A One Hour Wave file

Well in fact the full story is this (if it helps) .. I imported/opened 4 MP3s, mixed them down to one Wave, and then tried to export/save that. So recording in 32 bit is not in the equation, but the 'session' might be in 32 bit, and if this is the case, why does it say '16 bit mixing' in the bottom-right corner in Multitrack View?   The only thing I do in Audition in 32 bit is original voice recordings, could this setting be affecting everything I do in Multitrack?

I imported/opened 4 MP3s, mixed them down to one Wave, and then tried to export/save that. So recording in 32 bit is not in the equation, but the 'session' might be in 32 bit, and if this is the case, why does it say '16 bit mixing' in the bottom-right corner in Multitrack View?   The only thing I do in Audition in 32 bit is original voice recordings, could this setting be affecting everything I do in Multitrack?

Yes it is - and now you've explained it as well. Your voice recording is 32-bit, and if you have the default settings in place, the MP3s will be opened as 32-bit as well - and if all of these are in one session, then it has to be a 32-bit one, because you can't mix the bit depths in the same session.

Anyway, the solution was to save Waves from the Edit View rather than from the Multitrack View and then the Waves are the size they should be.

So, we've established what was happening quite clearly - you simply haven't done the 16-bit mix on your 32-bit files. But I would suggest that if you do this, you aren't doing the optimum job here anyway. What you should do is to set the mix so that it happens most accurately - in other words do a 32-bit mix. You then end up with a 32-bit mixdown in EV, and then you convert this to 16-bit. If you do a 16-bit mix, you are effectively throwing away any benefit at all from recording your voice in 32-bit mode. What you should do is to keep everything 32-bit whilst it's being processed - so you do a 32-bit mixdown, and then normalise it in 32-bit in EV, and absolutely last of all, convert it to 16-bit. It's important that any amplitude changes at all, or mixes, are done this way - this is the one area in which you lose effective resolution irretrievably. A 32-bit mix is a Floating Point mix though, and effectively lossless from this POV.

Some people might argue that with a correctly recorded voice in the first place, and MP3 files in the mix, that you won't actually notice the difference - and hey, they may be correct most of the time. But I'd say that as a matter of principle, you should do it correctly anyway - because at some stage, if you don't, it's going to bite you... I can't say how or when, but Sod's Law does seem to be remarkably consistent from this POV!

Quite frankly though, I can't see any reason for the 16-bit mixdown option even to be there in Audition when the session is 32-bit - it's completely pointless. FWIW, if you have a 16-bit session, then doing a 32-bit mixdown is even more important - this is the time when the differences can become quite clearly audible.

Yes, good, explained, sorted....now I'd like to know why Audition has the audacity to continue running and continue drawing on CPU recources after I've shut it down .. I wanted to rename the session folder and had to do a CTR + ALT + DEL to discover audition.exe was still running after shutdown! Rant over, off-topic  

I have that problem occasionally...takes me a moment to remember why everything is running so slowly...then I smack myself upside the head (figuratively--I'm delicate) and end the process.  I'm running Win2000, so the procedure is slightly different.
[/u]

Quite frankly though, I can't see any reason for the 16-bit mixdown option even to be there in Audition when the session is 32-bit - it's completely pointless.

I couldn't diasgree more Steve!

If you're doing final CD mastering from the multitrack and the last thing in the fx chain is the Ozone 3 dither down to 16-bit surely there's absolutely no reason not to save time and hassle and mix down to 16-bits directly, rather than go to 32-bit and then 16-bit in the EV without further dither?

Likewise using AA's own dither routine - if the final mix is not going to undergo any further alteration (which is normally the case with my work) why not have it mixed and dithered down to 16-bits with a single command?

Surely it's a matter of flexibility that this is an option - someone somewhere will always find a valid use for it!

[Cool Edit Pro (Adobe Audition)]

If you're doing final CD mastering from the multitrack and the last thing in the fx chain is the Ozone 3 dither down to 16-bit surely there's absolutely no reason not to save time and hassle and mix down to 16-bits directly, rather than go to 32-bit and then 16-bit in the EV without further dither?

Likewise using AA's own dither routine - if the final mix is not going to undergo any further alteration (which is normally the case with my work) why not have it mixed and dithered down to 16-bits with a single command?

Well I'm glad that you can guarantee that your mixdowns are going to be exactly at the final level you want them, because I certainly can't with anything involving more than a single pair of tracks! That's basically the first reason that it's pointless - the times that you could do this would be the exception rather than the rule. Secondly, a 16-bit integer mixdown is not exactly a high quality affair - I really don't think that I'd want to do what you are describing anyway, I'm afraid. I can't conceive of any situation where 16-bit mixing brings any sort of advantage at all - only disadvantages. Even a 16-bit session should really be mixed in 32-bit - the chances are that the sum and difference products, and very likely the IM products will be somewhat worse than they would be if this was mixed in 16-bit.

