Optical mic noise

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

crazydoc

Well-known member
Joined
Jun 5, 2004
Messages
1,397
Location
Lassen County CA
I have built an optical microphone that is so noisy as to be useless. It uses a reflective optocoupler with an IR LED and a phototransistor in the same package. I am attributing the noise to the phototransistor (possibly erroneously) and am looking for configurations to make it quieter.
The basic circuit is the common emitter as in (a).

optosensor2.jpg


With respect to noise I expect (b) is equivalent.


Does anyone know, or care to make a guess if any of the other circuits or those that follow in the next image would be any quieter, or have any suggestions of other circuits that might work?

optosensor1.jpg


Unfortunately, there is no base connection to apply any feedback to, unless it were optical feedback.

I am using a Vcc of 5 volts - would increasing this (to a max of 30v) have any effect on the noise?

Would changing the value of the load resistor make any difference - in other words, would increasing or decreasing the current through the transistor make a difference?

Thanks for any thoughts you might have - I'm trying to get an idea of what might work without having to build all the circuits. Of course, it may be that using a phototransistor in the design just ain't gonna work and I'm screwed from the get-go.
 
I think the type of device you are using (phototransistor) is really the source of the noise, but I'm not sure. If you're getting noise using circuit A and you already have a quiet amplifier (short out the transistor and see if it's quiet) then you know where the noise is coming from. Does the signal to noise ratio change with increasing signal source brightness?

My suggestion would be to try a photodiode in its photovoltaic mode (no bias current, with a current to voltage converter stage - you know you've found the circuit I'm thinking of when you find the application note that gives you an op-amp, a single feedback resistor, and the photodiode tied between the + and - terminals of the op-amp), and make the detector area as big as possible. You may be in the signal-to-noise ratio where photons are contributing to the noise, so maybe a brighter light might help?

Good luck!
 
> built an optical microphone that is so noisy as to be useless.

The noise you hear is the phototransistor.

BUT... what we care about is the Signal To Noise Ratio. Are you getting ANY signal at all?

If I understand your acoustic-optical system, you won't get detectable signal until a bomb goes off. If you shine a light on a diaphragm and detect the intensity of the reflection, for normal loudness, at any practical spacing, the variation of reflected brightness is essentially zero. Acoustic powers and diaphragm deflections are VERY small.

Soft sounds vibrate smaller than a wavelength of light. If your source/detector is many wavelengths away, the variation of signal is just too teeny to sense.

You can try an interferometer (a phase-meter for light). But the AIP Microphones book (or is it the other book?) has a whole chapter on optical mikes. Much of it over my head. But my understanding is that heroically complex optical systems just can't be as good, for audio microphones, as a simple capacitor mike.

> try a photodiode in its photovoltaic mode

Agreed, but I think light-intensity sensing is doomed to failure with any photodetector.
 
[quote author="PRR"]> built an optical microphone that is so noisy as to be useless.

The noise you hear is the phototransistor.

BUT... what we care about is the Signal To Noise Ratio. Are you getting ANY signal at all?
[/quote]
Yes, I'm getting a fair amount of signal - of course the problem is that in the amplification process I'm amplifying the transistor noise also. So obviously the S/N ratio is too low. I'm trying to find a way, if possible, to minimize the noise produced by the tranduction portion of the circuit

The two possible noise generators, as I see it in my somewhat ignorant mindset, are the LED and the detector. I assume, possibly erroneously, that the LEDs output is of such a magnitude that any noise will be many orders of magnitude smaller. The LEDs light is reflected off a membrane onto the detector, in this case a phototransistor, whose collector current is sensitive to the distance from the membrane, as the membrane vibrates in response to the acoustic waves.

current_distance.jpg


The noise, by the way, I would characterize as "hiss", though I haven't measured its spectral properties. I'm just trying to see if there is an electronic means to minimize this noise I am assuming is produced by the transistor.

Thank you for your input anyway - I guess I am just going to have to build circuits and use different brands of optoreflectors to see if the noise can be dealt with.
 
http://www.fas.harvard.edu/~scdiroff/lds/LightOptics/SchleirenOptics/SchleirenOptics.html

I been thinking about Schleiren Optical systems for a "light" microphone. Sound will cause a density change in air. Can we sense it and extract the signal from all the other noise(heat, light)?
 
