Implementing the Chorus Effect

I have implemented a code for flanging. Flanging uses a delay of 0-10ms (in this case 0-3ms) to produce the desired output. Now, I want to implement this same code for creating a chorus effect, which uses a delay of 25-30 or 30-50ms. Please suggest what I should do

clear all;
close all;

infile='acoustic.wav';
outfile='out_chorus.wav';

% current sample is 11kHz so 0-3 ms is 0 - 33 samples

% read the sample waveform
[x,Fs] = audioread(infile);
% parameters to vary the effect %

min_time_delay=0.010; % 10ms min delay in seconds
max_time_delay=0.025; % 25ms max delay in seconds
rate=1; %rate of flange in Hz


index=1:length(x);

% sin reference to create oscillating delay
sin_ref = (sin(2*pi*index*(rate/Fs)))';    % sin(2pi*fa/fs);



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

min_samp_delay=round(min_time_delay*Fs); %convert delay in ms to min delay in samples
max_samp_delay=round(max_time_delay*Fs); %convert delay in ms to max delay in samples

y = zeros(length(x),1);       % create empty out vector

y(1:max_samp_delay)=x(1:max_samp_delay); % to avoid referencing of negative samples

amp=0.7; % suggested coefficient from page 71 DAFX

% for each sample
for i = (max_samp_delay+1):length(x)
    cur_sin=abs(sin_ref(i));    %abs of current sin val 0-1
    cur_delay=ceil(cur_sin*max_samp_delay);  % generate delay from 1-max_samp_delay and ensure whole number  
    y(i) = (amp*x(i)) + amp*(x(i-cur_delay));   % add delayed sample
end


% write output
audiowrite(outfile, y, Fs);

figure(1)
hold on
plot(x,'r');
plot(y,'b');
title('Flanger and original Signal');

[Mathieu NOE]

hello

have you tried the examples given in matlab :

https://fr.mathworks.com/help/audio/ug/delay-based-audio-effects.html;jsessionid=f7cdf69ca482078f4a73fa86d55d

[Mathieu NOE]

also available info's here :

http://users.cs.cf.ac.uk/Dave.Marshall/Multimedia/PDF/07_Audio_Effects.pdf

[Vincent Abraham]

Yes I have. In fact that's where i got the code for flanging. But it doesn't contain the code for chorus. For the implementation of chorus it says that i need to use a low-pass noise signal but I don't know what that means. If you could help me with this it would be great

[Vincent Abraham]

And i have seen those examples, but they are implemented on Simulink

[Mathieu NOE]

and you don't have Simulink ?

[Vincent Abraham]

I do. But i need to implement this in code

Answer 1

[Mathieu NOE]

hello again

so this could be considered as a vectorized version of the original code. You can add multiple delayed versions of the original signal

just make sure the dimensions of rate and amp are consistent

hope it matches your musical expectations...

infile='DirectGuitar.wav';
outfile='out_chorus.wav';
% current sample is 11kHz so 0-3 ms is 0 - 33 samples
% read the sample waveform
[x,Fs] = audioread(infile);

% normalize x to +/- 1 amplitude
x = x ./ (max(abs(x)));

samples = length(x);
% parameters to vary the effect %
min_time_delay=0.010; % 10ms min delay in seconds
max_time_delay=0.025; % 25ms max delay in seconds

% nb_effects = 4; % assume we do 4 delayed versions of the signal that will be summed 
rate=[1 0.9 0.8 0.7] ; %rate of flange in Hz (as many as chorus effects)
nb_effects = length(rate); % assume we do 4 delayed versions of the signal that will be summed in the final section

