Sound filtering circuit - will this work?

[Bob_b]

 

[MrAl]

Hi Bob,

Can not view your attachments for some reason. Could you post them differently?

NEXT DAY:
Hey thanks for whatever you did because they are showing up now! Nice.
I still cant agree with the "corner frequency" calculation though as that is the frequency they usually call the "Natural" frequency not the cutoff. Maybe you can explain what you mean by those formulas.
I also like the 10x impedance scaling for the passive network in the second stage.

 

[audioguru]

The lousy old LM358 is too noisy (lots of hiss) to be a mic preamp, especially since you have WAY too much gain in your system. You should use an audio opamp like a TL071 single, TL072 dual or TL074 quad opamp instead. A single TL071 opamp with a non-inverting gain of 100 can replace your first two lousy old opamps.

The first opamp has a 10k input resistance that loads down the output of the mic. The impedance of the electret mic parallel with R1 is about 2.5k ohms so the input resistance of the first opamp should be about 25k or more. It is easy if it is non-inverting instead of inverting like you have now.

R1 and R2 should be fed a filtered voltage from a series 470 ohm resistor then a 100uF capacitor to ground to prevent power supply hum and noise from being amplified.

The volume control should be fed from a coupling capacitor to block the DC which causes it to "crackle" as it is turned.

Remove C12 to reduce the gain to a reasonable amount and prevent acoustical feedback howling.

 

[MrAl]

The lousy old LM358 is too noisy (lots of hiss) to be a mic preamp, especially since you have WAY too much gain in your system. You should use an audio opamp like a TL071 single, TL072 dual or TL074 quad opamp instead. A single TL071 opamp with a non-inverting gain of 100 can replace your first two lousy old opamps.

The first opamp has a 10k input resistance that loads down the output of the mic. The impedance of the electret mic parallel with R1 is about 2.5k ohms so the input resistance of the first opamp should be about 25k or more. It is easy if it is non-inverting instead of inverting like you have now.

R1 and R2 should be fed a filtered voltage from a series 470 ohm resistor then a 100uF capacitor to ground to prevent power supply hum and noise from being amplified.

The volume control should be fed from a coupling capacitor to block the DC which causes it to "crackle" as it is turned.

Remove C12 to reduce the gain to a reasonable amount and prevent acoustical feedback howling.

Hi there audioguru,

Well i think the reason people use the LM358 so much is because the power supply requirement is so so much simpler. The TLxxx amplifiers usually need a much higher supply voltage to be able to deal with their common mode input limitation. So we end up throwing an LM358 at it so we can get away with simpler power supply like +5v or +9v or something like that, which is much simpler than needing a plus and minus 9v supply or around there.

Perhaps you can recommend an op amp like the LM358 with similar power supply requirements but is more suited to low level audio work. It would be nice to have a lower noise version of the LM358 that can also operate from say +6v or even +9v if needed, but just that one single supply not a plus and minus supply or a higher supply like +18v.

 

[Nigel Goodwin]

Hi there audioguru,

Well i think the reason people use the LM358 so much is because the power supply requirement is so so much simpler.

AG is obsessed by the LM358 - yet it's used commercially in vast numbers of professional equipment and performs perfectly well.

Likewise the humble 741 works perfectly well - I even tried updating the mike preamps (on two channels out of four) on an old mixer/amp I have - you couldn't tell the slightest difference between the channels afterwards.

 

[audioguru]

The LM324 quad and LM358 dual opamps were the first low power opamps. Therefore they have outputs without enough bias current so they produce horrible crossover distortion and the low currents in their input transistors cause noise. The low currents also cause a very low slew rate so that they limit near maximum output swing to only 2kHz.

TLE214x single, dual and quad opamps are still available in DIL packages, are low noise, have extremely low distortion, have a power supply range of 4V to 44V, have a high slew rate so the output can swing perfectly as high as 450kHz and have inputs that work perfectly down to the negative supply voltage that can be 0V.

 

[Nigel Goodwin]

The LM324 quad and LM358 dual opamps were the first low power opamps. Therefore they have outputs without enough bias current so they produce horrible crossover distortion and the low currents in their input transistors cause noise. The low currents also cause a very low slew rate so that they limit near maximum output swing to only 2kHz.

And millions of them are in use in professional audio equipment, where they work perfectly - while I don't entirely deny your claims the problems are presumably only apparent under certain specific conditions?.

 

[Willen]

I simply guessed that the distortion or noise level (produced by LM358) cannot be detected by normal-maximum ears of listeners. I found many normal people they cannot hear any differences between '60kbps 22khz' and '128kbps 44khz' MP3s. So may be manufacturer are using this very cheap LM358 instead of comparatively expensive TLXXX version. But in some complicated device like recording studio, audio processors must have to use TLXXX version to get less than 0.1% distortion, isn't it?

 

[audioguru]

The output of an LM324 or LM358 is class-B for low current drain but it causes severe crossover distortion. The slew rate and frequency response are too low for negative feedback to reduce the crossover distortion much. You can add a DC load on the output so that one of its complementary emitter follower output transistors operates in class-A for low distortion.

Its low slew rate cuts frequencies above only 2khz when its output level is high. But its high audio frequencies are fine when its output level is low.

