Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Carl Taylor 43]

so if im running 24 bit audio files through VST plugins, and I mix down to a 24 bit file does this count as reducing the world length?

Also, does dithering only effect low level signals that you can harldy hear or does it effect the whole range?

[Bob Olhsson]

Ronny wrote on Tue, 05 April 2005 03:25

...Some people claim accumulated dither causes degradation.

Of course not dithering causes greater degradation that accumulates faster than the dither will.

"You can't hear it anyway" is often a cop out for lazy programming or inadequate computational power. Lots of things "you can't hear" become painfully audible after the audio has been diced, sliced and pulverized. MP3s are actually a good test. Really pristine audio sounds great while anything that has been truncated very much turns to dust.

[JackJohnston]

fuze wrote on Tue, 05 April 2005 04:57

Quote: Dither only one time only and always down. no. noise-shape only after all processing, once. dither always when reducing word-length. truncation always degrades more than dither. Quote: Dithering uses DSP techniques that try to get rid of the least neccesary parts of the signal. again no. dither is a random number added to the lowest bits so that, on rounding, the error  is random and doesn't track the signal. noise-shaping attempts to move even the dither noise into less audible places and shouldn't be done if other processing will take place. a daw should internally dither whenever it reduces word-length. you may have to select this... i don't know if cubase gives you the choice. always choose yes. steve parker.

Well, I guess we do have some disagreements. Maybe a bit more detail will clear this up for me.

Are you saying that dithering is not a DSP technique?

Or are you saying that it is not used to maintian as much perceived auditory information?

Are you saying that plug's don't have their own bit depth management when returning to 24 bits?

Do plugs always carry a bit depth of 32, 48, 64 or however much bit depth they have all the way to the mix bus?

Is all this audio data carried to the mix bus in structures, meaning is it all carried on the stack?

Thanks,

Jack

[LawrenceF]

Ok let's get back to the question at hand here.  I'm running 24-bit audio files in Cubase SX and a vocal track is being compressed and de-essed by a couple of Waves plugins.  Here's the specific question.

1.  Assume SX's dsp is operating say at 32-bit float point.
2.  I add a Waves RenComp and RenDeesser to a vocal track.

The initial poster's question was directed at something like this.  

Question:  When the plugin is "done" with that track what is the bit depth of the track that's "returned" to the mix bus?  24-bit 32-bit?  This really gets to the heart of the matter, or does it.  If the plug is returning 24-bits then it's performed so dither or trunication of it's own.  Or is it?

More directly to my interests...

If I have discreet tracks running from SX to a 24-bit digital mixer has each and every track that comes out of SX been truncated to 24-bit and do those tracks contain quantinization errors that might show up later as an overall slight degredation of sound?  The audio is (after dsp) going down from 32-bits or higher to 24-bit for output through a 24-bit lightpipe path to the console.  

I'm confused.  I'm gonna take this question to another forum.

Lawrence

[bobkatz]

3D Audio wrote on Mon, 04 April 2005 17:30

Dither should be used for only one thing. Repeat after me and memorize it. It will save you lots of questions and embarassment in the future. Remember IRB DUD. "If Reducing Bit Depth, Use Dither." That's the only time you have any excuse for using dither. Any other time, don't do it. Don't dither fades on 24-bit files in your 24-bit sessions. Just don't ever do it unless you're reducing bit depth and you'll be fine.

If you want to be totally correct, there's technically a bit more to it (pardon the pun) than Lynn's statement. The output of the fades (or any processing) in the 24 bit sessions will be longer than 24 bits. To be totally correct, they should be dithered down to 24 bits. But Pro Tools dithered mixer takes care of that automatically... in standard Pro Tools or HD. So Lynn's statement stands, but only because the mixer already takes the 48 bit output of the fade and turns that to 24 bits. I don't want to discuss rendered fades right now in this post, that's an interesting side turn).

However, in Pro Tools LE it is NOT taken care of automatically in the mixer, and a 32 bit float output is generated from any gain changes, off-center pans, eqs, processing. So it is technically correct to put in a 24 bit dithering plug on the output of the mixer. Is the problem audible if you don't do it? It will be extremely subtle at this level of dithering; you may not hear it, you may not notice it. The distortion is definitely measurable. If this were a perfect world and you weren't sending the material for 5 more generations of cumulative processing, each with its own attendant distortion, then I'm sure you could skip a generation or two of 24 bit dithering. But each step along the way adds its own distortion, and you might not discover that your material has gotten a little bit colder or hasher until then. Better to be safe and to be theoretically correct than sorry.

[cgc]

Quote:

Question: When the plugin is "done" with that track what is the bit depth of the track that's "returned" to the mix bus? 24-bit 32-bit?

VST and AudioUnit plug-ins receive 32 bit floats as input and output the very same.  I would say the vast majority of the available plugs use 32 bit single precision for all processing.  Any of them that use Altivec definitely do.

[Ronny]

Carl Taylor 43 wrote on Tue, 05 April 2005 09:10

so if im running 24 bit audio files through VST plugins, and I mix down to a 24 bit file does this count as reducing the world length?

Yes. It you know the plug-in is higher resolution than your native buss (it's often beneficial to convert 24 bit integer into 32 float for native processes) and the plug-in gives you the option to dither back to your 24 bit native buss, enable it. Many processors won't give you the option to dither the processors output, in that case it's taken care of by the processor and/or is out of your hands.

Quote:

Also, does dithering only effect low level signals that you can harldy hear or does it effect the whole range?

Truncation artifacts are most prevalent on fade outs and reverb tails that fade to black. Higher levels mask the artifacts, however they are still there and if you run through several non-dithered truncations, it can affect the total clarity, as truncation artifacts are accumulative and the resulting playback on a detailed system can give you a blanket sound robbing you of some detail. Can you hear this on the typical NS-10's or Alesis M1 mixing monitors, probably not, but when the material goes to mastering and can be heard on a detailed system in a tuned room, it beomes noticeable, especially if the ME is mastering further with digital processors. If you add dither to the noise floor you are only adding 3dB of noise, when the word is reduced, the dither noise is still at the LSB, if you reduce word again, the dither is again only peaking 3dB above the noise floor and still resides in the LSB. If your final destination is 16 bit and the dithered real level is -93dB, you are still going to be -12dB under the noise floor of a Neumann U87, common workhorse mic that has a very low equivalent noise of around -81dB with cardioid pattern switched in. Truncation artifacts on a fade out or reverb tail will sound like crackling, not unlike a gate that has the signal toggling back and forth across the threshold, making a break up in the audio type of noise as the gate rapidly cuts on and off and on and off until the signal falls below the threshold.  

[Carl Taylor 43]

thanks Ronny, thats cleared that one up for me