From vague and cobwebbed memory:

There was some sort of "laser microphone" built in the 1980's... a holographic disc was 'projected' into the air, then a laser was fired "through" the disc. The interaction of the laser "line" passing through the disc was modulated by the air pressure at the intersection.

Apparently, the disastrous effects were that full-scale output was generated by any interruption of the beam, so an energetic singer who moved and stood in the way of the "reader" beam would deafen himself in the headphones. Also, draughts and air currents from air conditioning, as well as opening doors etc. all caused massive rumble/low frequency signals.

Keith
 
[quote author="Gus"]http://www.fas.harvard.edu/~scdiroff/lds/LightOptics/SchleirenOptics/SchleirenOptics.html

I been thinking about Schleiren Optical systems for a "light" microphone. Sound will cause a density change in air. Can we sense it and extract the signal from all the other noise(heat, light)?[/quote]

Described in your link is basically the Foucault knife edge test to evaluate the surface of a telescope mirror during the process of "figuring" it.

http://www.cloudynights.com/howto/focault.htm

We go to great lengths to avoid any atmospheric disturbances during the test. I think that any differences in air density produced by audio SPL's would be minimal compared to temperature differences, and would not be detectable.

Then again, it may be that they are so transient (the eye can't detect luminosity changes greater than 30 Hz or so - that's why movies, TV. etc. go at 30 FPS or faster) that we can't see them.

Next time I'm testing a mirror I'll have somebody shout in front of it and see if I can detect any perturbations. :green:
Or set up a speaker at 10 Hz or so blowing across the optical path and see what it looks like.
 
Funny thing, I just heard a track last night on the college station, using a optical mic.
The guy said he used a light dependent resistor, (ala T4b module).
This makes more sense as far as noise is concearned.
He had some cool tracks of the leds and various lights on his mixing board being put thru the optical mic.
Your speed will be a lot slower on the decay side with an ldr, but I bet your noise goes way down.
cj
 
[quote author="cjenrick"]Funny thing, I just heard a track last night on the college station, using a optical mic.
The guy said he used a light dependent resistor, (ala T4b module).
This makes more sense as far as noise is concearned.
He had some cool tracks of the leds and various lights on his mixing board being put thru the optical mic.
Your speed will be a lot slower on the decay side with an ldr, but I bet your noise goes way down.
cj[/quote]
Was he miking an audio signal, or just holding a device up to different lights (maybe configured as a vco) and getting an audio tone output?

Per the specs I have seen, for example
http://www.hint.no/utdanninger/iu/linker/datablad/NSL19_M51(LDR).pdf
with rise and fall times in the 10 to 100ms range, this would be useless for audio applications. Also, the sensitivity is very low, and needs a relatively large change in Lux to appreciably change the resistance.

What originally attracted me to the phototransistor was the extremely steep slope of the collector current versus distance to membrane, indicating a very high sensitivity to change in illumination.
 
It was one of those "noise" tracks, you know, weird sounds generated in weird ways. He was actually getting some audio effects by the changing light patterns from different light sources.
 
Crazydoc,

Did you consider making it not as a reflector, but as an interaptor? This way I think you could make the distance between optopair shorter. As you know, I was thinking about using optopair with a ribbon, which excursion is much bigger than this of stretched membrane, so the S/N might be much better.

Everybody whom I ever spoke about optical mics, suggested interferometer.
 
Article on optical Mikes - 2 Meg PDF file, about 12 minutes on dial-up......

> he used a light dependent resistor... some cool tracks of the leds and various lights on his mixing board being put thru the optical mic.

Not a "microphone" at all, but a synthesizer controller.

When I was playing with Theremins, many years ago, I got disgusted with instability. I built an oscillator with a photo-resistor and golly, it can be played. We had a modular analog synth, so I wired some photo-Rs as control voltages. Someone thought it would be cool to take the rig outside under the trees and let Nature play it. Well, the trees were scrawny and anyway you basically get two notes: full sun, and north-sky light scatter. The most memorable part of that event was a rather young Laurie Anderson playing some of her own pieces (and also helping on the light-synth).

