Questions about bit depth and dither

Hi All, I'm a little confused about the functional difference between the bit depth of an actual audio file versus how it is processed in Sonar 8.

I'm using an FA-101 with ASIO drivers.  The FA-101 is designed in such a way that the sample rate is specified via a knob on the front of the interface.  I have it set to 44.1.

In Sonar, in 'Global' Options, the 'Record Bit Depth' is set to 16 and the 'Render Bit Depth' is 32.  (I presume these are defaults because I don't recall changing them, though I may have...Perhaps the FA-101 audio drivers determine these setting....?...but I'm getting a bit off track.)  In 'Audio' Options, the 'Audio Driver Bit Depth' value is displayed as 24, but the value is greyed out and cannot be changed.  So, here are my first two point of confusion:

1.  What is the difference between Render and Record bit depth, particularly as it concerns my quasi-mastering process, ie: bouncing audio out of or loading a wav back into Sonar?(more on that in a minute...)

2.  If Global Options is set to 16 for the Record bit depth, but Audio Options shows the 'Render Bit Depth' is 32, what is the actual bit depth of my project?  Perhaps this is a repeat of my first question.  I suspect that the actual bit depth is 16 simply because when I right click any of the wav files in my project and go to Clip Properties the bit depth is shown as 16.  Or maybe the bit depth of the project is 24(the 'Audio Driver Bit Depth' in Audio Options)?

As an example of what I'm trying to figure out, I've got a project with four instrument tracks(audio with VST's-eq, Waves S1, a little compression on bass) and an Addictive Drums VST-instrument/midi track.  The process I'm currently using for my "quasi(amateur)-Mastering" involves exporting the project as a wav, loading that wav into HarBal, then loading the HarBalized wav back into Sonar to apply limiting(1.5 db in this case with L3).  Now perhaps just looking at that basic overview, you might see an issue with the process I'm not aware of.  There's probably a "better" way to do this, this is just how I opted to go about working the HarBal step into the process, because I wanted to do the limiting last.

In the above process, I first exported the project out of Sonar at 16 bit depth, with no dither because I assumed since the wav's were 16 bit, no dithering would be necessary.  But after thinking about all this, I suspect that Sonar is doing some things at 24 bit or some other depth higher than 16, like eq and reverb busses.  Should I have exported as a 24 bit wav, without dither, done the HarBal step, then after I load it back into Sonar to apply the limiter, Export it again at 16 bit with POWr or Triangular dithering?(From what I've read, triangular and shaped dithering seem to be more or less functionally the same...).  Or is there some reason dithering needs to be applied if I initially export the project as a 24-bit wav? 

Perhaps I should also change 'Record Bit Depth' in Global Options to 24 so that it matches Sonar's 'Audio Driver Bit Depth' in Audio options and to benefit from the increased headroom?  And maybe I should also start working at a higher sample rate?  I've been aiming for a CD quality finished product(16bit, 44.1) for simplicity's sake, but maybe I should bump it up since I usually end up posting material in digital format online anyway, versus burning it to media.  I hope this is clear enough, I know dither is an advanced subject and I'm a neophyte but I just want to make sure I'm not making a mistake in my final product.  And typing this up has also got me thinking about moving away from a CD quality end result, since that's an outdated standard.    Any advice or thoughts are appreciated!

Ideally, you should have your soundcard / interface and sonar's record bit depth to 24 (this is the deepest your converter can do and is 'industry standard').

If your computer is powerful enough and your hard drive fast enough you might consider upping your sample rate but for now it is probably good to keep it at 44.1khz.

The render bit depth option defines the floating point space that Sonar uses to process your files, it's fine (and least problematic) at 32. The higher the numbers the more decimal places Sonar has to play with and therefore the fewer 'rounding errors' are likely to creep in (at some point it needs to decide if a number is one or t'other and will round up or down accordingly, the further a long a list of decimal places this is the lesser the impact).

Generally you want to keep your project at the highest possible rates until the final moment, when you want to prep the file to be burnt to CD or you're exporting it out to another format (mp3 etc), after you've done all your mastering treatments.

