Using Audio Toolbox to pick up an audio with the microphone from a speaker

Hello, Is it possible to create a Simulink model, where there are both a speaker and a microphone. The speaker is an output to a sin wave for example. the microphone in the model should pick the audio from the speaker. Then i want to be able to compare both original and the picked up signal by the mic. I then want to dispaly them both in a scope or data inspector for example and look for noise or distortion. If you have a model in mind could you help me with it.

This is sofar my model:

[Jimmy Lapierre]

Are you on Windows? If so, you might want to use ASIO (and a device that supports it).

[Walter Roberson]

If you were using ASIO you would probably be heading towards a loopback driver. However in this situation the user appears to want to measure distortion caused by the speaker and microphone and the sound hardware.

[Hamda Zarrouk]

Yes exactly, I won't be using ASIO, because I am measuring distortion caused by the picked up signal from the microphone. This model actually worked, I tested it with my headset and its microphone. Now I am looking for a way to compare these signals in an automated way.

Extra question: if the originated sound is sterio, is the signal picked by the microphone also sterio. because in its output parameters we can choose the numbers of channels. So choosing 2 would be having a sterio sound (since my input has also 2 channels)?

[Walter Roberson]

I would suggest that you use a protocol for the tests. Something along the lines of start with a rectangular pulse, then a brief silence, then a triangular pulse, then a brief silence, and then the signal you want to test, after a known delay. This would permit you to synchronize the signal while also protecting against accidental synchronization.

Chosing stereo in the speakers does not mean that you can choose stereo in the microphone, but it is not uncommon to be able to choose stereo for microphones.

I would, however, point out that most single microphones have poor stereo separation. The smaller (physically) the microphone, the more it acts like a point-like receiver. Unless the microphone was specifically designed to orient its main pick-ups away from each other, you will have difficulty telling the channels apart.

Which reminds me, that it might make sense for your protocol to use a different lead-in for the two channels, so that you can tell which output is being received on which channel, and also to be able to determine how well you can separate the channels.

[Hamda Zarrouk]

Thank you for the good explanation. Unfortunetly I don't know exactly how to proceed because I am new to Simulink. How should i place the different wave generators (rectanguar and Triangular) and set the parameters (Amplitude, frequency)? How to add or create a silence between the different pluses? Is there a specific block to use or a timer maybe?

Last, what do you mean with a different lead-in for the two channels? Which two channels are you refering to, because I want to be able to tell which output is being received on which channel. Thank you for your time and effort.

[Walter Roberson]

There are probably multiple ways but one way is to use https://www.mathworks.com/help/simulink/slref/if.html where you would use the clock as input to comparisons against constants that say when each block is to turn on.

https://www.mathworks.com/help/simulink/slref/pulsegenerator.html

https://www.mathworks.com/help/simulink/slref/repeatingsequence.html

The two channels I refer to would be your left and right channels.

For example suppose you have the left channel start with three square wave pulses, and the right channel start with three triangle waves whose peaks are right at the center of the square waves, but with exactly negative amplitude. If the microphones managed to achive perfect separation, then one recorded channel would be pure square wave and the other pure triangle wave. But if the separation is not perfect, then the amplitude at the center can tell you about how much mixing there is: an equal mix (no separation at all) would have 0 at the center; a positive value at the pulse is more square-wave side, and a negative value at the center would be more triangle-wave side. But you also need to deduce something about the relative delays between the two signals, which would give you information about how off-center the microphone is relative to the speakers.

I do not know if square and triangle are the best shapes to be able to decode the mixing; you should be able to use the same principle though.

[Hamda Zarrouk]

Thank you for the detailed explenation, I will use that principle for the testing.

So Until now I succeeded in having first the perfect sine in mono and then picked up by the mic and displayed in the scope. I need to compare these two sines that are supposed to be identical (maybe a small phase shift), and to be able to detect any glitches. The idea I have is to write a Matlab function that iterated through both matrixes and compare sample by sample until the end and reports as soon as two smaples are not similar. and in this case each channel have 1024 sample and they consitute one frame. if one of several samples are not the same the the frame as a whole is not valid for example. Then i would import it to simulink as a function block.

Do you have a simpler and efficient way or is this method good?

[Walter Roberson]

If you have two signals that are nominally the same except possibly shifted in time, then xcorr() to find the best mutual alignment.

[Hamda Zarrouk]

Is it possible to use this function with two array with the same size and data type but porbably not the same amplitude, because after correcting the phase shift I want to go ahead and compare sample by sample.