I also converted several transistor radios with LDRs in the top as pocket light-synths.

But to capture audio waves: way too slow.
 
[quote author="Marik"]Crazydoc,

Did you consider making it not as a reflector, but as an interaptor? This way I think you could make the distance between optopair shorter. As you know, I was thinking about using optopair with a ribbon, which excursion is much bigger than this of stretched membrane, so the S/N might be much better.

Everybody whom I ever spoke about optical mics, suggested interferometer.[/quote]
Hi Marik

As you can see, I'm still sloggin' along. I thought of using a ribbon also, but can't quite figure out how the get the flat, reflective area into a corrugated ribbon.

I'm trying to keep this simple and elegant, without esoteric parts or alignment problems, so interrupters and interferometers are out for now. Maybe a laser diode, but I think they're pretty noisy too.

I have managed to figure a way to get that aluminum leaf stretched evenly over the capsule (almost 4 inches in diameter), but not how to quantify the tension put on it so as to be tunable. That stuff is just too fragile. I still need some 1u reflective mylar. :grin:

[quote author="dale116dot7"]...My suggestion would be to try a photodiode in its photovoltaic mode (no bias current, with a current to voltage converter stage - you know you've found the circuit I'm thinking of when you find the application note that gives you an op-amp, a single feedback resistor, and the photodiode tied between the + and - terminals of the op-amp), and make the detector area as big as possible.\...[/quote]

While browsing thru some spec sheets I came upon this
http://www.taosinc.com/images/product/document/TrS17xx-E14.pdf
which looks like what you're suggesting, and all in one integrated package! Although the noise figure isn't that great, at least it's low enough to spec. With rise and fall times in the 160 usec range, the bandwidth ain't gonna be all that great, but I'll trade that for quieter. I've got a few of 'em on the way.

Thanks again for the suggestion.
 
PRR
you made me think of O Superman by Laurie Anderson. Imagine driving down the highway in heavy rain at night with the windshield wipers keeping time with the song.

Would you fall asleep or be hypnotized?
 
[quote author="crazydoc"]
I thought of using a ribbon also, but can't quite figure out how the get the flat, reflective area into a corrugated ribbon.
[/quote]

Doc,

Use thicker, probably 4-5um foil. Anything from 600V paper in oil capacitors would be good. Corrugate only ends (2 pleats would be enough), leaving middle part flat, then slightly curve the middle flat part in vertical plane, to make it stiffer. Should work.
 
From your curves posted in this thread, the static output current for very-small spacing is about 50 microAmp. The slope of the curve is about 0.4 millimeters (400 micrometers) for 1 milliAmp. From the article I posted, the deflection of the diaphragm at 140dB SPL is about 10 micrometers. So the current sensitivity should be around (10um/400um)*1mA= 25 microAmps at 140dB SPL, or 12 nanoAmps at my preferred reference level of 74dB SPL (gentle speech, harpsicord in a room).

Assuming a 5V supply and very-small spacing, 50 microAmp static current, 100K is a suitable load resistor. However working with a macroscopic package next to a fragile diaphraghm, it may be hard to get the spacing down to 0.1mm without slipping and damaging the diaphragm. So static current may be around 200 to 400 microAmps, and load resistor might have to be 20K or so.

Assuming a 20K load resistor and 12 nanoAmps signal current at 74dB SPL, the signal output voltage is 0.24 milliVolts. Hey, that's just about normal for a dynamic mike, though the impedance is much higher: 20K instead of about 0.15K. Ignoring the transistor's noise, the signal to noise level is already 20dB worse than a common dynamic mike because it makes similar voltage out of a much larger resistance.

Reducing the resistor reduces signal faster than noise, so is no good. Increasing the resistor increases signal faster than noise, but increases supply voltage needed. Reducing that 20dB noise figure to "negligible" requires 100 times the supply voltage, and 500V may be awkward in many ways.

