ALL/DSP

ALL/DSP technology combines equalization, compression and mathematical convolution to transform, in real-time, the frequency response captured by Alas pickups into that of a specific “real” microphone pointed at a “real” classical guitar resonance chamber.

A zoom into ALL/DSP software (3 Modules):

Module A: The signal’s frequency spectrum is analysed with regard to sensory consonance/dissonance (a key variable determining the emotional pleasantness of sound). High/low pass filters and notch equalisation are applied at strategic fixed frequencies to reduce the levels of sensory dissonance in the controlled signal supplied by the Alas guitar.

Module B: The signal undergoes a light audio compression (i.e. dynamic range reduction). The effect is a more intelligible sound that is often qualified as more comfortable to listen.

Module C: Finally, the compressed signal undergoes two mathematical convolution transformations. In the first place, the signal is convolved with an impulse response that captures the natural reflections occurring inside the acoustic chamber of a professionally miked classical guitar. The signal is subsequently convolved with the impulse response of a user-chosen room space (for example: a professional recording studio; ALL/DSP software includes all necessary impulse responses).

These three modules operate with zero latency. As a result the Alas guitar produces in real-time the sound of a professional classical guitar. 

Technical note: The ALL/DSP software technology was originally developed for the OpenGuitar project, as an educational software package that could run in very low performance systems, such as those available for children living in Argentinian deprived settlements called “villas” (analogous to favelas in Brazil). In other words, the algorithm applied had to be optimised for “old and slow” computers. It is well known that a block FFT implementation of convolution is vastly more efficient than the direct form FIR filter. Unfortunately, block FFT processing requires, first, an entire block of input samples to accumulate, and must then perform significant calculations before an output sample is generated. This process causes significant input/output latency, especially in low performance systems, which is undesirable for real-time applications as the one proposed. The ALL/DSP software overcomes this problem by combining direct form FIR filter (time domain convolution) for the early portion of the impulse response, and block FFT processing for its latter parts (Battenberg & Avizienis, 2011; Gardner, 1995). The result is a zero latency convolution technology that outperforms currently available VST software approaches, which rely on the implementation of time-distributed convolution for their operation as VST plugins within host digital audio workstations or DAWs (e.g. such as Digital Performer, ProTools, Logic, Cubase and Ableton Live, among others).