It's no accident that Ozone is based on 64-bit processing - what's the point of feeding it with a degraded 16-bit int. signal when you could just as easily send it a 32-bit FP one which is already in the format that it wants to deal with? This certainly isn't what iZotope recommend at all - they will strongly advise you against applying anything other than a 24-bit signal (which is basically what a 32-bit FP signal is, of course) to Ozone for dithering.

Here's the critical bit from their dithering guide about how to apply it in Audition:

Cool Edit Pro (Adobe Audition)
Same process as Sound Forge to start - open your 24-bit wave file, apply Ozone. To actually perform the conversion from the 24-bit wave to the 16-bit wave, go to Edit-Convert Sample Type.

{picture of sample convert page here}

The key checkboxes here are circled. Make sure that Enable Dithering is not checked, and select 16 bits as the output format. Click OK, then save the mix.[/list:u]
No, the whole idea of 16-bit mixdowns for any reason at all simply sucks, I'm afraid. It's only a throwback to days when this took simply ages to achieve on a slow machine - you could get an idea of what your mix sounded like rather faster like this. But you should never do it with a 'real' one.

It had never occurred to me that the maths of the mixdown might be 16-bit.  As a programmer I would naturally write only one bit of code (the best possible), and convert to the required format at the output stage.  As the Intel processors since the 386 are fundamentally 32-bit, separate 16-bit code would be unlikely to be significantly faster, if at all.

Are you saying that you know that AA does fully 16-bit arithmetic to generate a a 16-bit mixdown?  After all, they wouldn't be able to add dither if that was the case.

Paul

[When performing a mixdown, this is the bit resolution that will be used.]

It had never occurred to me that the maths of the mixdown might be 16-bit.  As a programmer I would naturally write only one bit of code (the best possible), and convert to the required format at the output stage.  As the Intel processors since the 386 are fundamentally 32-bit, separate 16-bit code would be unlikely to be significantly faster, if at all.

Are you saying that you know that AA does fully 16-bit arithmetic to generate a a 16-bit mixdown?  After all, they wouldn't be able to add dither if that was the case.

I ended up thinking about this instead of going to sleep, which I should have done... Of course, it's Audition, and there are always more options, which makes it more complicated...

What it says about the 16 and 32-bit Mixdown option (my bold):

"

When performing a mixdown, this is the bit resolution that will be used. Regardless of the session format (16 or 32 bit) mixdowns can be generated at either 16-bit or 32-bit quality with this option. The default is 16-bit." (ouch)[/list:u]
So that's quite clear - except that it's not quite clear, because underneath is a pre-mix quality setting:

"Determines the bit size used for the background mixing process. Best quality is achieved by leaving this at the default 32-bit setting. However, if you're using  multiple soundcards, it may be advantageous and faster to choose 16-bit for pre-mixing, as less data will be transferred across the drives. For single device situations, or faster hard drives, 32-bit is better because it provides optimisation at mixdown."[/list:u]
So, it quite clearly states that a mixdown can be performed as a 16-bit operation - the pre-mixing situation makes a significant difference to the background mix, but this is a different issue to the

actual mix - which is at the selected bitrate. My problem with what Andrew is doing though, is more concerned with the first issue I mentioned - one of the things that you cannot include in the MV process chain is normalisation - this is an EV-only option. My personal view about this is that the wider the dynamic range of the signal you are processing, the more important it is to determine that the peaks of it are using all of the available dynamic range when it's 16-bit. If you lose 6dB at the top of the range, and your dithering  is taking up 1 bit at the bottom, you've got yourself a 14-bit dynamic range recording at best, and quite a lot of what you've recorded ends up at a rather lower resolution still - and in a critical listening situation these differences become quite audible.

I just spent a day doing precisely this in very* quiet environment with a decent pair of monitors, and it's amazing just how much difference there is between optimised and non-optimised recordings at 16-bit resolution. If I am processing anything that critical, then the whole thing stays at 32-bit until the Ozone dithering process. The steps at this stage are to normalise the 32-bit signal, then add the Ozone dither as a discrete step, and then finally, do the 32 to 16-bit conversion - as recommended in the freely available Ozone Dithering Guide.

I still can't see any justification for retaining 16-bit mixing - there's absolutely no need for it at all, and having it set as a default is even worse!