I have neglected the transistor's noise. Because in effect the Base terminal is "open", Base-Emitter noise current is multiplied by Beta. At 50 microAmps I would expect the transistor Base-Emitter current noise to resemble a 30K resistance. I suspect that low-price object-sensing phototransistors may be far noisier than a plain transistor, because of the large exposed Base area and because a little noise doesn't matter if you are just counting boxes on a conveyer belt.

I have neglected LED noise because I just do not know. It might be utterly insignificant, but there is a lot of power there so a "very small" noise in absolute terms could be overwhelming at these low signal levels. A cross-check would be to un-power the LED and point it at a window, shading to get a similar static current in the transistor. Seems to me the Sun's noise should be small (far from zero due to sunspots and air-jiggle), or at least different from LED noise. Similar guesses for incandescent lamps, though you will get some power-line hum (try a car tail-light on a clean 12VDC supply).
 
Thanks, PRR, for your super posts. I couldn't respond yesterday as I had to go do something useful ($$$)work.

My only formal electronics training was Physics 101 about 35 years ago, so please bear with me and my lack of understanding at times.

You lost me with your first statement:
"From your curves posted in this thread, the static output current for very-small spacing is about 50 microAmp. The slope of the curve is about 0.4 millimeters (400 micrometers) for 1 milliAmp."
It looks like you're assuming that the units on the y axis are milliamps. My understanding of the graph is that the collector current is only a ratio of an arbitrary current through the collector by varying the distance from the reflector to the device: that is, there are no units. The IsubF noted on the graph is the current through the LED.

Here's the full spec sheet for this particular device:
http://www.fairchildsemi.com/ds/QR/QRE1113.GR.pdf

The maximum collector current is given as 20ma, though the typical current listed is 0.4ma. I'm not sure why the large discrepancy, unless the 0.4ma value is a quiesent current without a reflector in place.

Anyway, if I assume a maximum collector current of 10ma (by choosing a suitable load resistor of around 500 ohms at Vcc=5V), this should increase the sensitivity values by a factor of 10 to 250uA at 140dB or 0.12 uA at 74dB. But I am probably wrong.

My thought was to bias the transistor on at about the mid point of the upward slope of the curve by suitable transducer-membrane spacing (~0.25mm), and the signal would cause positive and negative deflections along the curve.

This does seem to happen - there's just a shitload of hiss. I guess I need to use variable resistances to control the LED and collecter currents and see if there is a variation in the noise. I've tried paralleling resistances across the ones I have in the circuit, but that doesn't seem to make a difference in the noise. Maybe I need to decrease the current instead.

Thanks again for your help.


Marik-

Do you know offhand the approximate excursion of a ribbon transducer at a given SPL?
 
> you're assuming that the units on the y axis are milliamps.

Yes, and that is not what it says. Me stupid.

However, if you bias the LED around 25mA, then I think the peak current (around 0.6mm) is around 1mA, so it isn't far wrong.

I'd "bias" the phototransistor by shoving the detector closer to the diaphragm. Even touch it (on a test diaphragm!) and back off just barely enough to be sure it won't hit. That will minimize your bias current, allowing a larger load resistor and probably more output voltage. Whether that improves S/N or not, I'm not sure.

I suspect you want a jig where you can micro-move the detector and also adjust the load to keep the transistor collector voltage in the linear range (1V to 10V, well below the supply voltage), and fool around with it seeing if spacing improves S/N.

Because all this fiddling also changes gain, it can be hard to judge when things are getting better. Try to rig a pulsed noise source (loop some hiss and silence on a CD track and set the player to Repeat). Adjust the playback volume so, as heard in the mike, the pulsed noise is about the same loudness as the mike's steady noise: the output gets slightly louder and softer. Then fiddle. You may be able to clearly hear the steady noise fall-away or rise above the pulsed noise a little, even though the total hiss level is changing with distance and load.

I suspect you want to run the LED as hot as possible, 20-30mA instead of the nominal 10mA which is adequate for package-counting and other large-signal chores.
 
Thanks, PRR (and Marik and dale116dot7). You've given me several new avenues to explore. That's what the fun of all this is - thinking things up and trying them out. Time to start trying them out.

I'll dig this thread up again when I get any results, positive or negative.
 
Back
Top