Guys! Should I use dithering (with/without noise-shaping) if I produce/synthesize/mix/master at 32/64 bit FLOAT (with use of 24-bit WAV files and VST-plugins/VSTi-synths that may work at higher bit depth with their double-precisions etc.) and render it all down to 24 bit?
ALSO: What's the aim of 24-bit dithering at tracking/mixing/master? Does it save something?
I use dithering only when 24 bit ->16 bit at final stage of mastering.

 

hey, mild pump milk,

hope this here me understands the workflow as you intended, meaning that for the mixing stage you are using 24 bit files. *yes*

use the best quality downsampling, bit depth changing, software that exists, then there is no real need for dithering until, to, 24 bit, the noise shaping, perhaps, even strictly prohibited (of course, my opinion, my preferences, my workflow, etc.). *yes*

from 24 bit to 16 bit depth it serves, even deserves, its own purpose, of which you seem to know of, so there is no need to going into it. *yes*

there might be very rare cases if, when, using older outboard gear, there are remains of near-nyquist-frequency problems to solve, but counting the last seven years, did not meet with any of these (caused, quite possibly, by sheer luck, which is one valid form of professionalism). *yes*

which software is the best is, again, perhaps the most personal choice, after comparing (ears, supported by most of the existing metering methods) nearly all of the possibilities, for me there exists one open source solution, plus one really hyperpriced but really good program, the gem of swiss engineering. *yes*

hope this is, will be, of some help, wish you all the best...

 

i Don't use noise-shaping i keep it off But if your going to put your Song on CD it needs to be 16 bit But if you are Going a Digital download go with 24 bit

 

No need for dithering when going from 32bit float to 24bit. Dithering has its rationality when going from a higher bit depth to a lower one.
As long as you might still edit the file (after rendering), you'll remain in 32bit float (IEEE) anyway. And if rendering means that you want to save the result, there is no bit difference between 32bit IEEE and 24bit integer. Why? Because 32bit IEEE is identical to 24bit integer when saving the file: The 32bit float word's mantissa has only 24 bits, and the exponent when saving the file is 0 by convention.
So what you get as 32bit float word when saving is: 24bit mantissa, multiplied with basis^exponent. As the exponent has to be 0, basis^exponent = 1. There we are, the 32bit float is nothing else than a 24bit integer word when saving.

Another thing is going from 64bit float into 24bit integer (or into 32bit float): There you have a bit difference from a 32bit mantissa down into 24bit, so dithering is theoretically justified. Whether it is really necessary in praxi at such bit resolutions, is another question.
And of course you have the bit difference and a dithering necessity when going down to 16bit integer (CD format).

You may check Stillwell PsychoDither, to my mind the best dithering software.

All the best to you and your projects...

 

You shouldn't need to dither from 64/32 to 24, ever. 24b has all the low noise floor you need.

Dithering at the very end master of the master bus is usual, since the bits shouldn't be changed afterwards. If you do anything afterwards, like a gain change, that totally screws the effect of the dither.

It is good practice to dither from 24>16bit, but I recently haven't done for several reasons (mainly influenced by 1 article which I can't find). If you have recorded material, that already has a noise floor, VST/is and other FX may also introduce a noise floor, meaning less need to dither.

Shaped dither theoretically makes the dither dither noise less perceptible. The standard shape that should sound not-crap is triangle, though there are more sophisticated shapes.

 

nadirtozenith
About SRC "sampling to 44k" I would choose full range with steep filter at 22 kHz (all high-cut filtering I can do when EQing, but weiss saracon does some filter at highs + some distortions/aliasing; sox cuts ultrahighs a bit, but it is very excellent) and I would choose without aliasing distortions (at -160 dB and lower) + linear phase (you will not hear pre and post rings about 22 kHz, will not be phase distortions, no peaks growing), so they are: FinalCD, iZotope 64-bit SRC, Brick 1.2 (for mac only).
Dithering (after SRC) I prefer: iZotope MBIT+, PSP Xenon 1.5/ PSP X-Dither, ToneBoosters Dither (fully customizable - do your own curves as you want).
But my question is "Use 24-bit dither or don't, if render down to 24 bit WAV (e.g. for further premastering tasks, final editing, level matchings, limiting etc.) before final-16bitditheredforCD-master?"

