digital audio neutral network (2019)
Trying to come up with a neural network specifically suitable for digital audio I managed to devise a fully connected graph that allows direct translation of audio signal processing schemes into this neural network.
floating plate reverb/cymbal (2019)
Thinking about possible suspensions for an experimental plate reverb it seemed a good challenge to disregard this problem by having a floating plate. This can be acheived by an active magnetic system that compensates for the forces acting on the plate, keeping it in place, as much as possible. The electronic compensation signal becomes a useful audio signal making this setup a potential floating electronc cymbal. Changing the magnetic flux using another electromagnet also excites the plate making it an experimental plate reverb effect.
botanical passive panning (2018)
Peculiar properties of passive summing mixers can be mimicked by using various plant leaves. By puncturing them with electrodes some variety in audio transfer can be achieved. A fresh aloe vera leaf has almost flat response with a slight high frequency roll-off and some harmonics produced in response to bass frequencies. As the leaf dries and gently retracts from the electrodes, the transfer gets worse. In this way one can control the fidelity of this passive botanical mixer. Different arrangement of electrodes and careful layering can yield passive panning of multiple sound sources.
recursive soundcard testing (2016)
The biggest challenge in testing seems to be the testing of the testing equipment. The self-referential nature of this problem may pose a technical challenge. Even the accurate testing results may not be sufficiently informative in choosing a soundcard. The audible difference is often minute among quality conversions. In order to be able to audibly assess the difference I designed a testing procedure that repeatedly plays and records the sound through a cable connecting the outputs and the inputs. This setup has allowed me to clearly hear that a PrismSound soundcard does beat RME, after about hundred iterations. Also, many cards are not calibrated and the re-recording has to be made with adaptive volum levels. Here is the link to SuperCollider code that performs this recursive test.
rhythmic derrangement (2014)
Having spent a great deal of time manually transforming music into more complex rhythmical schemes, it appeared that this process could be automated. Using Max for Live environment a technique was developed for rhythmical derangement and cross-mixing that yielded many pieces found in the music section. Here is a link to the published paper.
harmonic stability of soft cliping (2012)
Experimenting with different soft clipping algorithms it became clear that they behave differently based on the input amplitude. Driving the algorithms with analytic signals with increasing amplitude, it becam apparent that certain algorithms produce unstable harmonics. Details on this finding and the metric set up by Sean Enderby can be found here,
differential inversion subharmonic non-linearity (2010)
Octave halving seemed like an exciting challenge with good musical potential. Sketching around a sinewave with a pen allowed me to conceptualise an inversion algorithm that keeps a continuous gradient by oeprating on the differential. The details are published through AES and the SuperCollider code is here
zero-padding challenge (2009)
Having doubts about the theoretical stance that zero-padding will not improve the capacity of FFT to resolve partials I coded an example in SuperCollider showing two sufficiently close sinusoids to be separable only using zero-padding. Julius Smith claims that this does not contradict the theory, however it remains to be demonstrated how this example could work without zero-padding.
sonic quiz (2004)
Inspired by the complexity of psychoacoustics an experimental setting was explored to reflect on the primacy of different acoustic parameters.
recursive geometrical synthesis (2002)
Having internalised textbook synthesis I programmed a framework for recursive geometrical synthesis in SuperCollider. Each method name begins with lowercase letters indicating the inputs and ends with the indication of outputs. Class R contains methods that yield arrays while F class produces functions. The interest lies in constructing growth and limitation procedures in a recursion such that a simple envelope signal can grow into a chaotic occurence in both timbre and rhythm. Here you can find the code for this audio example.
warm sound (2000)
Rather difficult to reproduce without the needed setup, nevertheless a remarkable early invention worth documenting, is the warm sound. Two long strings were stretched to yield around 4 ohm DC resistance such that AC currents can be driven with audio amplifiers. Placed in a permanent magnetic field these strings faintly reproduce the audio signal they receive. Stronger currents induce heat and placement of the strings in a pipe with a hole makes that hole the origin of warm sound. It can be argued that there is the heat waves are modulated by the acoustic signal. The sensation for the ear resembles gentle earwax melting.