 

[audioguru]

I simply guessed that the distortion or noise level (produced by LM358) cannot be detected by normal-maximum ears of listeners. I found many normal people they cannot hear any differences between '60kbps 22khz' and '128kbps 44khz' MP3s. So may be manufacturer are using this very cheap LM358 instead of comparatively expensive TLXXX version. But in some complicated device like recording studio, audio processors must have to use TLXXX version to get less than 0.1% distortion, isn't it?

Here is a graph showing the crossover distortion of one opamp in an LM324 quad (the same opamps are in an LM358). At higher levels the distortion is as high as 3% which is audible and sounds bad. The negative feedback is as high as possible because the gain is 1. At higher gains the distortion is higher.

The noise is audible from SOME LM324 or LM358 opamps because they are noisier (or are made cheaper) than others.

 

[MrAl]

Hello again audioguru,

Well by now we all know that the output can be biased such that the crossover distortion goes away. I brought this up a long time ago and also there were things published by National Semi to the same end. So the question that remains now is not what the crossover distortion is (because we can get rid of that) but what is the left over distortion once the crossover distortion is eliminated knowing we can do that now.
Any idea what it might be after that?

Since it is not too hard to get rid of the crossover distortion, that should be the only question left.

The TLE amps sound interesting too i'll have to take a look.

 

['r-e-s-i-s-t-o-r']

I'm sorry to interrupt your discussion, but I feel like I have to clarify that we have little to no expectations considering sound quality. As long as we are able to hear something, we're happy. The circuit posted in #21 is a lot more complicated than previous ones. Is that really necessary, given that it's okay for the sound quality to be lacking?

 

[MrAl]

Hi,

Well if you dont need super good distortion spec's then you can probably get away with the original circuit or something like that of course. But you may want to restrict your circuit to using resistors and capacitors rather than inductors because they are easier to work with and obtain values in many cases.

What exactly will this be used for?

 

[Nigel Goodwin]

The circuit posted in #21 is a lot more complicated than previous ones. Is that really necessary, given that it's okay for the sound quality to be lacking?

#21 provides filtering, the earlier 'circuits' didn't - and isn't filtering what you're looking for?.

 

[audioguru]

As long as we are able to hear something, we're happy. The circuit posted in #21 is a lot more complicated than previous ones. Is that really necessary, given that it's okay for the sound quality to be lacking?

The circuit in post #21 has a high gain mic preamp using two opamps. Maybe you do not need its very high sensitivity then only one opamp can be used.
By adding one capacitor between pin 1 and pin 8 of the LM386 power amplifier its gain is increased 10 times then a mic preamp opamp might not be not needed.

It has a second-order Sallen-Key lowpass filter to cut low frequencies a lot and it has a second-order Sallen-Key highpass filter to cut high frequencies a lot. Maybe you do not need to cut low and high frequencies a lot then the filters can be very simple first-order with no opamps.

It has a two-pole lowpass filter that cuts high frequencies a lot. Maybe you do not need it.

Of course a much simpler circuit or a very complicated circuit will allow you to hear some sounds in speech and music but it will also allow you to hear many noise sounds.
The circuit does not "cancel" noises, it simply cuts some of them but it also cuts sounds that you want.

 

[MrAl]

Hi,

I just realized that he wants to also drive a *speaker*. That usually requires a power op amp or one made for use as an audio amplifier. So the circuit might require one op amp and one power amp.

A parallel LC in series with a resistance where the resistance is driven from the input and the output is taken across L and C make up a bandpass filter. This is shown in the first post and post #6. The center frequency is w=1/sqrt(L*C) where w=2*pi*f, and the gain at the center frequency is 1 and less than 1 at other frequencies.

 

[audioguru]

Hi,

I just realized that he wants to also drive a *speaker*. That usually requires a power op amp or one made for use as an audio amplifier. So the circuit might require one op amp and one power amp.

A parallel LC in series with a resistance where the resistance is driven from the input and the output is taken across L and C make up a bandpass filter. This is shown in the first post and post #6. The center frequency is w=1/sqrt(L*C) where w=2*pi*f, and the gain at the center frequency is 1 and less than 1 at other frequencies.

An LM386 power amp with its gain set to 200 can use an electret mic at its input and its output will produce plenty of acoustical feedback howling in a room.

 

[Bob_b]

Is this about right?

 

[audioguru]

C13 is a huge expensive 1uF film capacitor producing a cutoff frequency much lower than you need. Didn't you calculate it? Use 470nF for a cutoff frequency of 34Hz.
RV2 is not a volume control, it is a gain control. A volume control can be at the input pin 3 of the LM386 power amplifier.

Let us know if the hiss and distortion from the LM358 opamps is acceptable.

EDIT: The LM386 power amp is missing the very important RC in series from its output to ground to prevent high frequency oscillation.

 

[MrAl]

Is this about right?
View attachment 84068

Hi,

When R2=R1*10 and C2=C1/10 (R23,R24, C14 and C15 as per schematic), the cutoff frequency is very close to:
f=0.3/(R1*C1*pi)

where R1=R23 of the schematic and C1=C14 of the schematic and R2 and C2 are the other two components of the low pass filter section R24 and C15.

Using this simple formula we get f=1975Hz for the cutoff of the low pass filter section.

Using instead values of R1=R23=330 ohms and R2=R24=3.3k and C1=C14=0.1uf and C2=C15=0.01uf we get a cutoff frequency of:
f=2894Hz

So if a cutoff of 3kHz is needed these values get a little closer to that.