Unwanted piezo-sensitive amplifier circuit

While debugging a little noise problem in an amplifier, I noticed that the circuit is acting like a microphone. Vibrations in the PCB are showing up as noise in the final output.

The circuit is all surface mount, 4-layer, using CMOS MCP604 rail-to-rail op-amps with a very well-regulated and filtered +5v supply. There are three cascaded AC-coupled non-inverting stages with a combined voltage gain of about 3000. The input is currently disconnected, but returned to ground through a 1 Meg resistor. AC coupling is through a 0.002 uF ceramic cap. But I see about half a volt of dirty line-frequency noise in the output.


If I hard-short the input to ground, the noise goes away. If I lift the board off the workbench, the noise mostly goes away. If I play a loud tone generator a few inches from the board, I can see the audio waveform in the output. If I tap the board with something hard, a spike appears with every tap. Softer taps produce smaller spikes. It is linear.

What could be causing this piezo-sensitivity? A bad solder joint? A bad MCP604 amp? A bad ceramic cap or resistor? Whatever it is, I need to get rid of it.

-Robert Scott
Hopkins, MN

The ceramic caps could be the cause of your piezoelectric effects. And it's not necessarily a "bad" cap, it's just the nature of ceramic materials and I think some are worse (better?) than others. But that doesn't explain your noise pickup.

With that high gain, an input voltage 166 microvolts will give you .5V out. Further, with that 1Meg resistor, all you need is 166 picoamps of unwanted current to drive the output to .5v.

So, you've got several issues here, not just one. You've get mechanical sensitivity (probably the caps), and electrical sensitivity (probably bad shielding and/or layout). Do you have a guard ring around your input? Can you use a lower input resistor?

Ceramic caps are piezo electric. so never use for audio apps or S&H circuits.

It will pick up building vibration like a geophone.. The bigger the SMD cap, the bigger the signal I expect.

Cap type Touch screens and accurate S&H circuits use a plastic film cap for this reason too.

@Barry: I resoldered a few parts and connected the input to a low-Z source. The noise went away. I guess my mistake was assuming that the 1 Meg load would be sufficient without an input connected.

@SunnySkyguy: I didn't know ceramics were so piezo sensitive. The question now is can I use my existing PCB? The 0.002 uF caps were SMD 0603 packages. I see that Digikey has a Polyphenylene Sulfide cap in that same 0603 package size. Is that what you mean?

Not all ceramic capacitors have the same piezolectric sensitivity. Actually it's only strong with high Er dielectrica, e.g. X7R. C0G has about no piezolectric effect. You get C0G 0603 up to at least 4.7 nF.

Tunelabguy said:

@Barry: I resoldered a few parts and connected the input to a low-Z source. The noise went away. I guess my mistake was assuming that the 1 Meg load would be sufficient without an input connected.

@SunnySkyguy: I didn't know ceramics were so piezo sensitive. The question now is can I use my existing PCB? The 0.002 uF caps were SMD 0603 packages. I see that Digikey has a Polyphenylene Sulfide cap in that same 0603 package size. Is that what you mean?

Click to expand...

Yes that's what I meant except since they are low K , tend to be thruhole in large values. FVM is correct about COG.
We found manufacturing a touch screen that the ceramic reference cap caused noise with vibration so it was switched to metal film. but the noise was only in 10's of mV,

I also found in the past that ceramic caps used in S&H circuits had a memory effect that caused sag, perhaps due to high ESR and other effective parallel capacitance like some electrolytics. So I switched to film cap to eliminate the problem.

If a floating 1M input causes noise, that tells you the H field nearby is strong enough to induce a large voltage, so check your inpout signal loop to gnd (EDIT) meaning , check for conducted noise on gnd path and radiated noise on signal loop, then shields up if necessary,