Electret mic and phone for heart and lung sound for telemedicine

[Buk]

Another thought re: noise.

I can't see how you are mounting the electret in the bell, but if you haven't already done so, you should try to isolate it accoustically from the bell to reduce handling noise.

The simplest/cheapest option would seem to be mounting the mic on a piece of fine grained closed cell foam.

I looked around for stuff that might work that would either be cheap to buy or (as I do) you might find lying around.

I found some scraps of EVA (ethylene-vinyl acetate) floor tiles, and some polyethelyne pipe isulation:

Both are electrically insulating, so you could get away with mounting the mic by pushing its leads (or stiff wire if it doesn't have leads) through a small slug cut to fit in the end of your clear tube.

 

[audioguru]

Your recording produces no heartbeat sounds. The very low frequencies are missing. The background noise seem lowered when there should be a heartbeat.

Maybe the mic is a wrong dynamic type (coil and magnet) and maybe its very low impedance is shorting the 60Hz electrical hum pickup by the mic wires since you did not connect the mic with a very important shielded audio cable? Those ordinary wires will pickup LOTS of 60Hz.
Or maybe your recording system has AGC that reduces the gain during each high level heartbeat? Maybe the recording system does not pass low frequencies?
Maybe the 47k resistor R6 has a value much less?

How will a telephone system pass low frequencies?

 

[Buk]

Your recording produces no heartbeat sounds. The very low frequencies are missing.

Dunno what you're seeing/hearing/analysing?
I hear and see a very distinct heartbeat and lots of low frequencies all the way down to 1Hz in the heart sample:

For contrast, the lung sounds with lots of pops and scratching.

 

[audioguru]

Even in your graph showing the extreme noise, during each heartbeat the noise reduces instead of producing a low frequency pulse.
Is it an AGC circuit that reduces the noise during each heartbeat? Maybe the gain is too high and something is clipping?
Where is all that noise coming from? The stethoscope head should be placed on the chest, not on a foot.

 

[Buk]

Even in your graph showing the extreme noise, during each heartbeat the noise reduces

Okay. I see what you are getting at.

Maybe the gain is too high and something is clipping?

Wouldn't that be easily determined by replacing R7 with a potentiometer to allow the gain to be adjusted?

 

[doldett]

Some of my friends can hear and some cannot. I think the noise is little to much. I have no idea where it is coming from. It could be that I does not use the shielded audio cable. It could also coming from my shaking hand causing the stethoscope head to scratching with my shirt.

I also think I will find eva or sponge to isolate mic from the stethoscope head to reduce noise.

Do you think i can use the cable (cable with bare wire (sorry for language)) connecting to the mic?

Can i cut the ts male and use only the shielded wires? I found that these shielded wires are so thin it is really difficult to connect to the block terminal on the perfboard.

Wouldn't that be easily determined by replacing R7 with a potentiometer to allow the gain to be adjusted?

I have 1-turn 1M potentiometer and will put it in the circuit next.

 

[audioguru]

The cable with the plug does not say (when translated) the cable is shielded. It might simply have a black wire and a red wire.
The gain of the first opamp is only 1 + (1M/47K)= 22 times so I doubt the opamp is clipping. The recording device might be clipping or removing low frequencies.

 

[Buk]

The gain of the first opamp is only 1 + (1M/47K)= 22 times so I doubt the opamp is clipping. The recording device might be clipping

Probably not, but if the recording device is being driven into clipping, wouldn't reducing the gain prevent it from happening and improve s/n ratio.

The recording device might be ... removing low frequencies.

If that were the case, then the lower frequencies would be entirely missing; not just when the signal input is loud.

The obvious way to check the recording devices fidelity would be to record a known signal (an exisiting recording with a a good range of frequencies like pink noise) and then compare; but the spectra above show that it can record from 1 Hz through 21kHz.

 

[doldett]

Many thanks everyone for suggestion. I better find where noise source is coming from. I think what I am gonna do next in no particular order are

1. use ipad to record the sound and compare the result. If the noise level is relatively similar, the recording devices should be ok.
2. use the shielded audio cable for mic connection.
3. change mic to the ones bought from digikey (scheduled to arrive on 10th August together with opa2134) and use the audio cable.
4. change the stethoscope head as it, at the moment, is not hard plastic probably causing handling noise when pressed hard.
5. use eva or sponge wrapped around the mic to reduce handling noise

Any more suggestions guys I am all ears!

Thought on the noise. I have read that differential amp could help reduce the noise (differential inputs). I am not sure yet how it work, but could it be added in the circuit?

 

[Buk]

Doldett

Any more suggestions guys I am all ears!

To confirm or otherwise AG's suspicions that the phone is filtering the (lower) frequencies whilst recording, play this 192k sampled pink noise file (from here) on one phone; and record it using your electret on another phone.

Then use this online spectrum analyser on your recording. In log mode (checkbox) the results should look very similar to this:

If the phone is applying any filtering (low-/hi-/band-pass; companding etc.) to the audio recording, the resultant spectra will look very different.

 

[doldett]

I had put 1-turn 1M resistor in the circuit and the circuit decides to be noisy. Could not figure it out why changing 1M resistor to 1M potentiometer has cause such high frequency noise. It looks like the circuit has been oscillated at 1.6MHz at the output.

I had also changed NJM5532 to OPA2134. These high frequency noise can be clearly captured on the oscilloscope using ipad, but not on phone. However, high pitch noise is present on all recording. Using 9V battery or 9V DC supply does not make any difference. I am really puzzled as my recordings from 2 days ago was not these bad, though they had noise.

Anyway, I will recheck the circuit again tomorrow. Maybe I was doing something very wrong today.