* I measured the background noise level - only I didn't, because it was well below what my level meter will accurately measure, and at that level, we are into its own system noise. All I can tell you is that the Leq(15) was significantly lower than 20dB (around or below the threshold of hearing). This is eerily quiet, and only gets that way late at night. It's actually too quiet if nothing is going on - you can hear all the weird noises your body makes - including the blood circulating around your head, which is why as humans, we can't hear anything significantly quieter than this.

MV .. integer .. FP .. dither .. it's occurred to me that you really do lose a lot of people when you use very technical jargon, when this forum is supposed to have a wide Audition user purpose, and only a minority of users, including producers and creatives are that technical .. kind of defeats the object of this forum which is, I imagine, aimed at the majority of Audition users. So try, if possible, to remember that things are best explained in a way that everybody can understand it, and remembering that will ensure most of us will get a lot more out of the discussions and explanations, thanks techies

MV .. integer .. FP .. dither .. it's occurred to me that you really do lose a lot of people when you use very technical jargon, when this forum is supposed to have a wide Audition user purpose, and only a small number of users, including producers and creatives are that technical .. kind of defeats the object of this forum. So try, if possible, to remember that things are best explained in a way that everybody can understand it, thanks  

Oh, come off it! You're the one that was asking what amounts to a technical question in the first place. You are not obliged to read and understand every single thing written here, and neither does posting information in any way shape or form 'defeat the object of the forum'. In fact, it is an integral part of the forum, and always has been. If you don't like it, that's your perogative, but it's no reason for being snotty about it, is it? We don't 'lose' people this way - they either walk away, because it's not for them, or they are sufficiently interested and/or concerned to find out for themselves. We didn't choose the name 'AudioMasters' for nothing, despite what you might like to think.

So, if you want an explanation for any of the terms used, or any other terms, come to that, then ask, or look it up. You'll find that there's a more than adequate explanation for most, if not all, of them in the archive, or on the forum itself. You will find that there are plenty of examples where people have done just as I suggested - most people for instance, find out very rapidly that MV means Multitrack View. Yes, it's jargon. It's also, and more significantly, an abbreviation that most people using Audition use frequently and is, in this context, hardly technical jargon at all.

I shall effectively overlook the incorrect word in your last sentence ('it' should have been 'them', as it refers to the plural 'things'), but I will point out in this context that it is often just not possible to explain everything in the simplest of terms, and neither is it desirable. In fact, this is what causes most of the misunderstandings around. Different people learn best in different ways - there are no common denominators when it comes to how people best understand anything - it is entirely based on their own previous experience. It is not simple trying to explain even the simplest of things sometimes, and this I am very aware of, thank you. Of course it's not right for everybody - it never will be. But getting on your high horse and whining at me about something that's blatently obvious isn't exactly going to further your knowledge of anything, is it?

The danger of trying to address technical issues in simple terms is that one might end up reading something amounting to "take my word for it, it will sound better if you do this or that, it's too complicated to explain" - which many would find frustrating if not insulting.  I'd rather see the reasons and principles behind what I'm told set out; even if I don't grasp all the points, a little more rubs off each time I read.

But despite the huge amount of reading I've done here and on the related sites in the last couple of years, I've not taken on board enough of the detail of many of the technical answers to be able to respond to the technical questions at the same level as others here.  But the light that has been shed on these matters undoubtedly enables me to see further into the gloom, and I for one would be sorry to see any general 'dumbing down' of discussions.

Steve

You're not taking into account what I do. I am mixing from a single stereo source. There is no normalising to do as this is taken care of in the mix, usually by setting my maximum overall level in Ozone.

I'm using the multitrack to facilitate the chaining of perhaps 8 or 9 fx in the decrackling, denoising , EQing and further processing of a single track, taken from a vinyl or shellac source.

If I know that once this 'mix' is complete there will be no further work carried out - and that is often my intention as I want to be able to hear the 'final' sound prior to mixing of all my interrelating fx - then the final output of that fx send is already dithered down to 16-bit resolution by Ozone and is ready to be track-split and burned to CD, which is where it has to end up.

I have no need whatsoever at this point to be working in more than 16-bits. The job is finished with the final selection of mixdown. That is why it is useful to me. I don't always work this way - often there will be more to do in the EV. But there's enough jobs which do go this way for it to be a very useful function.

Like I said - it's a matter of flexibility for the job in mind - even if you can't imagine a scenario where it's of any use, doesn't mean to say others can't. For me it's something that's useful most working days...

My personal view about this is that the wider the dynamic range of the signal you are processing, the more important it is to determine that the peaks of it are using all of the available dynamic range when it's 16-bit. If you lose 6dB at the top of the range, and your dithering  is taking up 1 bit at the bottom, you've got yourself a 14-bit dynamic range recording at best, and quite a lot of what you've recorded ends up at a rather lower resolution still - and in a critical listening situation these differences become quite audible.

