The No.1 Website for Pro Audio
Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
15th December 2017
#511
Lives for gear

Quote:
Originally Posted by 5.333V
Thanks pentagon.
How about when going from floating to fixed? Is it also trivial math?
Trivial math but unlike a fixed to fixed conversion (which is either padding to go up bits or truncating with a lop-off of bits to go down), float to fixed leaves open how to handle rounding. In a stupidly simple example: 1.5 float could be 1 fixed or 2 fixed. That is left to the programmer. Dithering float is to randomise these quantisation noise errors of the resultant fixed number (ie, randomising the rounding direction)
16th December 2017
#512
Gear Nut
Quote:
Originally Posted by pentagon
Trivial math but unlike a fixed to fixed conversion (which is either padding to go up bits or truncating with a lop-off of bits to go down), float to fixed leaves open how to handle rounding. In a stupidly simple example: 1.5 float could be 1 fixed or 2 fixed. That is left to the programmer. Dithering float is to randomise these quantisation noise errors of the resultant fixed number (ie, randomising the rounding direction)
I thought you couldn't dither Float; e.g. dither 64 bit floating point down to 32 bit floating point.

Sorry if I'm misleading things here.
16th December 2017
#513
Lives for gear

One generally can't dither floating point. Because in this context, the quantization affects more than the LSBs. This process is somewhat chaotic, so the resulting partials (the generated distortion) practically turn into white noise.

Not that it would matter at this precision anyway, floating point or not. Music signals have a noise floor miles beyond the 32bit float target.
16th December 2017
#514
Gear Maniac

hmm... i checked kazrogs master dither again with voxengo span for LR phase correlation. this time i put alot of gain before the meter and evenually the phase meter moves the indicator from 0 up towards 1... so i think i was wrong in my statement of it being individual random between L and R...

i have analysed the dither on my cranesong hedd and with its DSP engaged a bit aswell. haha this is really funky stuff. i wonder if not, it is the dither that is being harmonized when using its harmonic effects, and not the actual signal in this one. in 16bit and 20bit mode, their is noise injected. they are a bit different to each other. but both have additional harmonic content starting with 500hz and spread through all harmonics up to band limit. 16bit has some noise shape, starting around 10khz and up plus a bump in the low en which i guess is a DC shift component while the 20bit is more flat with out the dc shift. both seems to be individually random between L and R. when u engage the triode/pentode/tape controls, the phase correlation shifts. with triode it gets full phase correlation just by turning the knob up a mm. pentode and tape changes the phase correlation more gentle and seems to be optimum with them hanging around 12oclock. this DSP is obtainable as a vst plugin called the Phoenix (aimed for pootools though). i wonder if it could be used only as a dither plugin with good results.
16th December 2017
#515
Gear Nut
Quote:
Originally Posted by FabienTDR
One generally can't dither floating point. Because in this context, the quantization affects more than the LSBs. This process is somewhat chaotic, so the resulting partials (the generated distortion) practically turn into white noise.

Not that it would matter at this precision anyway, floating point or not. Music signals have a noise floor miles beyond the 32bit float target.
When will we see a 64-bit-scaled Bit Meter from TDL? Is the only thing that's missing from the Metering section in Limiter6 if I may selfishlesly ask
20th February 2018
#516
Gear Maniac

i just want to correct my self.

i stated that reapers dither plugin was not generating randomized noise between L and R. while that is true, there is another dither plugin in reaper which actually does. and its called "Bit reduction/Dither w/noise shaping"

inside the plugin you can choose bitdepth up to 32bit, with/without noise shaping, tpdf/tpdf with highpass, and variable LSB.

i have found some possible advntages in reaper using 32bit dither at the end of all channels in a mix projekt.

i also have a 24bit dither on the monitor output but just tried putting another 32bit dither before it and the last mixbuss pluging. that would be going from 32bitFloat > 32bit Fixd > 24Bit fixd > DA

need to listen more to be sure but does anybody else do similar in reaper?
21st February 2018
#517
Gear Nut
Quote:
Originally Posted by 5.333V
i also have a 24bit dither on the monitor output but just tried putting another 32bit dither before it and the last mixbuss pluging. that would be going from 32bitFloat > 32bit Fixd > 24Bit fixd > DA
Why would you do that?

