DIGITAL AUDIO EFFECTS

BY ORCHISAMA DAS
BIEE, 3rd YEAR, ROLL-001211102017
JADAVPUR UNIVERSITY
 CONTENTS
TOPIC PAGE NO.
Abstract 3
The Guitar Processor 4-5
Spectrogram 6-7
The Effects 8
Overdrive/Distortion 9-18
Wah-wah 19-24
Tremolo 25-27
Flanger 28-30
Conclusion 31
References 32
 ABSTRACT
• Audio signal processing, sometimes referred to as audio processing, is the intentional
alteration of auditory signals or sound, often through an audio effect or effects unit. As audio
signals may be electronically represented in either digital or analog format, signal
processing may occur in either domain. Analog processors operate directly on the electrical
signal, while digital processors operate mathematically on the digital representation of that
signal.

• In this report, implementations and algorithms of digital audio effects have been explored
from a signal processing viewpoint. MATLAB codes of all effects have been included and the
results have been shown visually. Audio results had been demonstrated and discussed in the
seminar. The concept of the guitar processor is introduced first, followed by its set up. An
introduction to understanding the spectrogram is given next preceding the discussion of five
effects- Overdrive, Distortion/Fuzz, Wah-wah, Tremolo and Flanger.
 THE GUITAR PROCESSOR
• The guitar processor is essentially a DSP processor, which is used to create
various sound effects that are used extensively in all genres of music.
• Digital Processors are not bulky, come in an all-in-one console and are easy to
operate.
• Many effects can be cascaded in a chain to produce a variety of rich and unique
sounds.

BOSS ME-50
 SETTING UP THE PROCESSOR
AMPLIFIES PROCESSED
SIGNAL

PRODUCES
ELECTRICAL
VIBRATIONS

AMPLIFIER
ELECTRIC I/P O/P
GUITAR

PROCESSES INCOMING SIGNAL

PROCESSOR
 SPECTROGRAM
Before analyzing any of the effects, it is important to understand the spectrogram.

A technical explanation would be: A spectrogram is a plot of the Short Time Fourier Transform (STFT)
coefficients of a signal. STFT is used to analyze non-stationary signals (signals with time variant
frequency). The signal is partitioned into short stationary windows and Fourier coefficients
corresponding to each windowed frame are calculated to give the STFT coefficients.

A more intuitive explanation: It is a visual tool for observing frequency domain properties of signals. It
is a 3D plot of time, frequency and amplitude of a signal. However, we see it as a 2D graph, with time
on the x-axis, frequency on the y-axis and the amplitude is represented by the intensity of colour in
decibels. Deeper the colour, more is the energy content of that particular frequency in the signal. Red (0
to 10dB) indicates highest energy and blue (-100 dB to -80dB) indicates least energy.

For example: A sine wave has a single frequency component, its fundamental frequency. The
spectrogram of a stationary sine wave of 200Hz should ideally only have a single red line at 200Hz. Due
to various windowing effects, what we get is a frequency band around 200Hz. The frequency resolution
of this band can be improved by increasing the window length.
 SPECTROGRAM OF A SINE WAVE OF 200Hz OF DURATION 0.7s
IN ALL THE FOLLOWING MATLAB CODES, THE TEST INPUT HAS BEEN TAKEN AS A SINE WAVE OF 200Hz.
 THE EFFECTS
• Some of the essential effects available in every processor are:
• Distortion/Overdrive
• Wah-wah
• Flanger
• Tremolo
• Delay
• Reverb

At the heart of all these effects, lie simple algorithms and mathematics that is
used to alter the sound of a signal. These effects can be generated by simple
MATLAB codes, and we can create our own processor, at a minimal cost!
 DISTORTION/OVERDRIVE/FUZZ

• The most popular effect used in rock music is distortion.

• Distortion is the result of a non-linear characteristic curve, which causes

the input signal to be clipped at the output.