For CD you need to convert them to 16bit 44.1 (and apply dither if you're coming down from 24bit)

Child

But here's the question. If Sonar is recording at 32 bit, why does the project show 16 ( or 24) at the bottom and if you check the Wave files stored in the audio folder they are 16  (or 24) ? not 32.
I always figured the 32 bit was something to do with rendering the audio as you work. But once you save it is printed to the audio folder as 16 or 24.

Mistakewalk

johnnyV


But here's the question. If Sonar is recording at 32 bit, why does the project show 16 ( or 24) at the bottom and if you check the Wave files stored in the audio folder they are 16  (or 24) ? not 32.
I always figured the 32 bit was something to do with rendering the audio as you work. But once you save it is printed to the audio folder as 16 or 24.

Yeah exactly.  And if Sonar shows the project as 44.1, 16bit(at the bottom of the Sonar Window) then does that mean I don't need to dither if I initially bounce the project out as a 44.1, 16-bit wave?  Or do I need to apply dithering since Sonar is doing stuff at 32/24/20/18bit(lol) bit depth as bitflipper explained? 

No, you only need to dither if you're changing the record bit depth. Sonar handles the render bit depth changes itself, so if you're recording at 16bit and staying there you don't need to dither.

Child

Sonar works internally at either 32bit floating point or 64bit floating point.

Thus, theoretically at least, you would dither when outputting to anything other than a 32bit (or 64bit) float file.

See!  no one  knows the real answer! That's two completely different answers now!
Mistakewalks question ( and mine now too) is  Sonar recording at 32 bit , ( 64) which would need dither, or what it say's at the bottom of the screen and what you find when you look at the files stored in the audio folder.

I think that the 32 bit (64 bit) is nothing to do with the track wave files. But then why is the 24 bit greyed out on both my systems and his too. 

So when should one dither then?
I track at 44.1.  If I make an MP3 I always leave the dither box ticked when I Export/Audio because you must dither when converting from 24 bits to 16, and adopting what I thought was the credo 'dither only at the very last audio conversion stage'.  Am I correct in all this? 

OK so I just did a little snooping using 2 computers and 2 Audio interfaces.
AMD 4800 duo core 2.4 - M Audio Fast Track pro
Acer laptop 1.7 - Tascam us1641
Sonar Pro 7  Windows XP SP 3
For me this was an eye opener- I have been recording at inferior rates because of my own ignorance and and like all things Sonar, there are a LOT OF SETTINGS!

  So here was my lesson for the day-- I could change both my interfaces to 24 Bit but this is what shows at the bottom of Sonar on both Computers??  16 bit?? Ahh now I get it  FIXED! 
  I was missing a second step to the land of 24 bit-- It is under Global options/ Audio Data. I've messed with it before and it did not change , you have to set your interface FIRST.

So the other mystery is the grayed out 24 bit in ever situation including my office computer which has a on board card.

Looking at the FastTrack Pro manual on M-Audio's site, it states:

At 24-bit operation, the Fast Track Pro operates as a two-in, four-out device or a four-in, two-out device at a maximum sample rate of 48kHz, with analog OR digital inputs available:

< Analog In 1 and 2 or S/PDIF

< Analog Out 1 & 2 (available on both TRS and unbalanced Outputs 1 and 2)

< S/PDIF Out (mirrored at unbalanced Outputs 3 and 4)

Or

< Analog in 1 and 2

< S/PDIF In 1and 2

< Outputs 1 and 2 or 3 and 4

At 24-bit operation, the Fast Track Pro can also operate as a two-in or two-out device at sample rates of 88.2 and 96kHz. Analog or S/PDIF can be selected for input or output.

So this is probably a throughput issue and you should be able to get 24bit if you select only the 2 digital inputs.


And bit rate = bit depth * sample frequency

  Edit this post and move info.

