Circuit Surgery
Regular clinic by Ian Bell
Distortion and distortion circuits – Part 3
F
or the past two months we
have been looking at distortion
– the effect of non-linearities in
circuits such as amplifiers on the shape
of their output waveforms. Distortion is
often an unwanted characteristic that circuit designers make significant efforts to
minimise. The amount of unwanted distortion is commonly measured using Total
Harmonic Distortion (THD), particularly
in audio applications, which, along with
the basic concepts of distortion, was covered in the first article (PE, June 2022).
Last month, we looked at signal spectra
in the context of distortion. A spectrum is
a plot of signal strength against frequency,
thus showing the frequencies present in a
signal. Periodic waveform can be formed by
adding together a set of sinewaves of various
frequencies and different amplitudes
(known as a Fourier series). Therefore, the
spectrum plot of a periodic signal shows a
set of peaks at specific frequencies.
The spectrum of the output of a linear
circuit will not contain frequencies
which were not in the input (although
the relative amplitude of the original
frequencies may change). If distortion
occurs due to non-linearities there will
be frequencies present in the output
which were not in the input. For a
sinewave input, the additional output
frequencies due to distortion will be at
integer multiples of the input frequencies,
that is, harmonics of the input. This is the
basis of quantifying distortion using THD.
LTspice is able to plot signal spectra and
calculate THD, hence providing insights
into distortion. However, this requires
some care and attention in setting up
the simulation, and this was a key part
of last month’s discussion.
Although often unwanted, distortion
also has its uses, including in sound
processing effects used by musicians – the
most well-known example is probably the
distortional pedals used by electric guitar
players. In fact, this series of articles was
inspired by musical effects projects by
John Clarke in PE over the past year or
two. This month, we are going to look at
circuits which can be used to deliberately
produce distortion for creative purposes.
Clipping
Most musical distortion is obtained from
circuits which deliberately cause the
signal to be clipped – that is, the amplitude
of the signal is limited so that the peaks
of the waveform are flattened. This is also
referred to as saturation. We will recap
what we mean by clipping and look at
a couple of key variations (hard/soft
and symmetrical/asymmetrical) before
considering some of the circuits that can
be used to achieve these effects.
If we plot the transfer function (input
amplitude vs output amplitude) of an
ideal amplifier it will be a perfect straight
line for all possible amplitudes (the grey
line for v(y1) on the plots on the left of
Fig.1). Real amplifiers have a maximum
output amplitude which results in transfer
functions more like those shown by the
coloured traces for v(y2) and v(y3)
on the left of Fig.1. These two differ in
how sharp the transition is between the
linear region at low amplitudes and the
fully limited region at high amplitudes. A
relatively fast transition is referred to as
‘hard clipping’ (eg, for y2). ‘Soft clipping’
refers to a more gradual transition (eg,
for y3). The plots in Fig.1 were obtained
from a mathematically defined transfer
function discussed in the first article.
The grey traces on the right side show
the undistorted sinewave output.
For a given amplifier, or other circuit
which produces clipping, the amount of
distortion depends on the input signal
amplitude. This is illustrated in Fig.2,
which shows the output from the same
transfer function as in Fig.1, but with more
than twice the input amplitude (1.3V peak
in Fig.1, 3V in Fig.2). As amplitude is
increased for a sinewave input, the output
of clipping circuits will tend towards
producing square waves. Softer clipping
functions will produce rounder corners on
the square wave, but this will also occur
if the distorted signal is low-pass filtered.
The transfer function of a distorting
circuit does not have to be symmetrical.
Non-symmetrical distortion will occur
if the transfer function is asymmetrical,
or if a signal with a DC offset is applied
to a symmetrical transfer function. The
latter case is illustrated in Fig.3. This uses
the same signal amplitude as in Fig.1
(1.3V peak) but with a +0.65V DC offset
on the input sinewave. This means that
Simulation files
Most, but not every month, LTSpice
is used to support descriptions and
analysis in Circuit Surgery.
The examples and files are available
for download from the PE website.
Practical Electronics | August | 2022
Fig.1. Hard (top) and soft (bottom) clipping.
65
in the first article, the soft clipping
function used a behavioural source
with the function:
sgn(v(x))*uplim(abs(v(x)),
lim,0.4)
Here, lim is the limit or clipping
voltage. For asymmetrical clipping,
two such versions of the function
were used with limit values of 0.5
and 1.0. The two functions were
selected using an if function:
V=if(v(x)+0.5,a,b)
Fig.2. The same clipping function as in Fig.1 with a
larger input signal.
Here, a and b are the two limiting
functions. The LTspice function
if(x,y,z) returns y if x > 0.5,
else z. Refer to the first article for
the LTspice schematic.