• Depending on the non-linearity and extent of clipping, this effect can also
be called ‘Overdrive’ or ‘Fuzz’.
 TYPES OF CLIPPING-HARD AND SOFT
• WHAT IS CLIPPING?
In a power amplifier, if the Volume or Signal Gain is turned up so high that the signal
Amplitude or Gain becomes greater than that the amplifier can reproduce, the top of
the waveform (Sine Wave) can no longer be reached. Therefore, this part of wave
cannot be reproduced. The resulting waveform is chopped off or “Clipped”.
• There are two basic types of clipping – Soft Clipping and Hard Clipping.

• With Soft Clipping, there is a more gradual or gentle transition from the
normal part of the Sine Wave to the cropped off peak. Soft Clipping
provides a smoother more pleasing to the ear distortion.

• Hard Clipping is where there is an abrupt transition from the normal part
of the Sine Wave to the chopped off peak. Hard Clipping tends to provide
a less pleasing to the ear type of distortion.
 SYMMETRICAL AND ASYMMETRICAL
CLIPPING
• There are subgroups of Hard and Soft Clipping – Symmetrical and Asymmetrical
Clipping
• Symmetrical Clipping is when both the Positive and Negative sides of the wave are
chopped equally. In asymmetrical clipping the clipping on either side is unequal.
• All of the Harmonics are Odd in Symmetrical Clipping. Asymmetrical Clipping on
the other hand adds Even Harmonics to the equation.
 OVERDRIVE
• For overdrive, Symmetrical soft clipping of input values has to
be performed. A simple three layer non-linear soft saturation scheme
may be:
1
𝑓 𝑥 = 2𝑥 𝑓𝑜𝑟 0 ≤ 𝑥 < 3
3− 2−3𝑥 2 1 2
= 3
𝑓𝑜𝑟 3
≤ 𝑥<3
2
=1 𝑓𝑜𝑟 3 ≤ 𝑥 ≤ 1
 MATLAB CODE

function [y] = overdrive(x, Fs) if abs(x(i))< th, y(i)=2*x(i);end;

if abs(x(i))>=th,
% y=symclip(x)
if x(i)> 0, y(i)=(3-(2-x(i)*3).^2)/3; end;
% Overdrive simulation if x(i)< 0, y(i)=-(3-(2-abs(x(i))*3).^2)/3; end;
% x - input end;
N=length(x); if abs(x(i))>2*th,
y=zeros(1,N); % Preallocate y if x(i)> 0, y(i)=1;end;
if x(i)< 0, y(i)=-1;end;
th=1/3; % threshold for symmetrical soft clipping
end;
% by Schetzen Formula end;
for i=1:1:N, end
 OBSERVATION

TIME DOMAIN

Peaks are saturated. Output

is in phase with input.

FREQUENCY DOMAIN
Presence of odd harmonics
proves symmetrical clipping. The
amplitude of higher harmonics
decreases gradually.
 DISTORTION/FUZZ
• Hard symmetrical clipping, producing a hasher sound. A non-linear function
commonly used to simulate distortion/fuzz is given by:
𝑥 𝑎𝑥 2
𝑓 𝑥 = (1 − 𝑒 𝑥 )
|𝑥|
• This a non-linear exponential function. The gain, a, controls level of
distortion/fuzz.
• Output v/s input curve is more non-linear than that of Overdrive and produces a
180 degrees phase shift in output.
 MATLAB CODE
function [y] = fuzz(x, Fs)
% Distortion based on an exponential function
% x - input
% gain - amount of distortion,
% mix - mix of original and distorted sound
%1=only distorted
gain = 11;
mix = 0.1;
q=x./abs(x);
y=q.*(1-exp(gain*(q.*x)));
y=mix*y+(1-mix)*x;
end
 OBSERVATION

TIME DOMAIN
Hard clipping of peaks.
Output is 180 degrees out of
phase with the input.

FREQUENCY DOMAIN

Presence of odd harmonics. Amplitude

of odd harmonics greater than that of
overdrive. Hence the colour of odd
harmonic bands in spectrum is deeper.
 WAH-WAH
• A band-pass filter with time-varying resonant frequency and a small bandwidth.
Filtered signal is mixed with direct signal.

