martinholters / ADAA_Examples.jl

Example code for "Antiderivative Antialiasing for Stateful Systems"

Example code for "Antiderivative Antialiasing for Stateful Systems"

The example code is provided as a Julia package which can be installed with

import Pkg; Pkg.add(Pkg.PackageSpec(url="https://github.com/martinholters/ADAA_Examples.jl.git"))

Then, load the package with

using ADAA_Examples

which brings two sub-modules, DiodeClipper and TubeScreamer into scope.

They both provide three functions rumsim, runsimaaout, and runsimaa to run the respective model without antialiasing, with antialiasing of the output only, or with full antialiasing, respectively. They all take two arguments, the sampling rate (in Hz) and a vector of input samples.

To e.g. recreate the spectra of the tube screamer at 88.2kHz, first install packages for DSP (for window functions), FFT, and plotting:

import Pkg; Pkg.add(["DSP", "FFTW", "Plots"])

Then do

using ADAA_Examples, DSP, FFTW, Plots

f0 = 1000
fS = 88100
u = [sin(2π*f0/fS*n) for n in 1:round(Int, fS)]
w = hanning(length(u))
w ./= sum(w)/2
f = range(0, fS/2, length=length(u) ÷ 2 + 1)
kmax = floor(Int, 20000/(fS/2)*length(f))
f = f[1:kmax]

y = TubeScreamer.runsim(fS, u)
Y = rfft(w.*y)[1:kmax]
p1 = plot(f, amp2db.(abs.(Y)); legend=false, title="no aliasing mitigation", ylim=(-130, 10));

y = TubeScreamer.runsimaaout(fS, u)
Y = rfft(w.*y)[1:kmax]
p2 = plot(f, amp2db.(abs.(Y)); legend=false, title="output aliasing mitigation",ylim=(-130, 10));

y = TubeScreamer.runsimaa(fS, u)
Y = rfft(w.*y)[1:kmax]
p3 = plot(f, amp2db.(abs.(Y)); legend=false, title="full aliasing mitigation", ylim=(-130, 10));

plot(p1, p2, p3, layout=(3, 1), size=(600, 800))

Note that these functions setup the model and initialize its states on every invocation, so they are not suitable for successive processing of signal frames.