If I'm presenting a quiet piece of classical music, which should never get much above, say, mezzopiano, I'm most certainly not going to normalise it up to -.1dB. It would be ridiculous to have the effective volume level of this on someone's stereo at the same actual level as the CD of the Motorhead they'd just taken out of the player.

It may sound stupid, but in some ways that headroom has sometimes to sit there empty, taking into account the fortissimo that the composer never wrote. As I'm normally coming from vinyl or shellac and I'm dithering very effectively down from 64-bits (final part of the chain is always Ozone) I reckon this is a legitimate alternative to normalising everything to absolute peak level.

MV .. integer .. FP .. dither .. it's occurred to me that you really do lose a lot of people when you use very technical jargon, when this forum is supposed to have a wide Audition user purpose, and only a minority of users, including producers and creatives are that technical .. kind of defeats the object of this forum which is, I imagine, aimed at the majority of Audition users. So try, if possible, to remember that things are best explained in a way that everybody can understand it, and remembering that will ensure most of us will get a lot more out of the discussions and explanations, thanks techies

So, you're speaking for the majority ? Is that it ?

I think what you really meant was that you tend to get lost easily when the words get complex. Or when you don't know what they mean. And since you don't really have any interest in learning any more than you have to, you'd try to get everybody to "dumb-down" the discussions so that you can continue to leech bits of personally useful information - provided that no effort is required on your part.

"So try, if possible, to remember that things are best explained in a way that everybody can understand it, and remembering that will ensure most of us will get a lot more out of the discussions and explanations..."

That's the  - who's everybody ? How far would you like to see it go ? Should 7 year olds be able to participate ? Or the mentally handicapped ? Or did you mean, just, mostly you ? Anyway, your statement is completely backwards. To explain things in the simplest possible terms ensures that very little will be gotten out of it, if anything.

The whole thing reeks of pure laziness on your part, BFM. If you don't understand something, try FIGURING IT OUT. You haven't even bothered to become familiar with commonly used terms for the software, or computing in general. Like your insistence on using the word "exporting" to describe a save operation. Pick up a book !

If "Integer" is very technical "jargon" (which it isn't, at least not for a typical 12 year-old) maybe you should be using "GoldWave", or better yet, Windows comes with it's very own Audio Megaworkshop, it's called "Sound Recorder", and all you have to do is click stuff.

Anyway.....BOOOOOOOOOO !  

You should get a prize for the biggest turd left floating in the toilet bowl.

 

If I'm presenting a quiet piece of classical music, which should never get much above, say, mezzopiano, I'm most certainly not going to normalise it up to -.1dB. It would be ridiculous to have the effective volume level of this on someone's stereo at the same actual level as the CD of the Motorhead they'd just taken out of the player.

It may sound stupid, but in some ways that headroom has sometimes to sit there empty, taking into account the fortissimo that the composer never wrote. As I'm normally coming from vinyl or shellac and I'm dithering very effectively down from 64-bits (final part of the chain is always Ozone) I reckon this is a legitimate alternative to normalising everything to absolute peak level.

I wasn't suggesting that you would normalise a mezzopiano section - that would be a part of the wide dynamic range of the entire piece, after all. It has long been suspected though, and I think with some justification, that it's what 16 (or whatever it actually ends up being) bit reproduction does to quiet sections of music that really upset the majority of analog diehard enthusiasts. And that's why I'd rather consider the dynamics of the entire piece as a 32-bit file before committing myself to a 16-bit version, and optimise an entire CD accordingly. But I'm quite happy that other people's mileage may vary, and yes, if you are using Ozone to truncate your file at the 16-bit level, there isn't much point in saving it as a 32-bit file any more. I don't use Ozone in MV though - it's enough of a resource hog already without including it at the end of a processing chain, although I realise that this won't be anything like as bad if it's only a single stereo pair you are processing.

I'm quite intrigued about this person who switches between Motorhead and  a rather more subtle classical piece without thinking about checking the volume, though. Most people with catholic tastes in music are aware of the problems of doing this, and whilst I don't necessarily require to have wide dynamic range music normalised to -0.1dB (-1 or -2 dB is fine), I don't really want large sections of it at -35dB either, just because it's a quiet movement. With a lot of recordings, it's quite normal, and accepted, that the recorded level will be raised for quiet movements - organ music especially, because it's very easy in a large instrument to reproduce music with a very wide dynamic range - I recall that Ozpeter had a slight problem with this a while back...

Trouble is, the more choice you give people, the more they will disagree!