Actually, that's another myth.
I did a test a few months ago where I intentionally clipped a recording of an acoustic guitar in the digital domain by raising the level to be 2 dB above 0dBFS. Then I lowered the volume to match the original for a fair side by side comparison. Yes, it sounds slightly different from the original - not unlike slightly overdriving analog tape - but it's not horrible as some would have you believe. I'll be glad to share the file if anyone cares. Or, better, just do the same experiment yourself in any two-track audio editing program.
Well, in this we have to disagree in terms of perception. I find even 2dB of digital clipping a fairly unpleasant sound but that's just my opinion. However, it's also worth saying that the effect gets worse rather rapidly as you go above your 2dB test point. In any case, I don't think anyone here is arguing that digital clipping is a "good thing" or that proper gain staging to ignore it shouldn't be one of the absolute basics of recording.
Well, don't ignore that altogether because it's hugely important and relevant to this discussion. But I agree there's no need to record as close to zero as possible. I've been arguing that for years. Even with "only" 16 bits, the noise floor is 20-30 dB quieter than analog tape. So recording with peaks around -10 or even lower is fine. I object only when people claim that recording at -20 dB sounds "better" than recording closer to zero, because it doesn't. Unless gain-staging is wrong somewhere else in the chain. Recording at conservative levels is mainly for safety and convenience.
I only disregarded 32 bit float operations for the sake of keeping the discussion in this thread simple. In my own "real world" I use a DAW that works natively in 32 bit float. In any case, that was just a detail--I fully agree that, as long as clipping is avoided, I can't hear a quality difference between -20 and -1. Actually, that's not quite true. If I have to apply significant gain to the -20 signal, I may well start to hear background noise (depending on what was used earlier in the chain).
Absolutely. I work entirely ITB, so this doesn't affect me. But outboard gear doesn't enjoy the luxury of 32-bit floating point math, so levels in and out definitely matter.
I'm in a slightly "hybrid" position on this. In my studio, I'm 100% ITB. However, much/most of what I do is sound for use in live theatre--so very often my mixes (even sometimes stems) are played back live which implies at least some analogue stages.
Agreed, and I mentioned that earlier in this thread.
I disagree. The most compelling reason for lower levels is to keep the analogue components in the signal path from producing distortion on input and, in the digital domain, to prevent inter-sample distortion at the DA output. Clip point is not necessarily the point at which distortion will be produced. It may manifest many dB BELOW clip point, depending on the quality of the components.
No disagreement that analogue levels are important. If, to satisfy a silly quest for your digital recordings to peak at some silly-high level you have to push the analogue input stages too high then clearly this is a bad thing. However, that's not the sort of situation I was referring to. At least on my set up I could achieve overly high digital levels without pushing my analogue stages at all. No, I obviously don't do that--but, for the sake of my previous post, I was assuming sensible gain staging throughout the system. Indeed, gain staging is something that I get pretty OCD about--it comes from far too many years in the broadcast industry where we were constantly aligning levels from source to destination via every intervening stage.
I agree but what about the distortion that might have been introduced by using preamps with inferior components that introduce distortion at 6 or 9dB below clip point? That distortion will still be there no matter how much you turn it down and at 20 or 30 tracks, the cumulative effect will kill your recording. You'll no doubt get harmonic buildup somewhere in the spectrum no matter what your digital levels are.
At the risk of sounding cavalier, if somebody has a pre amp so poorly made that this is an issue then they should save up for better. Reducing your digital levels to compensate for deficiencies in your analogue gear should be a stop gap only--it doesn't mean it's a good way of working in the long term. It's a bit like driving your car at 30 mph on the freeway because you know your brakes aren't good enough to do the speed limit. In any case, I was talking more about general principles than trying to compensate for every bit of defective or badly built gear.
Exactly, so, if you don't have the time to null test every plugin (like most working engineers) the best thing to do is just keep your levels conservative and get on with the job which, of course, mostly entails LISTENING. It's amazing how the more listening experience you have, the less you need to know the results of a null test. Yes, they're interesting but for fuck sakes, who's got the time and the energy to test ALL of their plugins this way? How exactly will knowing the results make my mix any better? Did engineers of old refuse to work with an 1176 until they knew the results of a null test? Fuck that.
I basically agree--but my conservatism starts earlier. I work with a relatively small range of hardware (and tend to have investigated the quality before I put it to actual production use) and an even smaller selection of plug ins--but that's just me. However, if I came across a plug in that started to sound bad with levels that should be "legal" I'd probably just stop using it rather than compensate for its shortcomings by adjusting all my levels downwards.
In any case, we're getting to the stage of debating how many angels can fit on the head of a pin--or how many dB can fit in an audio file. In the "real world" I think most of us actually work in pretty similar manners and set levels in similar ways. The only thing I take real exception to is the assertion that ALL ELSE BEING EQUAL -20dBFS sounds better than -1dBFS. Yes, there are factors that can make this so and mitigate towards more conservative levels, especially when considering everything else in the chain. However, purely in the digital domain there should be no difference at all--and my experience of listening confirms this.