% amplitude is now a vector
amp=[0.7 0.6 0.5 0.4 0.3];; % suggested coefficient from page 71 DAFX; 
% Nb the first value applies to x and the remaining values apply to the 4
% delayed versions of x) so make sure this vector has length = nb_effects  +1

index=(1:samples)';
% sin reference to create oscillating delay
sin_ref = (sin(2*pi*index*(rate/Fs)))';    % sin(2pi*fa/fs);  % make sure this is a matrix dimensins =  (nb_effects , samples)
% sin_ref = 0.5*(1+sin(2*pi*index*(rate/Fs)))';    % sin(2pi*fa/fs);  % make sure this is a matrix dimensins =  (nb_effects , samples)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
min_samp_delay=round(min_time_delay*Fs); %convert delay in ms to min delay in samples
max_samp_delay=round(max_time_delay*Fs); %convert delay in ms to max delay in samples
y = zeros(samples,1);       % create empty out vector
y(1:max_samp_delay)=x(1:max_samp_delay); % to avoid referencing of negative samples
% amp=0.7; % suggested coefficient from page 71 DAFX

% for each sample
for i = (max_samp_delay+1):length(x)
    cur_sin=abs(sin_ref(:,i));    %abs of current sin val 0-1
    cur_delay=ceil(cur_sin*max_samp_delay);  % generate delay from 1-max_samp_delay and ensure whole number  
    % add the amp weigthed multiple delayed x values
    y(i) = amp(1)*x(i) + sum(amp(2:nb_effects+1)'.*x(i-cur_delay));   % add all 4 delayed sample (in one shot !)
end

% write output
% normalize y to +/- 1 amplitude
y = y ./ (max(abs(y)));

audiowrite(outfile, y, Fs);
figure(1)
hold on
plot(x,'r');
plot(y,'b');
title('Chorus and original Signal');

[Vincent Abraham]

Thanks a lot for this! Now there's only one problem, as we are using a sine signal as a reference, there's still some flange in the audio. In the pdf it was written that a lowpass noise signal should be used. Could you tell me how do I do that ?

[Mathieu NOE]

ok - I'll modify the code - see you soon

[Mathieu NOE]

so this is it

I must admit the listening of the effect is strange

Now you have plenty of parameters to play with :

• nb of effects : 4 (now)
• amplitudes of delayed signals : amp = [0.7 0.3 0.3 0.3 0.3];

• initial randon signal amplitude (0.2) : rand_ref = 0.2*rand(nb_effects , samples);

• you can apply a low pass filter effct on this "brute" random values to get a LP filtered random. This is driven by : b = 0.2; % coefficient for 1st order IIR low pass filter
• you can see the effect of this LP filter on figure (2). at the beginning there is no filtering (no worry as it is not applyed yet on the signal), then when i = max_samp_delay+1 at the beginning of the for loop , we have the LP effect coming into play

infile='DirectGuitar.wav';
outfile='out_chorus.wav';
% current sample is 11kHz so 0-3 ms is 0 - 33 samples
% read the sample waveform
[x,Fs] = audioread(infile);

% normalize x to +/- 1 amplitude
x = x ./ (max(abs(x)));

samples = length(x);
% parameters to vary the effect %
min_time_delay=0.010; % 10ms min delay in seconds
max_time_delay=0.025; % 25ms max delay in seconds

nb_effects = 4; % assume we do 4 delayed versions of the signal that will be summed 

% amplitude is now a vector
amp = [0.7 0.3 0.3 0.3 0.3]; % suggested coefficient from page 71 DAFX; 
% Nb the first value applies to x and the remaining values apply to the 4
% delayed versions of x) so make sure this vector has length = nb_effects  +1

index=(1:samples)';
% rand reference to create delay
rand_ref = 0.2*rand(nb_effects , samples);    % rand returns a single uniformly distributed random number in the interval (0,1).
                                            % make sure this is a matrix dimensins =  (nb_effects , samples)
b = 0.2;     % coefficient for 1st order IIR low pass filtering or rand modulation amplitude
             % b = 1 => no LP filter, b = 0.01 very slow filtering 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
rand_ref_filtered = rand_ref; % init
min_samp_delay=round(min_time_delay*Fs); %convert delay in ms to min delay in samples
max_samp_delay=round(max_time_delay*Fs); %convert delay in ms to max delay in samples
y = zeros(samples,1);       % create empty out vector
y(1:max_samp_delay)=x(1:max_samp_delay); % to avoid referencing of negative samples
% amp=0.7; % suggested coefficient from page 71 DAFX