• The Wah filter can be also called a state variable filter. A state variable filter is a
type of active filter. It consists of one or more integrators, connected in some
feedback configuration. One configuration is such that it can produce band pass,
low pass and high pass outputs from a single input.
 LET’S UNDERSTAND BAND-PASS
FILTERS
• A high pass filter followed by a low-pass filter gives a band pass filter.

• Centre Frequency Equation

𝑓𝑟 = (𝑓𝐿 𝑥𝑓𝐻 )
• Where, ƒr is the resonant or centre frequency
• ƒL is the lower -3dB cut-off frequency point
• ƒH is the upper -3db cut-off frequency point
 FREQUENCY RESPONSE OF A SECOND ORDER BAND PASS FILTER

In case of Wah filters, the resonant frequency (f center) varies with

time and the bandwitdth is small.
 BLOCK DIAGRAM OF WAH FILTER

DIFFERENCE EQUATIONS:
𝑦𝑙 𝑛 = 𝐹1 𝑦𝑏 𝑛 + 𝑦𝑙 𝑛 − 1
𝑦𝑏 𝑛 = 𝐹1 𝑦ℎ 𝑛 + 𝑦𝑏 𝑛 − 1
𝑦ℎ 𝑛 = 𝑥 𝑛 − 𝑦𝑙 𝑛 − 1 − 𝑄1 𝑦𝑏 𝑛 − 1
Where
𝑓𝑐
𝐹1 = 2 sin 𝑝𝑖 ∗
𝑓𝑠
𝑄1 = 2𝑑
fc=cut off frequency, d= damping.
 MATLAB CODE
% must be recalculated each time Fc changes
function [y] = wah(x, Fs) F1 = 2*sin((pi*Fc(1))/Fs);
%EFFECT COEFFICIENTS % this dictates size of the pass bands
% lower the damping factor the smaller the pass band Q1 = 2*damp;
damp = 0.05; yh=zeros(size(x)); % create empty out vectors
% min and max centre cutoff frequency of variable bandpass filter
minf=500; yb=zeros(size(x));
maxf=3000; yl=zeros(size(x));
% wah frequency, how many Hz per second are cycled through % first sample, to avoid referencing of negative signals
Fw = 2000; yh(1) = x(1);
% change in centre frequency per sample (Hz) yb(1) = F1*yh(1);
delta = Fw/Fs; yl(1) = F1*yb(1);
% apply difference equation to the sample
% create triangle wave of centre frequency values for n=2:length(x),
Fc=minf:delta:maxf yh(n) = x(n) - yl(n-1) - Q1*yb(n-1);
while(length(Fc) < length(x) ) yb(n) = F1*yh(n) + yb(n-1);
Fc= [ Fc (maxf:-delta:minf) ]; yl(n) = F1*yb(n) + yl(n-1);
Fc= [ Fc (minf:delta:maxf) ]; F1 = 2*sin((pi*Fc(n))/Fs);
end end
% trim tri wave to size of input %normalise
Fc = Fc(1:length(x)); maxyb = max(abs(yb));
% difference equation coefficients yb = yb/maxyb;
end

• EXPLANATION: In the code, the center frequency (or resonant frequency) is varied as
a triangle wave with values between 500Hz to 3000Hz with a step size of 2000/fs
(sampling frequency is 44100Hz for all audio applications).
 OBSERVATION

TIME DOMAIN FREQUENCY DOMAIN

The effect of decaying amplitude also
Gradually decaying amplitude envelope evident in spectrogram. Colour intensity of
indicates amplitude modulation. band at 200Hz decreases gradually.
 TREMOLO
• Modulation is the process where parameters of a sinusoidal signal
(amplitude, frequency and phase) are modified or varied by an audio signal.

• Amplitude Modulation (AM) is defined by:

𝑦 𝑛 = 1 + 𝑎 × 𝑚 𝑛 𝑥(𝑛)
a is the modulation index.
x(n) is the audio carrier signal
m(n) is a low-frequency oscillator modulator.

• When x(n) and m(n) both sine waves with frequencies fc and
fx respectively ,we get three frequencies in the AM wave: carrier,
difference and sum: fc; fc - fx; fc + fx.

