I've commented a number of times in the past on my desire to create a long-form modular synth composition. By 'long form' I mean something around a traditional album side. There have been a number of technical limitations that stood in the way, many of which have been unlocked by Volta. Some of the remaining issues center around a working process, project management and the realistic limits of what the software and hardware can handle.
Calculus, at three minutes or so, generated over 8GB of project audio data. I'm in the very early stages of this project and I've already accumulated over 4GB. I fully expect the entire thing to weigh in over 100GB when I'm done - for a single piece of music. The amount of this data isn't important, but it should give you an idea about the scale of the project.
However, the larger a project is, the more unwieldy it becomes. Managing and organizing the project is a project itself. I have to work on it in pieces and sections as my G5 is already groaning and creaking under the strain. If you're assembling a complicated, multi-layered modular piece one monophonic line at a time, you have to have a macroscopic view of the project, and know exactly where you're going. I've never attempted something this long using this technique and I'm unsure what sort of problems this will present beyond complicating the already considerable long view.
Even the best VCAs on my modular probably spec out around 90dB signal to noise ratio, which should be representable within a 16-bit audio file. (I haven't actually tested this, it is a guess, but I'm not going to lug home the Audio Precision) But, I record all my tracks at 24-bit for some additional headroom, or to allow for some sloppy recording technique by recording at lower levels.
Typically, when I do the final mix, I'll generate a 24-bit master and a 16-bit master as a matter of process. When summing multiple 24-bit files (or, in my case, a possibly absurd number of 24-bit files), you're effectively generating audio data with meaningful resolution that exceeds 24-bits, which is one reason all native-processor DAWs use a 32-bit floating point mix bus architecture.
This has a lot in common with shooting in RAW on a digital camera. There total dynamic range exceeds our ability to perceive, (or represent) at a single time. Shooting in RAW allows me to shift or compress the dynamic range to suit my needs. I've found I have 1-2 stops in either direction of the captured exposure that I can use without any loss of detail.
Back in the audio world, that full 32-bit floating point resolution represents a dynamic range that can't be described in terms of human experience. We can only examine a portion of this range with our ears from the threshold of sensation to the threshold of pain. Typically, we take the output of the 32-bit floating point mix bus, and use compressors and limiters to squish the dynamic range into something reasonable, and quantize the output to a 16 or 24-bit word.
OK, so what?
When DP changed to broadcast wave files for its native file format (yay), support for 32-bit floating point audio files was added as well. You can even record all your tracks using this format if you you enjoy wasting disk space. So what is a 32-bit floating point file good for?
In the days when the typical recording studio recorded on 2" 24-track analog tape and automated analog consoles, music production was typically fractured into discrete processes: recording (basics and overdubs), mixing and mastering. DAWs made it possible to collapse recording and mixing into a single organic process. While productions may still follow the traditional method, more people are embracing this new process. But, what about mastering?
Mastering is exerting more of an artistic influence on the sound, and this is especially true for electronic music where creative mastering can produce dramatically different results. I see mastering has creeping its way into the production process like recording and mixing are now. It begins innocently enough by strapping a limiter on your mix bus to prevent clipping. Then, it starts to become part of the 'sound' of the track. Admittedly, mastering is one of my weakest points, and, historically, under the traditional advice I just figured I'd leave it to someone who knows what they're doing. It makes good sense, right? Another pair of ears, yadda yadda yadda. Well, the reality is I can't afford mastering any more than I can afford a new computer, so this is something else I can do myself and, even better if mastering is drifting farther from the technical to the artistic side of things.
There are many points across the editing process where the results are interesting, but unfinished. Then, I may go 'too far' and need a reference. The logical response to this is to bounce a mix down to preserve that moment. But, what if I wanted those mixes to be more than a reference, to be… useful. Perhaps I want to edit a some of these pre-composed sections together in a new composite? If I want to do this, I want to operate on the pre-mastered result. If only there was some way to capture the raw output of the 32-bit floating point mix bus.
TA DA! A use for 32-bit floating point audio files. Now I have a lot of flexibility and options and can still experiment with mastering settings at a later time.
OK, moving on to an issue related to this project, but completely separate from this bit depth nonsense. Yesterday, I was lamenting the lack of delays and reverbs without an automatable method to null out the delay lines.
Native Instruments Spektral Delay had a 'delay matrix reset' switch, but since I loved this product dearly, clearly,
it needed to be discontinued. (I'm sure NI had a good reason for axing this product, like, the amount of work required to bring it into the next architecture would be equal to a total rewrite and it wasn't pulling its weight to begin with) The delay matrix reset switch wasn't automatable, but it was there. Typically, this feature is a technical convenience, like for clearing delays prior to starting playback from a different location, but I want to use it artistically.
Automating the input or send doesn't work entirely as you still have to wait for the delay lines and regeneration to clear out. Automating the output or wet/dry mix doesn't work entirely because you want to take advantage of the new delays being created without the leftover garbage from the prior harmonic content clouding things up. If the effect does not have this feature, there is no satisfying way to fake it using these methods.
The only way I've been able to achieve this effect in the past is by recording the delay output segment by segment and editing them together which is as tedious and time-consuming as it sounds, even more so if you're batch processing 20-minute modular tracks. Cough, cough. Speaking of which, I've found, sadly, that my method of recording music using a modular is very compatible with doing laundry if you plan ahead.
Ideally, you'd be able to trigger the delay reset using a number of methods, but if it responded to MIDI note on messages, I could simply route a MIDI track to the effect. This would allow me to clear the delay lines with each new note without creating any additional automation. As a bonus, this could be played in real time. If you know of an Audio Unit that does this, let me know.
