Simple test bed for dsp experiments. Plugins, stacks of plugins, tools, measurements, graphics.

At the moment it can read aiff/wave files, process audio data with simple plugins (LP filter and RMS measurer is present), display signals with pylab and write results to aiff/wave files or export to .mat.

Obviosly it's not a real project but a set of supporting tools for DSP programming, experiments and rapid prototyping. However even now it's helpful to me and hopefully it'll grow a little bit more.

• Plugins. Very similar to those you might seen in VST (except the multichannel input is interleaved, not stripped). You are instantiating them with audio params (samplerate and number of channels) and plugin params (cut off, Q, etc). You can feed data into them using process method. You can query inner states of them using state method.
• Stacks. Chain of plugins, where output of previous one is connected to the input of the next one. Also stacks are capable to read data from signal sources (see below) and write down for you important values using tool called probe. You are instantiating stacks with list of plugins and for each of plugins you can specify a list of state variables to track You can feed input data using process method or by directly connecting signal source to the stack with process_source You can take some measures using probe method or just by accessing probe_results field (which is prepopulated with probe values gathered after processing a batch with process_source).
• Descendants of AbstractReader providing input for Plugins or Stacks. Think Aiff/Wave reader, sine generators, etc. Implemented at the moment: DiracSource, AiffReader, WaveReader, SineSource, CompoundSineSource. endless property of class specifying if the data is generated forever.
• Graph tools like ProbeResultsPlotter which you can use with probe results produced by Stacks.
• You can write results into wave/aiff file (using various bit depths) and export to .mat file which you can later read with Matlab(tm), GNU Octave, Scipy etc. Check writers.py for details.

DSP test bed is working with Python 2.6, 2.7 and PyPy

Core files (plugins.py/stacks.py/readers.py/writers.py) doesn't require anything. processing_test.py is using matplotlib for data visualisation recording_test.py is using PyAudio and basically is heavily facelifted script from http://www.stanford.edu/class/linguist278/notes/interactive-recorder.py (kudos to it's author) with some beat detection based on this code: http://musicdsp.org/archive.php?classid=2#200

Well DSP test bed is built for comfort not for speed. However I have plans to profile it deeply once major APIs will settle down. Also as core of the lib is pure python you can use DSP test bed with pypy.

DSP test bed is using pylab to display signals. Because speed of pylab is an issue when you have tons of data points (like you usually do with audio signals) DSP test bed has simple yet effective signal decimation implemented in pylab_tools.py. Results are pretty neat (and it's muuuch faster).

• DSPMaster[at]free[dot]fr for beat detection
• Christopher Potts (at least it's the only name I found around interactive-recorder.py)

Not so many of external code has been used but it was a great inspiration for me.