Peddling terminology
Distortion effects pedals used by
musicians are often described
as providing ‘boost’, ‘overdrive’,
‘distortion’ and ‘fuzz’. The term
overdrive was originally used to
describe turning up the volume (gain)
of a guitar amplifier sufficiently to
cause it to clip, thus giving a fuller,
grittier sound. In the early days of
electric guitars (before the widespread
availability of effects pedals), this
was necessary to obtain a distortion
effect in the amplifier itself. These
early amplifiers were built using
‘valve’ – vacuum tube – technology
(not transistors/semiconductors),
which produced a relatively soft
clipping due to the characteristics of
Fig.3. Asymmetrical distortion obtained by applying the devices. Thus, a pedal marketed
a sinewave with DC offset to the transfer functions as providing overdrive will typically
provide soft clipping and aim to
shown on the left-hand side in Fig.1.
emulate the sound obtained by
overdriving a valve amplifier.
the negative peak is at –1.0V and only
A ‘boost’ function simply provides gain
just reaches the limiting voltage of the
(amplification) without any clipping. Tone
clipping function. The negative peak is
controls may also be provided to control
not significantly distorted, but the positive
frequency response (provide different
peak is heavily clipped. The signal in Fig.3
gain at different frequencies). As noted
has DC offset, but this is easily removed
above, and seen in Fig.2, increasing signal
using a coupling capacitor.
amplitude produces more distortion,
Fig.4 shows an asymmetrical soft
thus a signal boost can be used to help
clipping transfer function and the
force an amplifier to clip, or increase the
resulting output waveform for a 1.3V
amount of distortion provided by the
sinewave with no offset. As discussed
Fig.4. Asymmetrical distortion from an asymmetrical transfer function.
66
clipping circuit (distortion effects unit)
connected to the output of the boost unit.
The terms ‘overdrive’, ‘distortion’ and
‘fuzz’ all refer to signal distortion caused
by clipping. As noted above, overdrive
typically refers to softer clipping. Effects
units marketed as providing distortion
and fuzz typically provide harder clipping
and/or drive the clipping circuitry at
relatively high amplitudes, so are likely to
produce outputs which tend more towards
square waves, with ‘fuzz’ often implying
a more extreme effect than ‘distortion’.
As well as introducing distortion the
nonlinear transfer functions of clipping
circuits (eg, the left side of Fig.1 and Fig.4)
affect the dynamics (volume range) of the
output signal. Soft clipping causes smaller
signals to be amplified more than larger
ones – a process known as compression.
For a guitar player (for example) this
causes sustain of the notes played. When
a string is plucked on an instrument the
volume decreases as the vibrations decay.
If the gain of the signal increases with
decreasing signal this output will tend to
remain at a more constant volume than
the signal directly from the instrument,
resulting in longer lasting notes. This
relates to something like the situation in
Fig.1, assuming the sinewave shown is
around the typical maximum amplitude of
a note. In this situation, only the loudest
parts of the sound will be distorted. If even
relatively quiet notes result in the more
extreme clipping like that shown in Fig.2
then there will be relatively little change
in output volume for different inputs.
As discussed in depth in the previous
article, distortion changes the frequency
content of a signal by adding harmonics
and (for other than sinewaves) other
frequencies. To obtain a good sound
from a distortion effect, filtering of the
higher additional frequencies using a
low-pass filter may be desirable. Also,
using frequency-dependent circuitry,
distortion can be applied differently at
different frequencies. Many distortion
units have some form of tone controls
(filters) which can adjust the sound, along
with the amount of distortion applied.
Diode characteristics
Clipping circuits are essentially amplitude
(or voltage) limiters. Diodes are well
known for having a near constant
voltage across them for a wide range
of currents and therefore naturally
act as voltage limiters. We can make
distortion (clipping) circuits using
diodes instead of forcing a full
amplifier into saturation. Diode
clipping is the basis of many, but
not all, distortion effects units.
We can plot the current against
voltage (IV) characteristic of diodes
using a DC sweep in LTspice – for
Practical Electronics | August | 2022
Fig.5. LTspice circuit to obtain diode IV characteristics.
example, using the setup shown in Fig.5. Here we investigate
the IV curves of a very commonly used silicon diode, the
1N4148 and also a germanium diode, the OA91. Germanium
diodes were widely used before silicon versions replaced
them in most application. Silicon diodes have much higher
performance in terms of reverse leakage, maximum reverse
voltage, stability, maximum operating temperature and cost.
However, germanium diodes still have some niche uses and
one of them is in distortion circuits. The OA91 was somewhat
arbitrarily selected as an example to use here because the model
was found after a quick online search. The current set of diode
models provided with LTspice does not include any with
type=germanium. The .model statement is shown in Fig.5.
