A bunch of research
Posted by Stefano at 20/09/2022, 16:11:45 UTC.
So, what are we doing at conferences?
The DAFx conference is arguably the most important purely scientific conference worldwide on music DSP, and it's certainly the most fun to attend. The legend says that it was started in 1997 (I was still in primary school), originating from a EU project. Since then, many of the most important scientific breakthroughs in this field were published and presented there. All published papers go through a proper peer review process, hence the quality of works tends to be relatively high. It is organised by a different hosting institution each year and consequently held in different places. Scientists and institutions from all over the world participate.
This year we presented two papers there. One is named "A General Antialiasing Method for Sine Hard Sync" and describes a new technique to reduce aliasing noise when digitally generating hard-synced sinusoid signals. The math behind it can be scary, but the basic idea is relatively simple: the corresponding continuous-time signal is described analytically, then a continuous-time FIR lowpass filter is virtually applied through convolution, and finally the output signal is sampled into the digital domain. We performed the calculations for different types of filter kernels and expressed the results in terms of residuals to make them easier to apply in practical applications. The resulting method is much more efficient and effective than the best previously methods (as a side note, at the time of writing we were not aware that this paper also dealt with a somewhat related problem, but it's quite far from being actually useful for sine hard sync).
The other paper is named "Antiderivative Antialiasing with Frequency Compensation for Stateful Systems" and is based on the observation that it is possible to apply frequency response compensation to nonlinearities treated with the antiderivative antialiasing technique based on IIR filters, as long a simple first order filter is used. This allows to use such method also in stateful (recursive) systems without side effects. Unluckily the method has problems with higher order filters, so an investigation on why that is the case is included and it seems like the virtual analog-to-digital reconstruction process is to blame. In other words, this is a start and more research is needed.
As usual prizes were awarded for the best papers, I guess according to the Commitee, yet I also wanted to indicate three more that caught my interest:
- "Deforming the Oscillator: Iterative Phases Over Parametrizable Closed Paths" by Georg Essl;
- "Combined Derivative/Antiderivative Antialiasing" by Martin Holters;
- "Time-Varying Filter Stability and State Matrix Products" by Kurt James Werner and Russell McClellan.
XIII CIM 2022
CIM stands for "Colloquio di Informatica Musicale" (that would be Colloquium of Music Informatics) and is a national-level Italian conference organised by the AIMI association that involves mostly young researchers and musicians. The conference has a very long tradition, with its history beginning no later than 1976. All contributions are peer-reviewed. This year's edition is hosted by Università Politecnica delle Marche and Acusmatiq MATME and our dear friend and colleague Leonardo Gabrielli is heavily involved in it. It will take place in Ancona, 25th—28th October 2022.
Paolo submitted a paper describing the implementation details of the Ciaramella Web IDE and we'll make it freely available for download right after the conference. Then, I'll have a talk on everyday technical matters when crafting music DSP algorithms and engines. And since we were at it, we decided to also officially sponsor the event.
The Audio Developer Conference has quickly established itself as one of the most important industrial events for music technology development. IIRC, it stemmed from the 2015 JUCE summit into its first edition in 2016 and is annualy held in London. Save this year's dates: November 14th-16th.
This is the first time I participate and I don't yet know what to expect in detail. Anyway, I'll have a talk on "LEGOfying audio DSP engines" with the pretense of teaching how to make DSP development scale.
Extra: an IEEE paper
As if all of this was not enough, we also found the time to co-author quite an important paper with Leonardo Gabrielli and Stefano Squartini, both from Università Poltecnica delle Marche, named "Antiderivative Antialiasing for Arbitrary Waveform Generation", which shows how to apply antiderivative antialiasing methods to waveform generation. The method can be used to generate any signal defined by a piecewise-linear waveform (that is, it can also be applied to sampler oscillators), with arbitrary antialiasing filter order, and without requiring to compute ad-hoc analytical expressions. The results we obtained are, IMO, astonishing.