Its not analogue

One of the comments voiced by several people at trade shows is that they love the analogue (or analog for our US friends) sound and that a digital synth is not appealing to them. Another was that they had already invested in analogue modular or that they really quite like the patch cables. I fully expected this and had decided that this was not the target audience for riban modular but I recently considered the differences between digital and analouge modular synthesis and I think there is a subject to be explored here.

Digital gives a lot of benefits like reduced cost, simpler multi-voice, etc. I have been sceptical of the “analouge just sounds better” argument because analogue and digital may sound different, but “better” is very subjective and I have wondered how much is perception led by bias, e.g. knowing that a signal path is analogue can lead users to believe it sounds better. But there is a very valid argument for some workflows where digital sounds very different (and to most ears, wrong). This is clearly demonstrated when modulating audio with high frequency signals, e.g. modulating an oscillator with an audio frequency signal. This causes supersonic frequencies to be created which lead to aliasing. Audio band aliasing is usually filtered out of the output signal, but the expected result (of fold-back frequencies in the analogue domain into the audio band) does not happen. Instead, the signal is distorted by the supersonic aliasing and what is heard is perceived to be wrong. This is undesirable behaviour. We want to hear the odd noises that occur when we do odd things like this and there is an expectation (from our prog-rock forebears) of what such workflows should sound like.

This is a tricky issue to resolve but I am starting to condisder how we might provide an analogue like sound using digital technology. This may required some of the DSP to run at much higher frequencies. I think that analogue circuitry is likely to diminish signals in the MHz range so that might be a sensible sample rate range for such DSP. Of course, this increases peak CPU requirements and may prove challenging to implement in COTS hardware, like the Raspberry Pi SBC currently used by riban modular, but we can but try…

Regarding patch cables, well there may be an opportunity to provide some analogue inputs and outputs to allow interfacing with other hardware. That was always in my mind but took a lower priority to making this thing work in its own domain.

I spent some time looking at software emulation of various modules and it looks like the most challenging is filters. It is relatively simple to design a technical filter but this soon fails in ways that are not expected or maybe sufficiently pleasing to hear. Filters can be intrinsically unstable but how instability manifests in the analogue world can differ significantly to that of the digital world, e.g. supersonic oscillation may be limited by the bandwidth of the analogue components and saturate in more subtle ways. This may result in sounds that initially sounded unexpected and possibly even bad but these anomalies have become part of the lexicon of analogue sound synthesis and the expected behaviour under such conditions. Emulating this in the digital world becomes increasing complex with even the smallest of unhandled error having an observable effect on the sound which, due to our learning over decades of what synths should sound like, can be objectionable.

I have built a Moog style ladder filter that works much better than the technical first order one under many conditions. I will probably make both available and see what can be done to avoid either from behaving unexpectedly, although unexpected behaviour can lead to desirable outcomes!