough for your application.
Often, amplifier hum can be caused by either or both of two things:
1) Too much enclosed loop area, within which a time-varying magnetic field is able to induce a time-varying current, and vice versa. For that, try to keep the two input signal conductors always as close together as possible. Run board traces very close together and tightly twist signal wire pairs. Also, tightly twist all power supply wire pairs and keep power traces as close together as possible.
2) Sharing of ground-return current paths: Conductors all have inductance and resistance, so currents returning to the power supply can induce non-zero voltages at a “ground” point, away from the power supply ground. And those “ground bounce” voltages can be significant, especially since the inductance makes them proportional the the rate of change of the ground return currents. In the case of a sensitive input of an amp, the input’s ground connection might need to have its own private path back to the power supply ground. Otherwise the ground point voltage itself might get bounced around. Look up “star grounding”.
Since the noise seems to be affected by the capacitance of my finger being nearby the input HPF, I'm wondering if I need to connect said HPF's resistor to a different ground, such as the chassis bolt, rather than the PCB's ground plane.
edit: In response to 1), I tried to keep my input leads short, twisted to the board, and my traces as direct to the inst amp as possible. I'm suspecting my grounding choices.
edit: In response to 1), I tried to keep my input leads short, twisted to the board, and my traces as direct to the inst amp as possible. I'm suspecting my grounding choices.
OK, things I am going to try in this order:
1. Change input hpf capacitor from 390pF to 10nF.
2. Change input hpf resistor grounding, connecting it directly to chassis.
3. Completely start from scratch, cursing.
1. Change input hpf capacitor from 390pF to 10nF.
2. Change input hpf resistor grounding, connecting it directly to chassis.
3. Completely start from scratch, cursing.
First, use a battery supply and see if the hum is still there.
The DC/DC converter must be ruled out first.
The DC/DC converter must be ruled out first.
Last edited:
Lots of good ideas everyone.
I have worked with many ultra high Z circuits. My bet is capacitive coupling. Locally shielding the high Z node often solves the problem.
A small square of copper, tin or even aluminum foil (1 in square for instance) can be connected to the local ground with a test clip or fly wire to the foil and be used as a diagnostic shield. This foil may be insulated with tape to avoid shorting to other circuit nodes. It can be moved around and even be bent into needed shapes to shield either the noise emitter or the high Z node.
Good luck!
I have worked with many ultra high Z circuits. My bet is capacitive coupling. Locally shielding the high Z node often solves the problem.
A small square of copper, tin or even aluminum foil (1 in square for instance) can be connected to the local ground with a test clip or fly wire to the foil and be used as a diagnostic shield. This foil may be insulated with tape to avoid shorting to other circuit nodes. It can be moved around and even be bent into needed shapes to shield either the noise emitter or the high Z node.
Good luck!
Yes. Your input caps are 390pFd. That's way tiny. Normally we might assume the source impedance is relatively "low". If your body has 1V of induced 60Hz (typical) and 3.9pFd of proximity capacity to the input circuit (after the 390pF caps), then you have 10mV of hum on the input. Depending on your instrument, maybe "medium loud", way intrusive.
Hi-pass at the input is almost always a bad idea. Aside from 1/f hiss considerations, a filter with 20Meg nominal impedance will be wide-open to capacitive hum injection. As you see.
You can try to shield but it may be like keeping porcupines out of my pear tree. (We now have 2 layers of wire mesh and assorted baricades, and they still get up there.)
Unless your sources are incredible subsonic thumpers (I'm thinking of a badly ventilated concert hall), let the subsonics through the first stage and filter at a later point, Here, input of IC5 suggests itself. However a dedicated filter stage is often a best idea.
Ribbon mike is low impedance.
I ordered some 10nF caps for a different project, could these be a better solution? I know that the cutoff is now <1hz, but that would still block dc.
Or should I completely omit those input caps, and just keep the resistor? I can take care of filtering at a later stage.
Or should I completely omit those input caps, and just keep the resistor? I can take care of filtering at a later stage.
Use 1M instead of the 20M resistors, then 0.1uF is fine, and 10nF is acceptable.
Noise should be lower.
Noise should be lower.
So update, I completely bypassed the input caps for the hell of it. No audible hum now. Almost imperceptible hiss. I need to attach my piezo to an instrument now, but it seems to have fixed the problem, sort of.
I will still try a higher value cap when I get one.
Next iteration will be an INA series amp instead I guess.
I will still try a higher value cap when I get one.
Next iteration will be an INA series amp instead I guess.