If you're mixing in a 64bit floating point enviroment as REAPER, as far as I have understood it, the way to go is to dither down to 24 fixed-point before going out the DAC.

I don't see the point of dithering down to 32 bit except for the few modern 32bit DAC that appeared in the market recently or maybe at times when in need to go in and out of the DAW itself - like when I reach out for iZotope RX for processing (I don't understand why RX, Sound Forge, and many others just won't support still opening a 64 bit WAV file).

Even so, as stated in ths forum by @FabienTDR , @chrisj and many others, there's no actual dithering possible for going from 64 floating point to 32 bit floating point, so I actually just render my 64 bit stems to a 32 bit floating point file (which RX will open) with no dither or noise shaping whatsoever, and I presume no truncation or tangible degradation is happening to my audio.

Hope I'm making it clearer for you or else somebody else will please correct me!
21st February 2018
#518
Gear Maniac

hi morfi!

doing a 32 dither before the final 24 bit dither before my monitor output is more or less experimenting. i wanted to see if a stairway dither approach would benefit when going from 32/64 floating to 24fixd.

i kind of like what i hear so far but still wount say anything definitive yet.

yes reaper is working in 64bit floating internally, but that is only when any prosessing has been done with any of reapers own prosessors for ex faders or their own plugins. i use some TDR plugins for ex which is 32bit in 32bit out floating. so from having a source soundclip in 24bit fixd, the wordlength will jump up and down multiple times through my hole chain.

dither is said to be used when wordlength is reduced. going from 64 to 32 is just that and going from 32float to 32fixd is as far as i understand a kind of truncation. so i have been experimenting with this latly.

ignoring when wordlength expands while paying attention to when it shrinks.

please do some experimenting your self and tell me i hear things that are not there. listen to bass flow, stereo width, and harshness everywhere.
21st February 2018
#519
Gear Nut
Quote:
Originally Posted by 5.333V
dither is said to be used when wordlength is reduced. going from 64 to 32 is just that and going from 32float to 32fixd is as far as i understand a kind of truncation. so i have been experimenting with this latly.
Thank you 5.333!

I'm also very interested in this particular subject of 64 => 32 and will have to do some experimenting myself too. By all means I'm not saying you're wrong and I'd like to be corrected, or confirmed, in my opinions as well by the people who really know about this more deeply than me.

I'm confused here in this: is 64 => 32 bit really "going down". I tend to think that there is no TRUNCATION taking place in such procedure, but I do wonder what a 32-bit-processing plug-in (such as TDRs) will do when receiving a 64-bit-wordlenght signal.
21st February 2018
#520
Lives for gear

32bit I/O doesn't imply any processing precision. It can be anything.

A truncation from 64bit to 32bit happens around -150dB and is likely self-dithered anyway due to the quasi chaotic truncation behavior of floating point, and the likely presence of noise. You'd have a hard time finding any trace of an issue with in.

64bit is a completely hilarious format for audio transport. The only reason why it is offered is that modern CPUs don't care much, by default, they process 32bit like a 64bit number. It costs nearly nothing. But to be clear, 64bit has 32bit times as many steps as 32bit, not just twice as much.

32bit gives you around 150dB of precision, the threshold of audibility (roughly -100dB) is ~316 times higher.
64bit gives you about 300dB of precision, the threshold of audibility is ~3.162.277.660.168 times higher.

That's *slightly* exaggerated for audio transport.

Now when you calculate a recursive algorithm such as a standart IIR filter (which feeds-back previous values and quickly accumulates their little errors), this increased precision can make sense. But this is totally independent from I/O precision. The dev is free to do whatever he finds adequate for his task.

The main reason why so many plugins still don't support the 64bit VST extension is that both VST and the other formats still use 32bit as default, and the sheer lack of technical necessity for 64bit transport. Once most support 64bit by default, we'll follow. It's just a matter flipping a switch, but not worth the trouble for us today because we'd have to support more cases, and test them every time we release something.

Last edited by FabienTDR; 21st February 2018 at 09:11 PM..
24th February 2018
#521
Gear Nut
Thank you so much Fabien!

Quote:
Originally Posted by FabienTDR
32bit gives you around 150dB of precision, the threshold of audibility (roughly -100dB) is ~316 times higher.
64bit gives you about 300dB of precision, the threshold of audibility is ~3.162.277.660.168 times higher.
So are you implying that putting a dither plugin for going from 64 bit floating-point (the DAW's resolution) to 32 bit floating-point (a plug-ins resolution) is needless?