I want to know a lot of professional opinions

 

You may want to consider this:
http://www.youtube.com/watch?v=IRlohQw-1DY

http://productionadvice.co.uk/dither-or-distort/

http://productionadvice.co.uk/when-to-dither/

And this:
https://www.xiph.org/video/vid2.shtml

 

LONG Story Short
its like PICTURE QUALITY when u record a song, you need to export it down to 16-bit(CD,itunes,beatport,etc.-QUALITY). So, converting a 32,24bit song to 16 bit makes some quality lost. So, its like putting some LOW LEVEL white noise(dithering) in the song (mp3) to fill up the holes made by the conversion..

IT SHOULD BE PUT ONLY ON THE FINAL BOUNCE/EXPORT

 

It is *like* putting low level noise into the audio, but dithering is not a normal kind of noise at all. Noise have *some* properties of dither and vice versa, but it is *not* dither.

That's why editing gain after applying dither destroys the effect, and you should dither the audio file again. That's why no matter how noisy your recording is you should add dither when converting to a lower bit-depth anyway.

It is very desirable to dither even when converting 64/32bit to 24-bit conversion to avoid any digital quantisation errors, but any simple and free triangular dither will do a great job and no need for any noise shaping as dithering happens at about -144dB in case of 24-bit audio.

Regarding bit-depth, dithering, and encoding to MP3 and lossy formats, you can encode a 32-bit audio file directly without using any dither.

Cheers!

 

hey, mild pump milk,

sorry to have misunderstood your intended usage, your question (thought you use one program for composing then one other for mixing)...

this here me nearly always tries to write without traces of orthodox opinions, but... sometimes it comes up still, again then again...

as for the answer to your question quoted, do not (written as dogmatically as possible, of course). *yes*
if you want to be convinced, do couple of tests, with as different audio materials as possible, perhaps also through editing the files then redithering them to 16 bit, comparing these too, then decide. *yes*

bellegear has clearly explained why the bit depth actually does not change till down to 24 bit, but of course, you decide. *yes*

there might be theoretically convincing articles pro, con, me thinks using the ears, eyes, we are endowed with is more productive than reading too much of such material.

there also might be some more factors to consider, the intended audience, the intended end product quality, as, for example, the quality level of the average music consumption (smartphones with cheapest sounding earphones, mono radios in kitchens, baths, etc.) makes the whole problem nearly questionable, while at the audiophile end of the spectrum still might have some level of validity. *yes*

all the best for all of us, including non-dithered non-noise-shaped 24 bit files...

 

I know thats why I said "ITS LIKE"(to understand) i attended a live Course of DITHERING 3 hours hehe

 

You don't need to dither your sample rate hasent changed, and your bit depth is high enough.

 

Yeah, in case of 24 bit audio, quantisation errors are imperceptible down at -144 dB. It's hard enough to hear stuff at -80dB, let alone -144dB.

However, they are there and dithering can correct them making the audio [slightly] better/smoother sounding.

We had a similar discussion a year or two ago here @ AS. Very in depth and great discussion about dithering. Maybe look it up? It's a thread in which we learnt that noise and dither are different things, and you cannot apply noise of your choice instead of dither. I was interested in recording some noise from a state of the art analog device and using it as a dither. "No way Jose", no worky. But that doesn't prevent some guys on the net from selling "dithering CDs" with different noise samples of quality analog gear to poor people who are not so much in the know.

