Here is what Paul Frindle says about my theory. I am wrong but I am right and even he is confused about the whole concept.
You see in the top of this post, frindle says, as Jeff has pointed out that it was him who propigated the myth of recoding hot but he also says that he also sugested what I do which once it's in the box use the trim to mix:
Ok - this has become controversial and discussions about this have raged for years - and I feel that it's largely my fault :-(.
I originally presented the idea that operating at absolute max levels was one of the major reasons people obtained worse results ITB in comparison to analogue systems.
The primary reason for this IMVHO was a workflow issue, in that almost anything you did (apart from moving a fader down) resulted in higher peak levels - which would put on red lights and involve you in messing up your balance permanently to put them back out again.
Think EQing a track in the mix - the red light comes on (even if you've cut something) - so you have to reduce it - so it's now quieter in the mix and has messed up your balance and drastically interferes with your artistic instincts etc.. Are you really going to reduce the levels of every other track in your mix by the exact same amount to get back your original balance? Not likely :-(
All of this could be avoided by reducing your levels (with a trim plug at the head of your mix channel if required) and carrying on with the mixing without the distraction - much as analogue systems always did.
Analogue systems had headroom above max operating levels for a good reason, which the digital community growing from HD samplers and editors had entirely missed :-)
----------------------------------------------------------------------------
As simple as this might seem, the discussion has ballooned and been made confusing again. I could never have predicted the bombshell response this simple idea provoked :-(
For starters people would not accept that simple things like cutting some range of freqs or rolling off the extreme LF could increase level. So I posted some examples this and other things such as intersample peaking in some plug-ins even if the response does not change. Showing that if you are aiming for max modulation through your mix and trying to get the loudest possible result - all this stuff got in the way.
Then people objected to reducing level as a concept - because they wrongly assumed (and had even been taught) that doing so would 'reduce their bits' and produce distortion!! This was the fall out from the widespread marketing mis-use ot the term 'resolution'. So this resulted in me doing scores of posts basically trying to get people to actually gain an understanding about sampling theory in layman's terms, i.e. trying to undo one of the biggest myths in the industry :-(
Other discussions widened the issue further and got completely off track - for instance the floating point 'it can't clip' guys missed the point entirely.
Also people artificially widened it to analogue operating levels (which was not what I was saying in the first place), ADCs and DACs as well - rightly arguing that running analogue outboard stuff at lower levels could result in not being able to get the effects (analogue distortion) they were after. So this resulted in scores of posts about converters, their input/output levels, their max output levels (sometimes insufficient) and so on and so forth - endlessly - not least of all since many people were trying to use consumer quality converters to drive pro analogue gear - and of course even pro converters are starting to cut corners these days.. Nightmare discussions that are still raging 3 years after, without even a hint of general widespread understanding still.
Then it transpired that engineers mixing on small OTB 'summing products' were actually clipping their DACs (sometime without even knowing due to inter-sample peaking) - mixing it OTB and flying it back into the digital domain again. A process that produced distortion of all kinds which was more important than the OTB mixer itself!! The cause of much of what the proponents of the summing boxes actually heard as an improvement. We even had amazing suggestions from people that computers may not be able to add up - a subject not even worth discussing, to anyone who just thinks about it for a single moment without the spin!! :-(
Then it transpired that some mastering engineers were actually doing this so they could use the ADCs as signal clippers - and increase the modulation levels (a secret weapon in the loudness wars)!
So what followed was scores of posts about the ways in which converters may (or may not) distort under these conditions - and then we find out that converter products are actually being selected depending on how they sound when deliberately over driven. At which point all bets are off - and any technical discussion becomes pointless :-(
----------------------------------------------------------------------
Ok to cut short the history of confusion opened up by a simple suggestion and get back to your question:
Because people could not bring themselves to reduce levels (the effing 'resolution' concept could not be cleansed out of their brains) - they started looking at plug-ins to ask the question as to whether or not they could be legally overdriven anyway, continuing to work 'correctly' even if the red lights were on - thereby conserving their precious notion (religion) of 'resolution'!!!!
Sorry if my frustration is showing through here - I am only human :-(
Now as completely irrlevant as this may be to the original argument - it is none the less complex, if you are hell bent on driving stuff over the limit and cannot be persuaded it's a bad idea and gets you absolutely nothing...
The problem is that whether a plug-in can or cannot stand this kind of thing depends on many factors including the process itself, the platform it's running on, what the platform may be externally connected to and so on. For anyone not understanding overload (let alone basic sampling theory) this is a nightmare of confusion.
I feel for the users, who after all should be able to work freely and artistically without messing with all this crap - i.e. my very original sentiment for reducing levels in the first place!! :-(. It makes me want to weep TBH.
In short my original intention in suggesting reduced levels and headroom had backfired completely. The marketing dis-information was so deeply ingrained and relentless that people were now required to be technically proficient (in ways no forum post could ever achieve), just to understand how to use things - that already had red lights and warnings aplenty. It was a totally impossible job for one person. This is an example of where knowledge gained by a whole industry from decades of design will follow us to the grave - because it has no place in the current marketing environment... It's a lost cause :-(
Anyway what is the bottom line (finally)?
Generally plug-ins running on a floating point environment can stand being overdriven (with some minimal loss of quality), providing the signals are not sent out to fixed point systems, like expansion processes (PT TDM, PowerCore) and interfaces DACs, CDs, SPDIFF, AES etc..
Plug-ins running in fixed point may stand being overdriven if the inputs are not themselves overdriven, they have internal headroom and they have output level controls to reduce the level before it's sent to the next process. This also applies to the RTAS plugs in a TDM system.
Plug-ins that have to have internal references to real world output levels (Dynamics, limiters, character plugs with distortion) may produce different results depending on absolute level whether in fixed point or floating systems (i.e the float does not take away the need for real level references). So it's very complex and application dependent in a great number of ways I could go into, but this post is long enough! :-) But it's a bad idea to trust overloading them...
The only horrible fly in the ointment is that since all of the plugs that need internal reference levels are designed to match full sample value level (the original default broken digital 'operating level' which has caused the whole problem), they may need to be modulated to full level internally to get the intended results..
For instance; our DSM has an output limiter set to catch sample value overs above max. If you like the sound of the limiter working hard you have to make sure that the rest of the process internally produces enough modulation to do it. It will then output full level automatically - which you may want to reduce to send on to the next process, if the DSM is not last in your signal chain..
I am so sorry for this long post - but I felt that I couldn't make sense of any reply to the question without refering to the original source of the whole debate.
So this is what Paul Frindle has to say on the topic, if someone can decipher some of it for me I'd be happy.
Here is the link to the whole thread:
http://www.gearslutz.com/board/music-computers/542885-paul-frindle-truth-myth.html I haven't gone through the whole thing but I did find some interesting stuff about truncating harmonic distortion, which is what I was talking about and believed that it could be combated through recording hot, what is not clear from this thread is how to combat losing that harmonic content.
What is also clear digital converters can be be driven and can be pleasing so recording hot could possibly add some form of digital harmonic distortion, or that is how I read it.
Man, this thread show's how little we know about the digital recording process and how importent these discusions are!!
I'm sticking with the hot theory because I tend to mix in Sonar as if it was an analog mixer, eg I use the trim pot and Esp if there is even half a chance of adding some kind of digital harmonic distortion that is pleasing to the ears, into my mix.
Well enjoy, I will continue to analyse the thread over the coming days, they're are just too many pages!!
Neb