Should we configure our mixing projects in 64 bit floating point to begin with or will it sound "as good as" a 32 bit floating point session?

Quote:
Originally Posted by FabienTDR
Now when you calculate a recursive algorithm such as a standart IIR filter (which feeds-back previous values and quickly accumulates their little errors), this increased precision can make sense. But this is totally independent from I/O precision. The dev is free to do whatever he finds adequate for his task..
This is definitely true since there are already a bunch of plug-ins calculating their filters at 80 bit or even more but then again coming in and out at 64 bit floating point for the DAW.
24th February 2018
#522
Lives for gear

Quote:
Originally Posted by morfi
So are you implying that putting a dither plugin for going from 64 bit floating-point (the DAW's resolution) to 32 bit floating-point (a plug-ins resolution) is needless?
Yes, it would have no effect on floating point quantization because it happens at many signal levels, not just near the floor. Any attempt to randomize these levels creates much bigger trouble, because any strict structure or rule defies randomness.
Good news is that the irregularity of this quantization tends to self dither quite well, in particular with complex signals such as audio.

With fixed point, all truncation predictably happens at the lowest levels (like an instant gate). So preventing audible distortion is easy: You simply randomize the everything up to the target floor's 2 least significant bits, and thus prevent the production of partials/tones during the process. If anything appears, it will be fully random, i.e. noise. No partials appear.

Fixed point can be perfectly dithered, floating point cannot.

Quote:
Originally Posted by morfi
Should we configure our mixing projects in 64 bit floating point to begin with or will it sound "as good as" a 32 bit floating point session?
In doubt, try what happens when truncating 64bit float audio to 32bit float, or simply use what your daw offers by default.
24th February 2018
#523
Gear Nut
Thank you @FabienTDR. I appreciate the in-depth explanation on the subject (I think everyone here does). Posts like yours bring value to this forum for sure.

I use REAPER which is set by default to 64 bit floating point. Now I believe Justin Frankel and Cockos don't just put "more for the better", so I always thought there was a sensible reason to work at this precision in spite of not being the industry standard, i.e. it sounded better than 32 bit floating point mixing...
26th February 2018
#524
Gear Guru
I always try to settle on a decision.
28th February 2018
#525
Lives for gear
Reading gearslutz when I have a lot on.
16th December 2018
#526
Here for the gear

Hi, @chrisj!
I have a question. If I go from 96 Khz 64bit file straight to 16 bit. And use your Ditherbox. Should I use the same 16 bit dither, as I use when i dither from 24bit to 16 bits? Or you recommend doing it two steps 64-24 24-16? How it's possible doing that with dither and BitshiftGain?
Thanks!
16th December 2018
#527
Airwindows

Quote:
Originally Posted by i_magi_nation
Hi, @chrisj!
I have a question. If I go from 96 Khz 64bit file straight to 16 bit. And use your Ditherbox. Should I use the same 16 bit dither, as I use when i dither from 24bit to 16 bits? Or you recommend doing it two steps 64-24 24-16? How it's possible doing that with dither and BitshiftGain?
Thanks!
Some of that isn't a general dither question, the principle is the same whether or not you're using Ditherbox or my other stuff (and BitShiftGain is just a tidy way to do gain offsets that correspond to bits in the exponent of the floating point word, or indeed a fixed point word)

The general answer is, it's the same as going from 24 bit and you should go directly, there's no point going to intermediate stages. If you're going to 16 bit, dither to that and go straight to 16 bit. If you're going to 24 bit, dither to that and go straight to 24 bit: there's no use in going part way and then going and truncating some more and dithering again.

I will say this: you're going to want to SRC before final wordlength reduction. I assume you're also going to 44.1K and your 16 bit is for CDs? Go to 44.1K first, at your 64 (or 32) bit working resolution, like the DAW gives you (some will allow 64 bit, stuff on Mac based on CoreAudio is 32 bit) and once you've got 44.1K sample rate, that's when you go to 16 bit. If you went to 16 bit and then converted to 44.1K also at 16 bit, that's not as good.
16th December 2018
#528
Airwindows

Boy did this thread get weird since I last saw it

Quote:
Originally Posted by FabienTDR
One generally can't dither floating point. Because in this context, the quantization affects more than the LSBs. This process is somewhat chaotic, so the resulting partials (the generated distortion) practically turn into white noise.