 

hey, SineWave,

do not know for sure which thread might that be but noise shaping, dithering, problems, questions, came up in one totally differently themed thread, about waves' hybrid delay (if this spongiformificated goo in me cranium does not play its usual tricks), perhaps worth one search...

that must have been some fucking really wicked brain, coming up with such some idea, selling one actual trouble source as the sole solution for one nearly non-existent problem...

all the best for all of us, but, pray, rather excluding dithering compact discs...

 

I know that other sample rates have other dithering and noise shaping types, but in case of 96 kHz will be noise shaping type for 96 kHz

 

hello, mild pump milk,

in most programs in common use, the dithering type, the noise shaping method, has nothing to do with the actual sample rate, with the bith depth, in nearly all solutions, the algorithms for different sample rates, different bit depths, are the same for said processes.

in other words, the different types of dithering, noise shaping, rather serve as solutions for different sounding audio material, not for different sample rates, different bit depths.

for the conversion between different sample rates, there is some validity in the usage of dithering.

at higher bit depths, as so many times stated above by countless users, there is no need for these.

so, in the case of of better than 16 bit resolutions, if there is sample rate conversion, you test if the results are better with dithering.

if there is no sample rate conversion, only going from 64 bit or from 32 bit to 24 bit, you can test again, but in nearly all cases, you waste your time.

also in the case of your 24 bit files, if you still intend to edit them in any way, in any sense, with any action, the formerly done dither becomes totally pointless, if done again, the results will be only worse.

in the case of the mastering stage to 16 bits, the last process in your chain is the dithering, with the best sounding type chosen, with but rather without noise shaping, very carefully defined in its type, in its intensity.

noise shaping is better to be understood as one last resort in the action chain, used with audio material where there are loads of low level passages, fade ins, fade outs, in quite special places in the audible spectrum, very rarely required even then indeed, because in nearly all cases, the correctly chosen dithering type eliminates all problems.

pray do try to search for other threads about these questions, including the one mentioned by sinewave, the one mentined by this here self of mine, these might serve as enlightening reading material...

all the best for all of us, including the increasingly better understanding of mastering processes...

later edit, post scriptum, before forget, the usage of the high bit depth floating point audio formats within workstations, besides other purposes, aims upwards too, rather for headroom, quite similarly to the busses used in certain older analogue mixing desks, not only for the lower noise floor, for which the 24 bit depth is nearly enough... *yes*

 

Nadirtozenith
No, these things are dependent. Just test it and listen (8 or 12 bit as example to listen to - they are different at different sample rates, higher rate - higher shift of frequencies of noise shaping).
PSP X-Dither - analyze dithering with noise at 44k project and 192k project.
iZotope Ozone's MBIT+ - the same thing. 8 bit dithering with other noise shaping types @ 192kHz I don't hear, it is too low, ultrasonic.
It's so because of different ranges.

 

all right, or it will be so soon,

the topic has reached the point where one might really begin to feel rather compelled to leave it.

this here self of mine chooses to act the same way.

many thanks for the possibility to express the opinions expressed, many thanks for the site for providing with the means to do so.

furthermore, deluge of gratitude particles accelerates vectoring in your direction from the thank you emitter system, for your questions.

from all that possible is, wish you the best...

 

tested today
saved files to 24 bit, 32 bit int, 32 bit float without ditherings:
24 much noisier than 32 bit int, and 32 bit float made no distortions (32 bit integer noisier than 32 bit float). 64 bit float gave better results. (Exported and Tested with RX4)

Nadirtozenith
Have I said something bad?
Sorry.
Just my tests today and earlier show that projects with different sample rates give veeeery different results, also between 44 and 48 there is a difference.
I think it's all OK, nadirtozenith?

 

Please have a look at iZotope's great "Dithering with Ozone" guide (PDF).

Yes, there are some noise shaping algorithms that are designed for specific sample rates:

No, you cannot hear a difference between (real) 24- and 32-bit, because you would need a DAC with an SNR of more than 140 dB.