Re: Dithering from 32 Float to 24 Bit
 

I found a short explanation of why Avid has moved to the 32 bit float - 64 bit float mix setup in their Knowledge Base.

Q: Why did Avid switch from 24-bit processing/48-bit fixed summing to hardware that offers 32-bit floating-point processing/64-bit floating-point summing, and what does it mean for me?

A: There are several reasons for the transition. A 64-bit floating-point mix bus provides more than 1,000 dB of headroom, which is more than enough to handle the huge track counts that Pro Tools|HDX can deliver. Also, by moving the insert paths to a 32-bit floating-point format, Pro Tools|HDX offers much more dynamic range for plug-in processing, making it nearly impossible to clip the plug-ins, while also being able to handle greater than 24-bit audio file formats.

Many of you have told us that even with dual-precision plug-ins processing at 48-bit, the path between inserts on Pro Tools|HD was still limited to 24-bit — thus, limiting any gains. With Pro Tools|HDX, all data streams are maintained at full 32-bit floating point and then summed in the DSP mixer at 64 bits. Our beta customers told us that they could hear and appreciate the difference.

My guess is they will soon be adding a white paper explaining this in more detail. There is a thorough explanation of the PT 48 bit mixer at:

http://akmedia.digidesign.com/suppor...ixer_26688.pdf

The above white paper explains what used to be top shelf before HDX hit the scene.

The newer black HD I/O interfaces also have superior dynamic range specs in their converters. The ADC is 122 dB, the DAC is 125 dB. The 192 interfaces had ADC of 120 and DAC of 118 dB (A-weighted). So, with a very high quality amp and monitoring setup, and in a well treated listening environment, the specs would indicate they are an improvement.

But the style of music comes into play as well. Not all musicians are performing on a $3 million Stradivarius.

Re: Dithering from 32 Float to 24 Bit
 

If neither your recordings or your mix are clipping anywhere, you don't need to do anything.

What most people seems to not understand about the 32 bit float on PT10, is that it's an invisible feature. Your metters still work the same way as if you were working at 24 bit. You can actually switch the session parameters from 24 bit to 32 bit float without actual changes to the files.

The reason for this is that 32 bit float is the same as 24 bit, except it has a lot more headroom. If you clip a recording in 24 bit for 3 dbs, the file gets clipped. If you clip by 3 dbs while using 32 bit float, the mixer metter shows your recording clipping, but the file supports that clipping with that extra headroom (meaning that you can lower the clip gain and have no distortion at all).

If you convert your 32 bit float session to 24 bit, you don't need any dither because the only thing getting truncated is the extra headroom. Any wave information on the file bellow that extra headroom is perfectly compatible with 24 bit and show the same levels in the mixer.

The keyword here is "float". It's not 32 bit vs 24 bit. It's 32 bit float vs 24 bit. 32 bit float "peaks" at the same exact point as a 24 bit file. The difference is that it has the space to store more bits over that peak point, while 24 bit simply hits the roof and distorts.

HTH

Francisco

Re: Dithering from 32 Float to 24 Bit
 

That makes sense if dithering does nothing for a BTD from 24 to 24. Is this the case? What then is this X-factor Bob notices? Maybe that difference is just in comparing bouncing from a 32 float session to 32 float files versus bouncing to 24 bit files dithered from the same 32 float session? If that is the explanation, than you're saying that from 32 float to 24 I need do nothing. Exactly what I was wondering. I don't know if this is a contentious area or not, but it seems that there is a lot of misunderstanding...perhaps just my own.

Just put a good 24 bit tpdf dither like the one from psp xenon across your 2buss and listen while you hit play...

No bounce. Listen.
Bypass. On. Bypass. On.

Make up your mind for yourself afterwards...

Re: Dithering from 32 Float to 24 Bit
 

yep...exactly...

Re: Dithering from 32 Float to 24 Bit
 

Better than that... don't trust your ears. Trust exact science.

1) Send your final mix to 2 busses, both receiving exactly the same, both sending to monitors.

2) Insert your dither plugin in the first one.

3) Insert some plugin like Channel Strip on the second one and invert the phase.

Now, you can play your mix and start switching the bypass control on your dither plugin. If something is really different, you will hear sound. If everything is exactly the same on both busses, you will hear nothing.

Note: In any case, for your dither plugin to be able to work, it should be able to receive audio at 32 bits so it can dither to 24. If it starts by receiving only 24 bits (a therefore clipping over that), making it dither to 24 bit would be just silly.

------------------
EDIT: Make sure you are working with ADC enabled for this test to work properly!
------------------

HTH

Francisco

Dear Mister lab coat...

Do you know that channel strip is a bit transparent piece of code to start with?

Unless you use a "legacy" TDM system (where connections between inserts ARE at 24 bits - native, PTN and HDX are at 32 between inserts) the comment about "truncation" (not clipping btw) is irrelevent.

Of course the noise from dithering is always gonna show up as difference (as is any distortion products from truncation somewhere in the chain btw) invalidating the test.

Bob O. might have said it in this very thread, TPDF dither (flat triangular) adds 3dB of noise to the signal per bit reduced. However the audio information is completely preserved.

Truncation adds 6dB of distortion products per bit truncated. All info under the truncation point is lost.
Also by nature their inharmonic components these new components are very unmusical and disruptive to the pgm material even at very low level.


While you pounder this I have an album to master an a remix to do.

Cheers!

Re: Dithering from 32 Float to 24 Bit
 

In Pro Tools 8 and 9, BTD with and without Dither resulted in files that would CANCEL each other out when one was inverted.
This proves that if you "forget" to add the dither plug-in on your Master Fader before bouncing, that it would dither anyways!

This made it very hard to teach students the chapter(s) on dithering, bouncing to disk, etc.

I have ALWAYS been a fan of Buss-To-Track mixes, and then Exporting the subsequent Region / Clip as a File to get the "automatic" / Default dither when moving from 24 bit to 16bit.

I can't see why they would now change back to altering the "behind the scenes" dithering on Bounces from 32 to 24.


Does anyone have any idea how this all works in Pro Tools 10 if one BUSSES to a New Audio Track, BUT first changes Pro Tools Bit Depth from 24 to 16 BEFORE recording the new track? Dither or no Dither automatically applied or needed??

THAT is a good question.
Also did you check your null test with a bit scope?
I'm not sure residual products (such as dither noise) at lsb level would be detectable by ear unless you have an incredible monitoring chain and control room... Not sure but wondering?

Cheers!

Re: Dithering from 32 Float to 24 Bit
 

While I'm not discounting Bob O's observation while listening, I'm going to stick my neck out here and make a statement that may rile a few. Dithering when going from 32 bit float to 24 bit fixed is little different than adding dither after a file has already been truncated. 32 bit float only stores a 23 bit significand (+1 implied) and an 8 bit exponent. In other words, the audio portion of the 32-bit float word only contains 24 bits. It is understood that floating point has a variable noise floor since having an exponent just means that the decimal point can shift up or down. For the same reason, trying to dither upon conversion to fixed point would result in a file with low-level detail (within the dither) modulated by the amplitude of the signal. So if there really is detail to preserve, it may happen ok when the signal is -18 dBfs peak, but not when the signal is -2 dBfs peak.

Frankly, it's not realistic to believe that dithering from 32-bit float to 24-bit fixed is going to gain anything in the area of quality. That's why Avid has never bothered to implement 24-bit dither.