% for each sample
for i = (max_samp_delay+1):samples
    
    rand_ref_filtered(:,i)  = (1-b)*rand_ref_filtered(:,i-1) + b*rand_ref(:,i);    % IIR low pass filtering of rand_ref
    cur_delay=ceil(rand_ref_filtered(:,i)*max_samp_delay);  % generate delay from 1-max_samp_delay and ensure whole number  
%     cur_delay=ceil(rand_ref(:,i)*max_samp_delay);  % generate delay from 1-max_samp_delay and ensure whole number  
    % add the amp weigthed multiple delayed x values
    y(i) = amp(1)*x(i) + sum(amp(2:nb_effects+1)'.*x(i-cur_delay));   % add all 4 delayed sample (in one shot !)
end

% write output
% normalize y to +/- 1 amplitude
y = y ./ (max(abs(y)));

audiowrite(outfile, y, Fs);
figure(1)
hold on
plot(x,'r');
plot(y,'b');
title('Chorus and original Signal');

figure(2),plot(1:1000,rand_ref_filtered(:,1:1000));grid


sound(y,Fs);

[Vincent Abraham]

Thanks a lot! Is there a way through which i could filter out more noise from the output ?

[Mathieu NOE]

hmm

i wonder if the code implementation is truly what we were looking for.

I tried the chorus effect on a separate application, and it sounds very different, without the this unpleasant grainy sound that we get with the matlab implementation.

I followed what the pdf says and triple checked the code. So it is doing what we asked for, but I'm surprised by the result.

[Vincent Abraham]

I kinda agree. What other application did you use ? Python ?

[Vincent Abraham]

And could we implement chorus without any reference signal ?

[Mathieu NOE]

I simply used Goldwave freeware that has almost everything you need to play with wav files.

just for fun, I will try waht the Simulink version gives and maybe find the reason behind that not 100% satisfying result

that's for tomorrow when I'm done with my official work....

[Vincent Abraham]

Okay I'll try from my end too. Thanks!

[Mathieu NOE]

forgot that I don't have the audio toolbox

in the matlab doc about audio effects with simulink

https://fr.mathworks.com/help/audio/ug/delay-based-audio-effects.html;jsessionid=4b2e185e10316f1614d0c4fc1908#:~:text=%20The%20chorus%20effect%20example%20has%20six%20tunable,modulator%20applied%20to%20second%20delay%20branch%20More%20

the chorus effect is implemented with sinus LFO and not random as the pdf says; so our first implementation with sinus (and not random) modulation is better ?

[Vincent Abraham]

Maybe but idk there's still some flange in the audio

[Vincent Abraham]

Is it because we are using a single reference signal and here there are two ?

[Mathieu NOE]

why two ? you're talking about the picture ? there is only one common reference

[Vincent Abraham]

I meant there are two LFOs

[Mathieu NOE]

funny

I tested again the basic sine flanger file

when the dealy is fixed the output sound is clean, not so grainy

when the delay are variable , the sound quality is poor; there must be somewhere a problem but I don't see it yet;

[Vincent Abraham]

In this chorus code you implemented using the sine wave, i dont think that graininess was the problem. There's a whooshing effect present in the output audio which is present only in flanging signals and not in chorus effect

[Vincent Abraham]

Were you still able to generate chorus with the fixed delay ?

[Mathieu NOE]

hello

yes , for the fixed dealy (4 effects) the sound is good

as soon as we make the delay variable, it's horrible... don't know what we are doing wrong here

[Mathieu NOE]

so what do we do next ??