I also provided some MAX objects to remote-control Sound Processor via MIDI with MAX. Click below to get a whole directory of helpful MAX patches (will be transferred as one selfextracting archive in BinHex4 format).

Short description of available DSP algorithms:

Tap delay

The tap delay distinguishes itself for its very long delay lines: the maximum delay length is only limited by the amount of RAM installed in the computer. Elektroacoustic works, that in the early days were performed with use of analog tape-delays (like "Solo" from 1968 of Karlheinz Stockhausen), can now be realized digitally.

Besides very late echo's, the "tap delay"-algorithm can be used to create resonator-effects (very short delay lines with high feedback) and some primitive reverberation-effects.

The tap delay consists of two separate delay channels: one left and one right channel. Each delay channel is preceeded by a (variable) noise gate to avoid the accumulation of inevitable input-noise, in case you are using high feedback-levels. Each delay has faders for input, bypass, output, and feedback levels. From one input signal, you can tap up to six output signals, which may be directed to the left or right output channel.

Pitch shift

This algorithm (also known as "harmonizer") provides two independant pitch shifters (one on the left channel, the other on the right channel). You can transpose the pitch of the left and right input signals, from two octaves downwards up to one octave upwards. You may type frequency ratios (between 0.25 and 2), use the sliders, or send MIDI messages, to change the amount of transposition.

All signal processing is performed in the time domain (apart from a sort of autocorrelation-measurement, no frequency analysis is done). The algorithm basically consists of 2 parts: A third order interpolator, which does the actual pitch shifting, and a kind of autocorrelation procedure which controls the time correction. When the pitch shift ratio is less than 1, the audio buffer will be steadily growing and once in a while a certain amount of samples is cut away. In the opposite case (when pitch shift ratio is greater than 1) the audio buffer will be steadily shrinking and sometimes a certain amount of samples is repeated.
Theoretically, this method only works properly for quasi-periodical signals. And that is the main disadvantage of this method: It doesn't work that well for less harmonic spectra. It's the best to apply this effect to single voices, do not use it for polyphonic music or chords.

Ring modulation

There are 2 different ring modulators: one for modulating two external signals with each other (multiplying left and right inputs), and one for modulating a stereo signal with an internally generated sine wave. Both ring modulators contain good digital filters (on the input signals) to eliminate unwanted digital by-products (sum-frequencies above the Nyquist-frequency). Furthermore, one can modulate with rather high sinewave frequencies (up to 2400 Hz). And to create tremelo-effects, one can use very low (sub audio) frequencies.

Spectral analysis

Real time spectral analysis over 7 octaves on a logarithmic frequency scale. Frequency resolution up to 24 bands per octave (50 cents per band, 168 bands in total).

Real time pitch to MIDI

Universally applicable pitch-recognizer based on a spectrum analyzer and a neural network. This method also provides the recognition of multiphonics, non-harmonic-sounds and other special playing-techniques. The first prototype is included in Sound Processor, however, much work has still to be done before it will be really applicable within a musical context (with regard to reaction-time of the network and writing an extensive user-manual on how to train the network).

Pieter Suurmond, 1996