Fig.6 shows the results from the simulation in Fig.5. There
are two key points. First, the germanium diode conducts at a
lower voltage than the silicon one – at around 0.2-0.3V rather
than 0.6-0.7V. Second, the germanium diode has a less abrupt
rise in current as forward voltage increases – this will equate to
a softer response when the diode is used in a clipping circuit.
Distortion effects circuits
Fig.7 shows the block diagram of a basic distortion effects unit.
This is more of a concept illustration of what happens along
the signal path rather than the true schematic structure as the
functions may be combined in some implementations. It is
also common to have a bypass switch to route the signal past
the whole effects circuit to allow the player to switch the effect
on and off when needed (not shown). More complex units may
have more complex signal paths; for example, to select or mix
different distortion effects.
The signal path in Fig.7 starts with a buffer which provides
the correct input conditions for the intended source (eg, electric
guitar pickup). It may include some filtering, for example to
remove very high frequencies such as radio interference. The
gain stage is a variable gain amplifier which drives the clipping
circuit. As discussed above, and shown in Fig.1 and Fig.2, the
signal amplitude into a clipping circuit affects the amount of
distortion produced – thus varying the gain varies the amount
of distortion for a given input. The clipping stage is a nonlinear
circuit with a limiting transfer function (as in Fig.1 and Fig.4).
It may be hard or soft, symmetrical, or asymmetrical, or have
switches selecting different options, or controls to adjust factors
such as asymmetry. The input buffer function may be provided
by the same amplifier as the gain stage – it is not necessarily
a separate amplifier.
After the clipping stage we have our distorted signal, but its
amplitude may need adjusting to be suitable for the destination
(amplifier input, or next effects unit in a chain) – this is provided
by the output buffer in Fig.7. Some distortion effects also feature
tone adjustment – typically a filter to cut high frequencies by
a variable amount. In some designs the output stage may not
include an active amplifier, and level and tone adjustment may
be part of the clipping circuit.
Diode clipping
The circuit shown in Fig.8 is a typical diode clipping circuit
used for distortion effects. Not all components are used in all
versions and variants exist. The input (typically from a variable
gain amplifier, which sets the distortion level) is coupled
Fig.8. Diode
clipping circuit.
In
C1
R1
D1
D2
C2
R2
Level
Out
Fig.6. IV characteristics for a silicon (green) and a germanium
(red) diode.
Buffer
Gain
Clipping
Buffer
In
Out
Distortion
Fig.7. Block diagram of distortion effects unit.
Practical Electronics | August | 2022
Level
Tone
Fig.9. LTspice schematic for simulating the diode clipping circuit
with two different diodes.
67
to the left. LEDs have forward voltages in the range 1.2 to
4V, depending on the colour.
R1 limits the current in the diodes, and R1, R2 and C2 form
a low-pass filter which can be used to remove the higher
harmonics from the distorted waveform. Potentiometer R2
allows a proportion of the diode signal to be output, providing
level adjustment. This can be directly used as the output or
passed to an amplifier circuit which may have a tone control.
R1 and R2 are typically a few kilohms and C2 is typically a
few nanofarads.
Simulation example
The circuit in Fig.9 is an LTspice version of Fig.8 without lowpass capacitor C2. This is so that the full effect of the diodes
on the harmonic content can be observed. The input coupling
capacitor (C1) is also not included – it is not needed with an
ideal signal source, and specifically in Fig.8 we need DC coupling
to obtain the transfer function. The results in Fig.10 show that
the transfer function is similar to the soft function in Fig.1,
with the germanium diode circuit having a softer response.
If we change the voltage source in the circuit in Fig.9 to
produce a sinewave and run a transient simulation instead
of a DC sweep:
V1 source: SINE(0 1.5 500)
Fig.10. Simulation results from the circuit in Fig.9 – the transfer
functions of the clipping circuit with different diodes (top:
silicon, bottom: germanium).
via capacitor C1, which removes any DC offset, prevents the
diodes from disrupting the bias of the amplifier stage (where
applicable), and reduces gain at low frequencies. C1 is typically
a few microfarads.
