Hello Everyone,
I am back with more dread audio questions that I hope you gurus can help me with. This time I am trying to figure out when to transition from multiple offset domains. In my application, I have an audio input that needs to be cleaned up and then amplified and sent back out to an audio output. Based on what I have read I would expect the following flow of things.

Microphone input: You want to have a DC bias if you're using a capacitive microphone. You have your DC bias added to the input signal and then that is pumped into a preamplifier to make the input usable. At this point we have went from an analog signal based around ground to a offset voltage (let's say 2.5V). Now all our changes are around the 2.5V and then we have a preamplifier that is set to amplify between 5V and 0V. That means our sine wave is now all in the positive territory.

Filtering: That positive signal is now fed into the audio bandpass filter. In my application, I am using a 3rd order Butterworth filter with a sallen-key topography. Now two questions arise from this. One being what do you reference for the sullen key grounds? The other being is what voltage to I use for the op amp supply?

I have done some LTSpice simulations and actually have built the circuit with some different feedback but curious on what the thinking usually is. The application approach I did was to reference the signal grounds to the 0V ground and use a dual supply amplifier from 2.5V to -2.5V. The thought being that the input signal being referenced to ground is already from 5V to 0V but the output for the filter is now being brought back into the dual supply range to be used in the next sections. Maybe this is a terrible way to do this but that's what I am not positive. I can see from one side that things can be clipped because of the dual supply on the op amp. Obviously clipping a signal isn't great. In my application, I do see some clipping if the input gets over 500mV from the input source (microphone). My application can't be hard proof though because I made some changes awhile ago and had some op amps that are made for single supplies set up as dual supplies and had some clipping there.

Since this is a bandpass, it has two or three stages depending on how aggressive you want it to be. In my application, I am using three. If we follow the same flow from above, after the first stage, we have started to come down from the sine wave being around 2.5V to back to ground.

Mixing: This builds on my last question on here about the summing circuit. I understand that part but my question now is built on the above. Do I reference the non inverting input to ground or -2.5V and do I also do a single/dual supply for the amplifier? I am guessing it depends on what my offset is from the filter outputs.

Amplification: My amplifier is designed to take in a signal from 2.5V to -2.5V and amplify it. Its an I2C controller digital amplifier so no filtering or anything needed and I know that part works. At this point, if there is a DC offset, it would have to be removed.

I know this is a wall of text but as you can see, I am kind of lost when to remove the DC offset so signal quality.

 

An electret microphone has a field-effect transistor (FET) built-in to amplify the signal from the condenser microphone. This needs DC power. You apply a capacitor at the output which serves as a high-pass filter. The DC bias is now removed. The audio signal is now referenced about GND. Problem solved.

 

Microphone input: You want to have a DC bias if you're using a capacitive microphone. You have your DC bias added to the input signal and then that is pumped into a preamplifier to make the input usable. At this point we have went from an analog signal based around ground to a offset voltage (let's say 2.5V). Now all our changes are around the 2.5V and then we have a preamplifier that is set to amplify between 5V and 0V. That means our sine wave is now all in the positive territory.

If this were the case, then your DC bias would become about 500V after going through the premap. Fortunately, it is not true.

Bob

 

Thank you both for the response. The microphone part I was okay with really. The preamp chip I selected basically spoon fed me how to set it up, hah. As for the rest, this post was probably a bit on the working on this for awhile side with frustration. I spent the rest of the day running some LTSpice simulations on different setups and think I figured out my answers.

What I did end up doing is the following: Microphone -> Bias -> Preamp -> 3rd order bandpass filter -> Summing Amplifier

The microphone is biased to 2.5V and fed into the preamplifier. Unfortunately, I am not 100% sure what microphone is being used so this might need to be tweaked a little bit. I have been my real testing with just a function generator with a sine wave up to 500mV and the simulations are using a wav file scaled to predicted voltage levels.

The preamplifier (MAX4466) has a gain of 5. It is a single supply setup to keep the DC bias. With the gain, anything over 500mV will be clipped.

The bandpass filter was designed in the TI Filter Design web app. I am looking for voice filtering so I picked a center frequency of 1.85kHz with a passband bandwidth of 3.1kHz. This was based on the old telephony specs on what to filter for. Maybe this is a bad idea? It's a 3rd order butterworth with sallen-key topography. I did just pick unity gain for now since its amplified on the output. There are some losses I've noted with the real setup and simulations but nothing too bad. The op amps are dual supplied from 2.5V to -2.5V. This is essentially removing the bias and bringing the signal back down to ground reference.

The summing amplifier has each source coming through a capacitor and resistor to the op amp that has its feedback resistor to set the gain for each input. The setup is using an inverting solution with the input referenced to ground (since that's the potential we're at now). The op amp itself is dual supplied since we're now at the ground reference. The gain for now here is also at unity but I will probably add some trimmer pots to be able to tweak this if needed.

After doing simulations and using an input wave file and some noise, I have been able to record the output sound and it does sound pretty decent. Decent as in how the waveforms look and even listening to the resulting wav file versus the input wave file. It does sound a little bit quieter and crisper but I also had to scale down the input signal.

Overall I think I figured it out and I am pretty happy with it. The next version of this project might use a DSP with all the digital sampling and fun but its been a voyage to get back into the analog world versus being most digital since college (a decade ago). It's still a black magic demon to me and you guys have all my respect on being able to handle it and understand it well.

 

Audio signals, do they have a DC offset ?

Just AC coupe the audio stage,

 

If you want to go the DSP route all you need to do is AC-couple the output of the microphone into an anti-aliasing filter and go straight to the ADC. Band-pass filtering can be optimized digitally.