 

[audioguru]

The layout of the circuit on your perf board will cause it to oscillate at the radio frequency, especially if the input with no shielding is near the output.
I hope your ICs are not from ebay or Ali..., Bang..., or Amazon that could be defectives of fakes.
A while ago an Oriental manufacturer made some "improved" TL072 dual opamps that oscillated at a radio frequency so they were recalled, but many were sold and never came back (probably ended up on ebay).

 

[Buk]

I will recheck the circuit again tomorrow.

I am going to be talking beyond my level of knowlege here, so take everything I say with a huge handful of salt.

I think the circuit you are using is way overkill for your application.

Many smartphones come with a simple earphone headset -- that has no batteries -- which whilst not sufficient to record heartbeats, is perfectly adequate for voices.

According to this wikipedia page they are:

• Devices that use a "plug-in powered" microphone: an electret microphone containing an internal FET amplifier. These provide a good quality signal, in a very small microphone. However, the internal FET needs a DC power supply, which is provided as a bias voltage for an internal preamp transistor.

Plug-in power is supplied on the same line as the audio signal, using an RC filter. The DC bias voltage supplies the FET amplifier (at a low current), while the capacitor decouples the DC supply from the AC input to the recorder. Typically, V=1.5 V, R=1 kΩ, C=47 μF.

I'm fairly convinced -- on the basis of the spectra from the recording you made, and running this simulation -- that the circuit you are using is overdriving the input stage of the phone and producing severe clipping.

(I also suspect that a couple of the capacitors -- C2 and C10 -- are wrong/unnecessary; because I had to remove them in order to get the simulation to work.)

And I suspect there is a conflict with driving into a device that supplies a dc bias voltage.

I think you only need a fet or at most a single stage opamp in order to drive the mic input of your phone; and the rest -- designed to drive a pair of active headphones? -- is overkill.


I'm fully expecting to get beaten up here for expressing my unqualified opinion; but remember guys its the OP that needs help. I'm just a bystander trying to help him.

 

[audioguru]

I agree that the signal level from this electronic stethoscope circuit is way too high to feed the mic input on a phone. The phone must be clipping like crazy.

Buk, where is all your noise coming from? The first opamp has a gain of only 3.0 and the Sallen-Key lowpass filter is 1.6 for a total gain of only 4.8 times.
Line level inputs in circuits like this have a gain of more than 5 times with noise so low it can barely be measured.

C2 is very important to prevent an opamp from oscillating which is probably happening in your simulation.

 

[doldett]

Thanks AG and Buk for your input. I was also suspecting that the phone might be clipping. It is why I was probing the voltage across R13 (5.1k). The dc voltage across this R13 is 1.25V and I was afraid that with negative cycle this voltage could be driven to 0V. Unfortunately, I was greeted with high frequency noise instead T__T and totally forgot that I were to test clipping.

The layout of the circuit on your perf board will cause it to oscillate at the radio frequency, especially if the input with no shielding is near the output.
I hope your ICs are not from ebay or Ali..., Bang..., or Amazon that could be defectives of fakes.
A while ago an Oriental manufacturer made some "improved" TL072 dual opamps that oscillated at a radio frequency so they were recalled, but many were sold and never came back (probably ended up on ebay).

NJM5532 was bought from reputable supplier in Thailand, while OPA2134 was purchased from digikey. So i think they are genuine, not the fake ones.
I try to solve the issue with unshielded cable. I am waiting for **broken link removed** and this male to male 3.5mm trrs audio cable to arrive. The former is from china so it might take few weeks before I get my hands on them.

 

[Buk]

I agree that the signal level from this electronic stethoscope circuit is way too high to feed the mic input on a phone. The phone must be clipping like crazy.

Phew! At least that conclusion was right.

Buk, where is all your noise coming from? The first opamp has a gain of only 3.0 and the Sallen-Key lowpass filter is 1.6 for a total gain of only 4.8 times.
Line level inputs in circuits like this have a gain of more than 5 times with noise so low it can barely be measured.

Literally no idea.

But if the circuit is wrong for the purpose, then it doesn't seem worth pursuing. Better to attempt help the OP come up with a better circuit.

C2 is very important to prevent an opamp from oscillating which is probably happening in your simulation.

Problem is, if I add C2, the simulator complains "Capacitor loop with no resistance". (Bottom left of the screen and C2 highlight in light blue) and will not run.

 

[audioguru]

The 9V battery or power supply is missing its low output resistance.
Your opamps are not powered.
ALL electronic circuits need a decoupling capacitor across the power supply pins of ICs.

Your Falstad simulation with C2 is showing a correct zero output signal when the input signal is zero.

 

[Buk]

Your Falstad simulation with C2 is showing a correct zero output signal when the input signal is zero.

I showing zeros, because it isn't running! (If the "run/STOP" (top right) is red, it is stopped.)

Try clicking it. Nothing will happen. Because the as the text bottom left explains. C2 forms a capacitor loop with no resistance.

Your opamps are not powered.

The (ideal) opamps are powered. Right click the icon to see the voltage.

The 9V battery or power supply is missing its low output resistance.

What makes you think this?

 

[Buk]

Correct the C2 problem by adding a resistor inline, feed the electret pink noise at 1.5v and watch the spectrum of the output.

Start with the gain of the first opamp set to minimum and sweep up to max and back again.

Observe severe clipping.

 

[audioguru]

This circuit needs an output volume control to avoid any clipping. Then when played into the mic input of a phone, the other end of the phone call will not hear any low frequency heartbeats but will hear mid-frequency wheezes and squeaks from sick lungs.