Have you noticed any oddities since changing to 24 bit (especialy interested in the Acer laptop, as that is what I use)?
I upgraded to a Tascam 1800 recently, but at the same time I changed to 24-bit. When recording I get odd stutters, and a general lack of response/more audio drop-outs etc.

johnnyV


Yep I figured that out today about the no digital outs at 24 bit. It's even more complicated because it is always been slaved to my Yamaha 01v which is 44.1 / 24 bit. so I don't mind not having 48 Hz but the 16 bit part sucks. I hate to have to use the Tascam for both computers but looks like that's whats going to happen.  The only reason I kept the M-Audio was it has SPDIF. Nothing else about appeals to me.

Actually if you have the original 01V like me, then the I/O is all either 18 or 20 bit converters but you can set the S/PDIF output to anything from 16 to 24 bits as you wish.  It doesn't give you any more resolution though!  Also, if you slave the 01V to an external clock it can also operate at 48kHz.  It's only fixed at 44.1kHz when using the internal clock.  Check out the "Set Up" pages 1 and 5.

Mr Ease, yes it's the original 1993! I tried clocking Via the external and it's keeps defaulting back to internal every time you power down. So it's a pain to have to re set each session. I don't think the difference between 44/.1 and 48 is a biggy. 
The pour old girl is starting to have a few issues. I can't afford a new digital board and I could never go back to analog for live sound.

Thanks for the input everybody, good stuff!

Yeah, the tricky thing about the Edirol FA-101 interface is that while the sample rate is set via the knob on the front, there is no knob or anything to specify the bit depth, and the software panel for the interface only has buffer setting options, nothing alse, very minimal.  So I don't know how I'd go about changing the 'Audio Driver Bit Depth' setting in Audio options(currently greyed out).

Drewfx said "Sonar coverts everything to 32bit floating point (or 64bit) and always uses that internally. Whenever it leaves Sonar (even temporarily), it is converted to either whatever your audio outputs' format is, or whatever file bit depth you set it to." 
So does that mean that since my 'Record Bit Depth'("audio output format"?) in Global options is set to 16 that all the other project audio will convert to 16 bit automatically(applying dithering as needed) if I export the project as a 16 bit wav? 
Or would it defer to the greyed out 'Audio Driver Bit Depth' setting of 24 bit in Audio options, which I'm not sure how to change(and don't think I want to).  If it's the former, it looks like I can skip dithering.  If it's the latter, it seems I should apply dithering if I export the project at anything lower than 24 bit, but I'm still fuzzy on that.

Also, when I bounced the Addictive Drums track to a wav, it created a 32-bit wav, so it looks like it uses the Render Bit depth not the Record Bit Depth to bounce the AD to audio.  Of course that means now I will be reducing the wordlength of that track if I Export the project at any bit rate lower than 32, so I guess I'll dither when I export.  Or I could leave the Addictive Drums track as a live VST when I export(instead of bouncing it to a track first) in which case it seems like dithering upon export would also be in order.

So in summary, even though my recorded audio is at 16 bit, and the bottom of my Sonar window says 16-bit, I will still apply dithering(triangular or shaped; probably Pow-r1) when I export my project at 16(or even 24)bit since Sonar is still doing things at a bit depth higher than 16/24/#.  Good choice, yes?


On somewhat of a side note, but still related to my final process:
As far as my use of HarBal goes, I agree with bitflipper that it's best used just as a reference.  But in this case I like what 12% of 'Air'(HarBal's stereo widener) did to the mix and it did let me make a tasteful correction to a minor dip in the upper range and I'm not really sure what lives there in the mix.  Maybe you know of an alternative to HarBal 'Air' that could be used as a plugin on my Master bus, then I could just skip the HarBal. ( I could sacrifice the HarBal step to avoid another run through Sonar after the HarBal process to apply limiting- I'm still not sure doing a second run through Sonar is problematic but I suspect it isn't).  Waves S1 is the closest thing I can think of.  I actually used an instance of S1 on each of the tracks in this mix, somewhat experimentally(what isn't?), to play with the imaging and liked the result.  Maybe I could slap another S1 or something on the Master instead of using HarBal's 'Air' but Air seems to get a good transparent spread that I don't think I could achieve with S1.