Guitar - Photodiode Amplifier, Signal-to-Noise ratio

[Yarrrrr]

Hi, my first time posting here. I am a beginner at electronics.

I am building an optical guitar pickup, infrared diodes bouncing light onto the guitar strings back to photodiodes.

The circuit works but the signal to noise ratio makes it unusable, the latest design attempt uses parallel amplifiers with the hope of reducing noise but that didn't help at all. The final mounted part is wrapped in copper tape and grounded to shield from outside noise.

Components:

The op-amps are NE5532

IR Diode - TSUS5202
Photodiode - WL-TDRB Series

I'll attach schematics and a sound example of the current output.

Maybe there is some glaring mistake in there somewhere.

Audio:

https://soundcloud.com/don-stroganotti%2Fpcb-test-1%2Fs-lEiGLM16XAX

 

[barry]

1) You don't show what opamps you are using. Are they low noise? You've got a gain of 4000, and you're multiplying the uncorrelated noise by 2000 (if my calculation is right).
2) It sound's like you've got wideband noise, not hum (which shielding would take care of.) How much noise do the photodiodes generate? How about the infrared diodes? How about the power supply (which you don't show).
3) Does your PCB have a ground plane?

 

[Audioguru]

The amount of IR reflected from a guitar string is extremely low, the photo-diodes are used as tiny low output solar cells feeding a very low resistance then the high voltage gain of your opamps amplifies their noise.

Instead of using paralleled inverting opamps with extremely low input resistor values, try using a single non-inverting opamp with a very high input resistance so that the tiny voltages from the photodiodes are not loaded down so much. Then reduce the gain.

Or use a reverse bias on the photo-diodes? Then reduce the gain.


Or use photo-transistors instead of photo-diodes? Then reduce the gain.

 

[Yarrrrr]

1) You don't show what opamps you are using. Are they low noise? You've got a gain of 4000, and you're multiplying the uncorrelated noise by 2000 (if my calculation is right).
2) It sound's like you've got wideband noise, not hum (which shielding would take care of.) How much noise do the photodiodes generate? How about the infrared diodes? How about the power supply (which you don't show).
3) Does your PCB have a ground plane?

The op-amps are ne5532, I've tried lower noise more expensive ones without any discernible difference, I've no idea how much noise the diodes generate.

The power supply has some high frequency switching noise but I am assuming the power supply is not an issue because switching to battery power doesn't change the noise at all.

There is a ground plane yes.

 

[Yarrrrr]

The amount of IR reflected from a guitar string is extremely low, the photo-diodes are used as tiny low output solar cells feeding a very low resistance then the high voltage gain of your opamps amplifies their noise.

Instead of using paralleled inverting opamps with extremely low input resistor values, try using a single non-inverting opamp with a very high input resistance so that the tiny voltages from the photodiodes are not loaded down so much. Then reduce the gain.

Or use a reverse bias on the photo-diodes? Then reduce the gain.


Or use photo-transistors instead of photo-diodes? Then reduce the gain.

Thanks for the response, maybe I have been looking at this the wrong way, trying to use low resistor values with the assumption that most of my prototypes had issues with Johnson noise getting amplified.

I don't think I've tried a non-inverting design since I solved most EMI noise issues with shielding.

I wonder if I could possibly first try reverse bias with the current PCB design before going back to prototype boards. How exactly would I reverse bias?

 

[betwixt]

The photo-diode arrangement is somewhat strange. No current is allowed to flow from the steady state light picked up and by connecting them to +V/2 I assume the idea is they produce positive and negative voltage swings. That isn't going to happen.

A trans-impedance amplifier with a high value feedback resistor would work better and would allow you to ground one side of the diodes, making it easier to eliminate interference. Look up TI's reference data 'sboa220a.pdf' for a design.

If the final stage is to provide an inverted signal for an XLR connection, it would be better to use two amplifier stages, one inverting, the other not. In the present configuration you could face instability if the 'cold' output is capacitively loaded. If you use two amplifiers and add a resistor (~1K) in series with each output it will avoid the problem and give better balance to the output.

Brian.

 

[Yarrrrr]

The photo-diode arrangement is somewhat strange. No current is allowed to flow from the steady state light picked up and by connecting them to +V/2 I assume the idea is they produce positive and negative voltage swings. That isn't going to happen.

A trans-impedance amplifier with a high value feedback resistor would work better and would allow you to ground one side of the diodes, making it easier to eliminate interference. Look up TI's reference data 'sboa220a.pdf' for a design.