The maximum output signal is limited by (clipped) at the
forward voltages of the diodes (D1 and D2). Two diodes are
used to cover the positive and negative half cycles of the
waveform. Asymmetrical clipping can be achieved by using
different types of diode for D1 and D2, or by using a different
number of diodes in series in each direction. Diode options are
standard silicon didoes, Schottky diodes, germanium diodes
and LEDs. Silicon and germanium diodes were compared
above. Schottky diodes have a similar characteristic shape to
silicon, but a lower forward voltage – closer to germanium at
around 0.2V – this is like shifting the silicon curve in Fig.6
Simulation command: .tran 0 50m 0 10n
We get the results shown in Fig.11, which shows that the silicon
diodes result in harder clipping (or a more square-wave shape)
than the germanium diodes. The results from the transient
simulation can be used to obtain spectra of the output signals
using the LTspice FFT function (taking account of the requirements
for doing this effectively, which were discussed in detail last
month). The results are shown in Fig.12; using a linear frequency
plot here makes it easier to see which harmonic is which. They
show that both circuits add odd harmonics, but not even ones.
The relative levels of different harmonics are different for the
two types of diode. The difference in harmonic content will
result in different effects on note timbre – implementing the
circuit with different diodes will sound different.
In order to make the spectra easier to compare they have been
normalised so that the fundamental (input sine frequency of
500Hz) is at 0dB in both cases. This was done by measuring
the fundamental peak on the initial result and calculating the
scaling factor required to shift up to 0dB. The scaled signal was
then plotted. For example, initially the fundamental peak of the
silicon circuit was at −6.34dB. The scaling required is 1/10−6.34/20
= 2.075, so the trace was edited to become 2.075*V(outsi)
– right click the trace title to do this.
Listening test
Given that we are discussing sound processing circuits, it
would be useful to be able to listen to the results. Of course,
ultimately, we’d want to do listening tests on a real circuit if
we were actually developing a distortion effects unit. However,
just for fun, or for quickly investigating a range of possible
circuits, LTspice can provide listening opportunities via WAV
files. For example, you can use a short recoding of (say) a note
from a guitar as the input, and listen to the simulated results.
We discussed WAV files in detail in July 2020, so we will
be briefer here. For input, change the ‘value’ of an LTspice
source to the form:
wavefile=filename
Fig.11. Transient simulation results for a sinewave input to
the diode clipping circuit with different diodes (top: silicon,
bottom: germanium).
68
Where filename is the name of a WAV file in the same folder
as the schematic, or the full path to the file if it is elsewhere. By
Practical Electronics | August | 2022
measure the peak level so the scaling
can be set appropriately. The circuit in
Fig.13 is a version of Fig.9 configured
for WAV input and output. This was
tested using a guitar sample found
online (freewavesamples.com) – the
change in timbre of the output and the
difference between the two diode types
could be heard.
Other circuits
Fig.12. Spectra of the signals from Fig.11 (top: silicon, bottom: germanium).
default, the signal used will be the first
channel in the WAV file (typically there
are two channels for stereo). To write a
WAV file you need to place an LTspice
.wave directive on the schematic. For
audio signals it is best to configure this
to a standard format such as stereo 16-bit
44.1 kHz (as used for CDs). For example,
to output voltage out1 to both stereo
channels in file output1.wav use:
.wave output1.wav 16 44.1K
V(out1) V(out1)
One key issue is that the WAV file
maximum amplitude is 1V. We typically
have to scale both the input and outputs
to fit with the circuit amplitudes. This is
straightforward to do using behavioural
sources, but for the output an initial
simulation will often be required to
The diode clipping circuit discussed
here is not the only circuit that can be
used. Another common approach is
to use diodes in the feedback of an op
amp amplifier – this is similar to the
logarithmic amplifiers discussed in
December 2021. A typical configuration
is shown in Fig.14, but as before there
are variants of this circuit. In this circuit,
R1 and R2 set the gain for AC signals as
a conventional non-inverting amplifier
(gains of 10s to 100s are typical), but
C2 blocks DC giving 100% feedback
and hence unity gain for DC (not in all
variants). The diodes limit the output
amplitude to cause clipping. As before,
different diodes, or numbers of diodes,
can be used for asymmetry. R2 is typically
a variable resistor – changing the gain
controls the amount of distortion. C1 is
typically a small capacitor to reduce gain
at high frequencies to reduce the chance
of instability. The shape of the transfer
function of this circuit is a little different
from the diode clipper discussed above,
so it will sound different. Some distortion
pedals combine both circuits.
The Nutube Guitar Overdrive and
Distortion Pedal by John Clarke (PE,
March 2021) uses a different approach
– by driving an amplifier into clipping
rather than using diodes. The amplifier is
a common-cathode stage built using the
Nutube low-voltage triode. Asymmetric
clipping is achieved by varying the DC
bias of the input.
R2
C1
D1
D2
–
In
U1
Out
+
R1
C2
Fig.13. LTspice schematic for simulating the diode clipping circuit with WAV file input and output.
Practical Electronics | August | 2022
Fig.14. Op amp-based
distortion circuit.
69