• AM = (1 + a cos 2𝜋𝑓𝑥 t) cos 2𝜋𝑓𝑐 t

𝑎
= cos 2𝜋𝑓𝑐 t + *cos 2𝜋 𝑓𝑐 + 𝑓𝑥 𝑡 + cos 2𝜋 𝑓𝑐 − 𝑓𝑥 𝑡+
2

• In tremolo, the modulation frequency is set below 20Hz.

 MATLAB CODE

function [y] = tremolo(x,Fs)

index = 1:length(x);
Fx = 5; %modulation frequency
alpha = 0.5;
trem=(1+ alpha*sin(2*pi*index*(Fx/Fs)))’;
y = trem.*x; TIME DOMAIN
end
Amplitude Modulation by a modulating sine
wave of low frequency is observed. The
amplitude envelope varies sinusoidally.
 OBSERVATION

• To observe the effect of amplitude modulation visually, in the previous code, we

set Fx=1000Hz and Fc remains 200Hz;and in the spectrum, we find peaks at
200Hz,800Hz and 1200Hz.

SPECTROGRAM OF AM WAVE
 FLANGER
• A simple delay filter may be expressed as:
𝑦 𝑛 = 𝑎𝑥 𝑛 + 𝑏𝑥 𝑛 − 𝑀 ; where the delay is of M samples.

• In flange effect, M varies sinusoidally between 0-3ms with a frequency of 1Hz.

 MATLAB CODE
function [y] = flanger(x,Fs) % to avoid referencing of negative samples
y(1:max_samp_delay)=x(1:max_samp_delay);
% parameters to vary the effect %
% set amp suggested coefficient from page 71 DAFX
max_time_delay=0.003; % 3ms max delay in seconds amp=0.7;
rate=1; %rate of flange in Hz % for each sample
index=1:length(x); for i = (max_samp_delay+1):length(x),
% sin reference to create oscillating delay cur_sin=abs(sin_ref(i)); %abs of current sin val 0-1
% generate delay from 1-max_samp_delay and ensure
sin_ref = (sin(2*pi*index*(rate/Fs)))';
whole number
%convert delay in ms to max delay in samples cur_delay=ceil(cur_sin*max_samp_delay);
max_samp_delay=round(max_time_delay*Fs); % add delayed sample
% create empty output vector y(i) = (amp*x(i)) + amp*(x(i-cur_delay));
y = zeros(length(x),1); end
end

• EXPLANATION: Max_delay_dur=3ms
No of samples corresponding to 3ms=Fs*0.003=44100*0.003=132.3=132
Delay of 0.003s means delay of 132 samples
Delay will vary sinusoidally as a whole number between 0 and 132
samples.(0 to 3 ms)
 OBSERVATION
TIME DOMAIN

AM wave with modulation index>1 and

modulation frequency of 2Hz.

FREQUENCY DOMAIN

Intensity of frequency band at

200Hz increases and decreases
alternately indicating sinusoidal
modulation.
 CONCLUDING…
• There are many more effects that are commonly used in guitar processors,
like Reverb, Chorus, Phaser etc. Keep discovering them!

• These effects are readily available in any audio editing software like
Audacity, or any DAWs like FL studio. So even if you don’t have a guitar
processor, that won’t stop you from having fun with these effects!

• Now we know the signal processing that goes behind music

production!There is huge scope for creating new effects. There is no end to
the creativity, both musical and technical, that can be displayed in this
area.

THANK YOU
 REFERENCES
• DAFX: Digital Audio Effects by Udo Zolzer
Copyright q 2002 John Wiley & Sons, Ltd.

• Chapter 7: Digital Audio Effects by Prof David Marshall and Dr Kirill Sidorov,
Cardiff School of Computer Science.

• Multi-Effect Processor for Acoustic Guitar by Eduardo Fonseca Montero

Javier Galdon Romero, Ignacio Perez Pablos, Estrella Merino Siles, Lutz
Ehrig.

• www.electronics-tutorials.ws
• dsp.stackexchange.com
• www.wikipedia.org

• All codes written in MATLAB R2012a and audio edited in Audacity.