If the final stage is to provide an inverted signal for an XLR connection, it would be better to use two amplifier stages, one inverting, the other not. In the present configuration you could face instability if the 'cold' output is capacitively loaded. If you use two amplifiers and add a resistor (~1K) in series with each output it will avoid the problem and give better balance to the output.

Brian.

Thank you for the insight about the output stage.

I Do have a OPA380AIDG4 that I tried as a single amplifier with a high value resistor, it had the same noise characteristics as the current design which I attributed to it's worse voltage noise density in the specs compared to NE5532.

That led me to try this design with parallel amplifiers with the assumption that a lot of the the noise was from the amp.

But maybe I overlooked to try it with one side of the diodes grounded.

 

[KlausST]

Hi,

The photodiodes are not DC biased. I wonder if this can work

Klaus

 

[betwixt]

The other problem you might have is with how much IR light is actually reaching the sensor. If the light is insufficient you will get lower output and need more amplification but if it is too high you get 'dilution' of the signal because the sensor is blinded by the light level. In an ideal design you would have feedback from the sensor to the LED to keep it at optimum brightness.

Brian.

 

[Yarrrrr]

The other problem you might have is with how much IR light is actually reaching the sensor. If the light is insufficient you will get lower output and need more amplification but if it is too high you get 'dilution' of the signal because the sensor is blinded by the light level. In an ideal design you would have feedback from the sensor to the LED to keep it at optimum brightness.

Brian.

I've been worried from the start that the amount of light I am able to bounce on the strings and the incredibly small amount of movement they produce would be the limiting factor in the signal to noise. But there are commercial systems like this so I'm sure it's just my flawed designs or component choice.

 

[Yarrrrr]

Hi,

The photodiodes are not DC biased. I wonder if this can work

Klaus

I don't know what this means.

 

[KlausST]

Hi,

I don't know what this means.

It means - because of the capacitor - no DC current can flow. The nide is DC wise floating. Floating means it's value is not determined.

I'm not sure how the circuit is meant to operate. But I'd connect the node to some DC voltage via a resistor.

Klaus

 

[Audioguru]

You should do it like this:

 

[Yarrrrr]

Hi,

It means - because of the capacitor - no DC current can flow. The nide is DC wise floating. Floating means it's value is not determined.

I'm not sure how the circuit is meant to operate. But I'd connect the node to some DC voltage via a resistor.

Klaus

It is an audio circuit, I only want an AC signal.

 

[KlausST]

Hi,

Yes. The resistor is not in the audio path, it just makes the photodiodes to work properly.

Where do you have the circuit/schematic from? Give a link.

Klaus

 

[Yarrrrr]

You should do it like this:

Isn't that exactly what I have right now? but connected to my virtual ground because I don't have a negative rail.

+V/2 is my reference voltage. that is connected to the + of the op amps and at one leg of the photodiodes.

 

[KlausST]

Hi

Isn't that exactly what I have right now? but connected to my virtual ground because I don't have a negative rail.

No. Audioguru's circuit has the DC path I'm talking about...
Yours has a series C...

Klaus

 

[Yarrrrr]

Hi
No. Audioguru's circuit has the DC path I'm talking about...
Yours has a series C...

Klaus

I guess you are talking about my input high-pass filter? I've tried many different variations of this circuit with and without it and it hasn't made a difference at all in the photodiode output.

Forgive me for I am very new and ignorant when it comes to electronics, I've simply got this far by experimenting and googling.

 

[Audioguru]

Your circuit has extremely low value R31 and R10 completely squashing the tiny signal from the photodiode. They are not needed, remove them then reduce the gain which will reduce the noise. Look at my post #13.

As usual, your schematic has no supply voltage so we must guess that you play the guitar softly so the opamps do not produce clipping.

 

[Yarrrrr]

Your circuit has extremely low value R31 and R10 completely squashing the tiny signal from the photodiode. They are not needed, remove them then reduce the gain which will reduce the noise. Look at my post #13.

As usual, your schematic has no supply voltage so we must guess that you play the guitar softly so the opamps do not produce clipping.

The supply voltage is 9 Volts And I have a prototype with the transimpedance amplifier that looks and sounds like this:

Played louder compared to previous example, and slightly different diodes on this prototype. Less noise but far from acceptable.

Maybe I can try bypassing the filter and remove the input resistors on my PCB design and see if it gets closer to the transimpedance noise levels.

Is it wrong of me to have worried about Johnson noise, leading to using the very small resistors?

https://soundcloud.com/don-stroganotti%2Fmounted-test-2%2Fs-0prVJSL3ogI