Not that it would matter at this precision anyway, floating point or not. Music signals have a noise floor miles beyond the 32bit float target.
Not to contradict Fabien but nope

When you're working at 32 bit float, the loudest part of your signal (0 to -6dB) is always a smidge below 24 bit fixed point, because the exponent of the floating point works to constantly scale that up and down as needed. Then, other parts of the audio (say, a reverb tail) are using other scalings, still with that 24 bit mantissa, but the whole thing's scaled down so it's not unlike if you had 300 bit fixed point or more: the quiet stuff gets all the data it needs to stay pristine way down into the faintest imaginable vibrations.

If you could have audio 100dB above 0 and play it back and you were using floating point, the coarseness of it would be like 8 bit audio compared to normal-volume sounds. Floating point scales up and down, sample by sample. We just can't use more than a very small range of it, for audio.

You could dither floating point but it would be a CPU-munching pain in the butt, because you'd have to calculate what, 256 different dither levels, twice if you want TPDF, and then most of the time it'd be unnoticable. TPDF dithering the mantissa at levels of -60 or so would be nuts: that's effectively trying to dither 24 bits BELOW -60dB, which is very much 'don't bother'. Remember this is not 'on output' but (assuming float variables everywhere) on every math calculation. Thousands of very cumbersome 256-level dither calculations for every sample isn't reasonable (I think most of us just use doubles for plugin internal calculations)

You can noise shape floating point and I do that in pretty much every plugin. I've just put out a non-dither plugin that uses the same mechanism (quantization) and comes in a 24 bit version, and has noise shaping… which ended up compensating for the quantization and directly re-integrating it into the sound as audio rather than noise. I've had to zip out a bugfix that clips the output, because people were trying to use this post-limiting and if you retain the information lost in 24 bit truncation, you get overs of nearly 2dB… from the 24 bit truncation's energy alone. That's the same amount of undirected energy generated from the loud half of every 32 bit calculation done in a plugin or DAW algorithm.

It's just that if you don't dither it, you get hash, and if you dither it, you get uniform noise that can't build up in any way. Instead, I had an algorithm that redirected this energy… and got massive overs. We're really looking at a lot of energy from these tiny truncations, simply because there are so many of them per second.
16th December 2018
#529
Lives for gear

Tried a few, like the goodhertz dither.
16th December 2018
#530
Gear Guru
I try to not let folks know I'm undecided.
16th December 2018
#531
Motown legend

It's worth mentioning that there has been plenty of broken dither in software and some apps don't truncate after dithering. No dither sounding better means one of those two situations is happening.
17th December 2018
#532
Here for the gear

@chrisj Thanks for reply! It's very clear. Can you please explain about "like the DAW gives you (some will allow 64 bit, stuff on Mac based on CoreAudio is 32 bit)". Do you mean, that Reaper on Mac, which processes and renders at 64 bit actually "lies", and works on 32 bit for real?

Small question about SRC's. Yes, of course SRC is before bit depth convertion. I use mostly Izotope SRC's, and sometimes Apple's ones in program called sample manager. Which SRC you prefer working with? Did you think to develop your SRC algorithm (I would expect your SRC would be the most "musical" one in history, after what you did to dithers).
17th December 2018
#533
Gear Maniac

This is one of the best threads on here, thanks especially to Fabien, ChrisJ and Mr Ohlson for their knowledge.

Correct me if I am wrong, but after reading this I get the impression the best user case is to render/bounce everything in 32 Bit floating point (if you freeze tracks, if you bounce down two outputs, if you export etc.) and send your Mastering engineer at 32Bit floating point file at your project sample rate.

He can then dither this once when making the final Master file at whatever bit rate the final format demands.
17th December 2018
#534
Lives for gear

Quote:
Originally Posted by Mannymac
This is one of the best threads on here, thanks especially to Fabien, ChrisJ and Mr Ohlson for their knowledge.

Correct me if I am wrong, but after reading this I get the impression the best user case is to render/bounce everything in 32 Bit floating point (if you freeze tracks, if you bounce down two outputs, if you export etc.) and send your Mastering engineer at 32Bit floating point file at your project sample rate.

