The VPS oscillator that we featured in a blog post recently is one of the main new features of Circle². Future Audio Workshop’s Martin Gould worked on bringing the VPS oscillator to the synthesizer with the help of two of the original VPS paper’s authors, Joe Timoney and Victor Lazzarini. In a behind-the-scenes interview, Martin, Joe and Victor tell us about their field of digital signal processing (DSP) and how they developed the VPS technology.
Hi Joe, Victor and Martin. First, could you tell me a bit about your backgrounds - both professional and musical?
Joe Timoney: I studied engineering and signal processing at university, then went on to work in computer science. I’ve liked a lot of Jazz and Fusion music since I was a teenager, so I’ve always had an interest in music and music making, especially in regards to electronic instruments and synthesizers. I’m from Ireland and my hometown is just down the road from the university. Over the years, Victor and I have collaborated on various research papers and have been involved in the DAFX conference, which we hosted here at NUI Maynooth (pictured above) last year.
Victor Lazzarini: I’m from Brazil and my background is in music. I studied music and graduated with a composition degree, and then did my doctorate in computer music from the composer’s point of view. When I moved into computer music and started working with code, the most logical thing was to learn digital signal processing (DSP). I went into that area and moved to Ireland in 1998. One of my major projects next to the collaborations with Joe has been my work on Csound, which is a programming language for sound.
Martin Gould: I’m originally from London. I was always interested in musical performance and composition, and I learned to play piano for many years when I was younger. I studied mathematics at university, all the way through to my PhD, then started to work at Future Audio Workshop. That’s where I got more involved with the technological side of music.
Why did you enter the field of DSP?
Joe: If you’re interested in audio and making instruments, there are two directions you can go: The analog electronics route or the DSP route. Since the late 1990s, there has been a trend towards digital instruments. Things have really expanded in this field over the last 10 years. DSP is very attractive to us because we can use our mathematical and programming skills to build instruments that musicians can enjoy.
Victor: My interest in DSP came from the fact that I was programming and working with digital audio. I don’t have much interest in the mathematics of analog signal processing, but I am interested in synthesis and how to generate new types of sounds to use musically. I also like how DSP is a large area that goes beyond audio and video - for example, electrocardiography and control systems fall within the discipline.
In the field of DSP, are there any differences between having a background as a mathematician and a background as a music?
Martin: I think the only real difference is that an person’s background and training can influence the way that they approach problems. For example, when working with Joe and Victor on the VPS project, we all approached the same problems in a different way.
Victor: What I find interesting about DSP is that many mathematicians are not familiar with the kind of mathematics that we use. Many mathematicians work on very hard problems, for instance in group theory, but our tools are not commonly used by mathematicians. I have colleagues in the mathematics department who don’t have a clue about the things we work on. They would understand if we explained it to them, but they haven’t learned about these things in their studies. If you talk to people interested in mathematical physics, they know some of our models - but they apply them to the field of astrophysics rather than for making instruments!
How do you think DSP has contributed to modern music?
Joe: DSP has certainly made synthesis more accessible. In the 1970s, an analogue synthesizer cost the same price as a car, so the cost of making electronic music was very high. Thanks to DSP, anyone who has access to a computer can now make electronic music at home. The quality of these synthesisers is also getting better and better. DSP has also contributed to lowering the costs of hardware equipment, so the power you get for your money is far greater now than it was 30 years ago.
Victor: A classic example is frequency modulation (FM) synthesis, which was pioneered by an American composer called John Chowning. He figured out that the technique of radio FM - the transition of sound via electromagnetic waves - could be applied to sound synthesis. That was a great breakthrough, and he wrote a seminal paper about it that was published in the Journal of the Audio Engineering Society in 1973. He used FM synthesis in some of his work, and then Yamaha got interested in the idea, licensed it from him and developed the DX-range of FM synthesizers that we all know and love. The DX7 (chip pictured above) and its family were maybe their most successful range of synthesizers to date. In the early 1980s, the DX7 was everywhere. You would go to a pub, and someone would play a DX7.
Joe: Yes, the DX7 was he first big change that DSP made for synthesis. It was about a quarter of the price of a polyphonic analog synthesizer at the time. It also had 16-note polyphony, so it could be played by keyboarders who liked big, jazzy chords.
Are there any contemporary examples of how DSP is important for music making?
Joe: Recently, there has been a huge amount of work on physical modelling synthesis. The work hasn’t had a massive impact on the market yet, but it’s been a slow and steady growth. For example, some drum synthesizers have physically modelled drums available in them. The reproduction of real acoustic instruments or vintage analog synthesizers is completely DSP-based. DSP has also had a big impact on shaping sound: Having many analog effects units is sometimes difficult and cumbersome, but a DSP-based effect rack gives the possibility of choosing from a great variety of algorithms and allows a much bigger sonic palette.
What was the inspiration for the VPS oscillator?
Victor: We have been working in the general area of non-linear distortion for a number of years and had looked at many ways of generalizing distortion-based synthesis. We were looking at the question of phase distortion when a colleague from Finland visited us here for a few months in 2011. We were sitting in the office talking about the way we distorted phase and said: “Maybe we can generalize this and have one parameter that could change the phase and the waveform in a non-linear way?” That sounded like an interesting idea so we made a prototype in Csound. It was quite an empirical way to start - we worked out all the theory later! That led to the original VPS paper from DAFX in 2011.
Did you run into any problems or obstacles?
Victor: The major problem was the one that we revisit in the new paper with Martin: finding a simple way to describe a VPS waveform’s spectrum. It’s easy to see the different waveforms you can generate with a VPS oscillator, but we didn’t have a compact, straightforward way to predict what the harmonics would look like. Before, we mainly focussed on a more empirical way of thinking, along the lines of “When you change the parameters in this way, you get this type of wave shape, which produces roughly this type of spectrum”. But we couldn’t say exactly what things were. Joe had some ideas on how to solve this, so we sat down with Martin and started work. The practical result was not just the understanding of the spectrum, but also that it informed Martin how to improve the algorithm to avoid aliasing distortion, which can produce a lot of noise in a VPS oscillator.
Martin: The VPS signal contains lots of intricate harmonics at high frequencies. Retaining this sonic character was a big challenge. This is what inspired our collaboration. I had coded up an oscillator based on the original VPS paper, and I found that for many parameter choices the oscillator sounded great. But as I pushed the parameters further, I quickly reached a situation where everything sounded like a horrible screech. I contacted Joe and Victor to discuss the problem, and we set to work trying to understand and solve the problem.
Do you have any ideas for the further development of the VPS technology?
Martin: The new VSP oscillator within Circle is finished now, but the implementation could be generalized in many ways in the future. I’ve already discussed with Joe and Victor about possible ways to introduce additional parameters that could further shape the waveform and make the oscillator even more flexible. I like to think of the VPS oscillator within Circle as the first step towards a bigger goal of flexible non-linear waveform distortion.
Joe: There are also huge possibilities regarding modulation. We explored a couple of ideas in the first VPS paper, but there are many additional ways to generate even more complex waveforms. This is just the start!