Vegas takes the internal floating point numbers and rounds to the output bit depth. They don't apply dither. If you want dither you should add a dither plug-in on the master buss as the last element of the plug-in chain. SF includes a dither plugin for that purpose.
cheers,
bgc

"Vegas takes the internal floating point numbers and rounds to the output bit depth. They don't apply dither."

Uh...Please elaborate. I don't understand how this works. Where can I find any documentation for using Vegas in this way? Is it a recommendation to use a dithering tool? My CD sounded okay.

Thnks again, Kurt

Use Sonic Foundry Dither as the LAST effect on your master bus when rendering a final mix. This will do an optimal bit-depth conversion from the 32 bit internal floats back out to the bitrate you’ve chosen for your rendered file, typically 16 or 24 bit. Doing this will get rid of the audible effects of quantization error that can occur when you drop depth (change the playback level of any track to something other than what it was originally recorded at). You can add the dither effect to the master bus at any time, but make sure it's the LAST effect in the chain when you render your final mix: you only want to do this once.

"Pre-rendering" a sub-mix would be a candidate for this process as well.

When Vegas reads in it's source audio (16-bit, 24-bit, etc.) it converts these integer numbers to floating point numbers (i.e. 2945 becomes 2945.0). Why? For a couple of reasons. First your computer has nice floating point math hardware in it that's fast. Second, Vegas is basically doing lots of math on your audio (fading, effects, etc.) and these are best performed in the floating point domain for more resolution (the numbers after the decimal place make a big difference in audio quality). Vegas doesn't care about the bit depth of your project UNTIL it hits the master bus and has to be fed to your sound card. At that point Vegas looks at every floating point sample sent to the card and checks it against your bit depth. If a sample is set to 16532.3234234 it figures out how much resolution it can keep at the bit depth of your project and rounds the number (I think it rounds, it may truncate) to the resolution your output bit depth can handle. Lots of times this is completely OK and you won't hear the artifacts. The problem is that as the audio goes low in volume you start to hear this quanitzation or rounding. Dither is intended to add 1 bit or 1/2 a bit of random noise so that the last bit flipping on and off doesn't sound nasty. Dither is art and science and there are different types of dither and many folks have different opinions about it. If you want to train yourself, fade out a track of audio from loud to silence in Vegas and render it at different bit depths (8, 16, 24 bits) with dither on and off and you can hear it if you've got a good ear and decent equipment.
That's it for today's lesson class!
Enjoy,
bgc

One technical correction. The original sample is actually converted to a normalized floating point value between -1.0 and 1.0. This is done based on the raw format of the wave file being converted. It is easier to work with normalized floating point as it strips out the specifics of the orginal source material so that the later conversion back to fixed point PCM data is generic a soley based on the hardware or rendering format desired.

This is by convention, and something that everybody does.

PEter

Great responses. But the master bus is not being fed to the sound card in my situation. I'm using the CD burning feature of Vegas. I have a mixture of 24 bit tracks and 16 bit tracks with track markers set to burn a red book compliant audio CD using DAO. Should I pre-render the 24 bit tracks (i.e. in Sound Forge and maybe use Ozone or Waves) to 16 bit? I have iZotope Ozone at my disposal, which I'm learning how to use.

Thnks, Kurt

That's a good point, if you burn directly to CD, it's got to be 16-bit. My guess is that it still goes through the master bus and that if you put a dither plug-in as the last plug-in on the master buss fx chain that it will dither the output as it's written to cd at 16 bits. One way to test this is to create a cd track with the Vegas burn tool without dither. Then put dither onto the master buss and use dither. Then load both tracks from the cd into Sound Forge or vegas and look at the difference between the two. If it's low level "noise" at the least significant bit then Vegas is dithering your CD output.
(Peter is 100% correct on his previous post about the +/- 1.0 normalization of the floating point audio, I figured that might be too much detail but it does help make sense of the dynamic range allowed by 16 or 24 bit audio. +1 corresponds to the maximum integer value allowed by your bit depth 2^16 or 2^24. All the fractional numbers between +1 and -1 one are quanitized to the intermediate integer values (more bits means more detail means better sound).
bgc

Rendering is to a CD is just a faster than realtime (usually) form of playback. Instead of going to your audio card it goes to a file.

The mix path for playback and rendering are identical, so you have a "What You Hear is What You Get".


Peter