[mikepurvis]

I don't think anyone disputes that. The real question is, can a digital filter provide a close enough approximation of that colouring that going through the analogue path is no longer necessary? Are Instagram and Photoshop an adequate substitute for real life film grain?

I say yes, but I'm a software guy. And I've had this argument with a serious pro audio engineer who has a Pro Tools workstation built around a Neve 8816— even though every track is digital, and dozens of digital filters are applied, the tracks each pass through their own DAC connected to this old-school analogue summing mixer, and then are re-recorded on the other side. It seems totally foreign to me, but there are definitely recording professionals still spending a lot of money to set up and maintain these workflows.

[bsder]

> the tracks each pass through their own DAC connected to this old-school analogue summing mixer, and then are re-recorded on the other side. It seems totally foreign to me

There was, at one point, good reason for doing this. Digital summing is really easy to get very wrong. ProTools used to introduce some really nasty artifacts back in the 16-bit days if you had non-linear effects.

24-bit chains have made a lot of this moot. And using 32-bit float as intermediate calculation steps makes even more of it moot. And, increased computer power ... etc.

Then there is the always present "Magic Box X adds tonal coloring that I like." You can't actually argue against this very well unless the box really doesn't do anything at all.

[chipsy]

I think there are some answers to be found in technological histories, both creative tools and otherwise. The main point to consider is that new technologies augment the possibilities - outright replacement is actually the unusual case, because there's always some odd reason to have the old stuff around.

Artistic workflows definitely rank among those reasons; they do tend towards Taylorist efficiency when the artist doesn't want a restriction, but desire to be a maximally efficient artist in every aspect can lead to a deep probing of your motivations for doing any art to begin with, instead of just jotting down what the piece is supposed to be. So most artists will take on at least one technical restriction of some kind, consciously or no. That restriction subsequently guides and shapes everything else they're doing.

So for people who want their focus to actually be built on the limitations of analog, they need analog. Period. And that isn't most people, since there are many other ways to go about setting limitations, and what is usually needed when it's called for is an indication of analog, not its every nuance and imperfection. But like a car enthusiast who can enjoy old and inefficient designs, someone who really cares about analog audio will want to have the experience of the real stuff, despite the painstaking elements of doing so.

edit: And I welcome digital imitations. The newest VA synths really capture the imperfections of old stuff in a way I haven't heard from previous attempts, and I want to use them for that. But I don't want to have an analog box to tote around and maintain.

[mjn]

> can a digital filter provide a close enough approximation of that colouring that going through the analogue path is no longer necessary?

In principle yes, but it's surprisingly tricky, to the extent that even some very simple analog audio things, like '70s-era voltage-controlled-oscillator synths, are only recently able to be emulated reasonably well. The research field is called "virtual analog synthesis". I'm not sure if there's a canonical overview, but some googling turns up this 2014 PhD thesis: http://lib.tkk.fi/Diss/2014/isbn9789526055862/isbn9789526055...

[dharma1]

I've got a bunch of friends who have worked on this at some point, and I know some of the people cited in that PhD thesis. I can still hear the difference, even with state-of-the art current virtual analog emulations, like Diva.

Have to say the gap is narrowing though.

One approach I haven't seen used yet is using the latest AI/neural networks to learn and model the inherent nonlinearities in analog audio

[ShinyCyril]

I wish I had picked that for my masters dissertation - ML + DSP seems like a really interesting area!

[jschwartzi]

I don't think the issue is with the system's ability to replicate the effect. Rather, with a digital system you're now faced with the problem of explicitly representing the effect, where the analog system had the effect built-in. If you can't quantify the effect then you will be unable to replicate it, and if you can then you have to do the work to replicate it. If these effects are a signature of your work, I can see how it would feel like a step backward to suddenly have to quantify them, and it would be very tempting to switch back to the equipment that already applies the effect.

[snarfy]

For audio, yes, digital can replicate analog to 100% accuracy. You cannot differentiate the output signal using an oscilloscope. The signals are identical. People using workflows you describe are a cult, IMO.

[earlz]

I think this is true, but definitely not in real-time. Go observe how fast a complicated analog SPICE emulation is. Typically it's about 1000x slower or worse than real-time, though in the end the wave form might exactly match the actual circuit.

[donatj]

> Are Instagram and Photoshop an adequate substitute for real life film grain?

I thought you were making a good point about how filters were shitty needless imitations and then you threw it all out most unexpectedly.