If my audio is 16 bit, project set to 24 bit, should I MIX at 16 or 24 bit?

This older song was recorded in 24 bit mostly. Currently I have Sonar set to 24 bit for all projects but when I am mixing Multitrack down to stereo track (for mastering), do I Export audio at 24 bit or 16 bit?
 
For my newer projects that recorded at 24 bit I was recommended to Export at 24 bit, Master it with Izotope Ozone in new project, and then Export it from that project at 16 bit and MP3. Is it how you would do it?
 
In case of 16 bit audio- I just exported it in 24 bit stereo track for Mastering and now I am thinking- if it was recorded at 16, shouldn't I export it at 16 so it would not need to truncate it again from 24 to 16 after mastering? Exporting 16 bit in 24 bit would not improve it but that 24 bit new clip will be compressed to 16 bit after mastering with new loss. So should I initially (before mastering) export at 16 or 24? Thank you.

I used to record in 16 bit because my audio interface only offered 16 bit. I then switched to recording in 24 bit but was mixing for Mastering in 16 bit. But was told that it is better to export first in 24 bit, master and then render to 16 bit.
 
But this particular song is recorded in 16. I just listened what it exported at 24 bit- I can not tell if it would be any different if I set it to 16. But who knows, I am just thinking- if I export in 24 what recorded in 16, and then truncate 24 to 16 again, wouldn't I have losses as it has to take something out to make it 16 out of 24?

vladasyn
I just listened what it exported at 24 bit- I can not tell if it would be any different if I set it to 16. 

And you won't. When you export to a longer word length, all SONAR's doing is tacking on some zeroes to pad it out. It will sound exactly the same.
 
The reason for exporting 24- or 32-bit audio is so that the next software that gets it (e.g. a mastering program) will receive the maximum resolution. But if you're doing the mastering within the same project, there is no benefit to exporting 24- or 32-bit files, and no penalty for leaving it at 16 bits.

Yes of course. But make sure you select a suitable dithering algorithm.
 
The only time I render from 24 bit to 16 bit is if I want to burn a CD

I nominate this thread for the SurfBoard section of the next eZine.

This is important: http://www.dailyrindblog.com/technical-supporting-high-quality-audio/?utm_source=Sailthru&utm_medium=email&utm_campaign=Client%20Newsletter:%20week%20of%20May%2016%2C%202016&utm_term=All%20Clients%20%2B%20All%20Orchardites
 
The Orchard’s brand new standard audio specs are: WAV (.wav) / 16- or 24-bit / 44.1 kHz / Stereo.

• Be uniform when submitting your releases and avoid uploading a mix of 16-bit and 24-bit tracks on the same album.

I guess- this answers my question. So based on this article- can I say that I dont even need to turn it to 16 bit unless I make CD?
 
Then what would be the best way to make bith 24 and 16 bit files if I have Ozone? Thank you. 

Sonar and all wave editors allow you to save in just about any format. 
But the working environment these days is normally 24 bit. 
Sonar is working at 24 bit for me now since I bought the Tascam and then later the Scarlett. I think my M Audio interface ,10 years ago, was the only one that forced 16 bit. I think even my Sound friggin Blaster was 24 bit. 
 
Studios should be in the habit of keeping the highest quality master on hand. 
So there's no reason not to always have a 24 bit copy of each song on your album on hand. 
From there you can down grade them to Cd or MP3 as needed. 
I keep a main folder with sub folders inside for each mastered album. The sub folders contain the 16 bit or the 24 bit Wave and the MP3 version, I might have 2 Wave folders and 2 MP3 folders, one at 160 and one at 360.  
My old masters were all 24 bit but these day's I'm not recording anything important so all is just 16 bit from the minute I export from Sonar. 

vladasyn
 How do you use 24 bit files? it would not even play on anything (or may be it only applies to 32 bit?.

Your audio interface is almost certainly converting 16bit to 24bit when you play a 16bit file through it.

It is almost certainly doing the analogue-digital conversion at 24bit when you record. It is Sonar that will then be following your instructions and reducing the 24 bit data it receives from the interface to 16 bit.

I say "almost certainly" because there may be an interface or sound-card somewhere that still uses 16 bit convertor chips, like CD players use. I'm just not aware of one that does, even the modern cheap on-board sound cards use 24 bit convertors.

As for 32 bit, no, many media players can't handle it. And there's no point in them being able to handle it. 32 bit audio can handle a dynamic range of 193dB, which from an audio point of view is pointless because there are very few things that are that loud - nuclear weapon tests, Krakatoa-scale volcanic eruptions, that kind of thing. 24 bit handles 145dB, which is fine for all real-world audio recording and playback.

32 bit data processing, which is what many DAWs do, or 64 bit, is useful because the DAW pads out the 16 or 24 bit audio with lots of zeros to get the higher bit depth, which means processing it can be done more accurately with fewer unwanted side-effects. But your audio hardware will almost certainly be playing the result to you at 24 bits because that's what the hardware handles.

Saving projects at 32 bit means that Sonar or whatever processes the audio next doesn't have to do a conversion from 24/16 bits up to whatever it handles then back down again so avoids some processing and some potential, but very minor, side-effects. Though those side-effects are generally inaudible to humans. 32 bit files do take up a great deal of disk space though.

As for dithering, I forget the details, but someone (Ethan Winer?) did a test involving a lot of experienced engineers to see if they could tell the difference between audio dithered down using different algorithms and no dithering at all. Generally speaking they couldn't, especially between the different algorithms.

Record at 24 bit and leave the final mixdown at 24 bit unless you're making an audio CD or you need to distribute it as a wave file that's as small as possible while retaining CD quality. For all other purposes there's little point in dithering a 24bit file to 16 bit. There's an argument for mastering for mp3 separately than for CD, FLAC and other formats because mp3 benefits if the master has two or three dB headroom rather than the common 0.3dB used for CD. Mp3 is also likely to change the sound of the audio, especially at lower conversion qualities, which may also make a dedicated mastering job advisable,

Apple's stuff about mastering for itunes has some useful things to say about mastering for compressed formats, including Apple's own lossless compression. The big emphasis is on avoiding digital clipping.