He can then dither this once when making the final Master file at whatever bit rate the final format demands.
That’s what I do
17th December 2018
#535
Airwindows

Quote:
Originally Posted by i_magi_nation
@chrisj Thanks for reply! It's very clear. Can you please explain about "like the DAW gives you (some will allow 64 bit, stuff on Mac based on CoreAudio is 32 bit)". Do you mean, that Reaper on Mac, which processes and renders at 64 bit actually "lies", and works on 32 bit for real?

Small question about SRC's. Yes, of course SRC is before bit depth convertion. I use mostly Izotope SRC's, and sometimes Apple's ones in program called sample manager. Which SRC you prefer working with? Did you think to develop your SRC algorithm (I would expect your SRC would be the most "musical" one in history, after what you did to dithers).
Nope, I don't have a SRC. What I use is an old version of Audacity, with the 'Secret Rabbit Code' library, which is just sinc based interpolation. I don't use the highest quality version but the medium quality: for that particular library, it has ideal out of band rejection and differs in terms of how close to Nyquist it can get (more quality, steeper brickwall at Nyquist) and I liked the medium one.

I've also done this for audio that didn't have super-high-frequency stuff: put on an averaging filter at around 2 samples, then 'resampled' with no kind of SRC at all. This can sound fantastic but only if there is no audio information up over 20K in the first place, otherwise it's a complete disaster… and it can only sound 'fantastic' if you can appreciate dark, treble-lacking audio, otherwise it will seem totally pointless. But for special case audio that has no highs, it can be great. It's not 'correct', and it's totally inflexible: if there are super-highs you can't use it, full stop.

Regarding Reaper and CoreAudio: if Reaper has a 64-bit buss, that isn't CoreAudio. If Reaper is allowing you to place Audio Units on a 64-bit mix, then it's truncating the information to 32 bit before handing it off to the Audio Unit, and taking the Audio Unit's 32 bit output in turn. So, IF you can still use AUs on Mac Reaper at 64 bit, that's what it's doing.

VSTs in Reaper need to implement the 'double replacing' code (or whatever they use that takes double word lengths on input and output) to run on a 64-bit buss. If they don't, and they only have code that takes 'floats', they'll do the same thing. This is different from if the computer's RAM uses 32 bit or 64 bit: we're talking about how the DAW stores sample values, not about how big an area of RAM it can store them in.

Either way, it's not so much 'lying' as 'when you give the audio to this particular plugin, does it get truncated coming in and going out?'. The DAW can still be running 64 bit for everything it does, but if it needs to talk to a plugin that's only talking 32 bit words, it'll throw the remainder away. That, or not be able to talk to the plugin at all.
17th December 2018
#536
Lives for gear

Quote:
Originally Posted by t_d
this is an eye-opening presentation on digital audio, especially the part on dither (which begins around 12:00.. but do spend 25 minutes and watch the whole thing)

A succinct and useful explanation of the theory, and I found it fairly easy to tolerate the presenter's smug demeanor and slightly condescending attitude, but I take strong exception to his statement regarding the potential spoiling of masters by not employing dither, here:

I can think of one of my masters that was irrevocably spoiled by the lack of dither employed by the mastering engineer back in the day when the folk wisdom was "why add noise?" (about 20 years ago). Both the recording and the monitors were pretty high definition, though—I'm not sure I would have heard the damage under other circumstances.
17th December 2018
#537
Here for the gear

@chrisj Thanks, for totally amazing explanation. This information should be known to every user, before even start mixing!
First created a track with 64'bit wave file in Reaper and in the end of the chain putted an bit depth analyser of called "Bitter" by Stillwell Audio.
On the Master Bus I putted the same plugin once again. Here what I've got.
In case of Au's it always goes 32 bit, (from 64) you're right. If it's 24 i remains there.
In the case of VST's - the same story with most of the plugins, except most of Your's (not ADClip) and Black Rooster Audio's.
While using your plugins it goes 64 bit on a track. If the project is 64 bit, then on a Master bus also, but if the Project is say 32bit, on the Master it goes to the project's bit depth.
If the audio is 24 bit, still your VST's take it to 64 bit on a track, even if the project is 32 (then it goes back to 32 on a Master). In the case of the rest of the developer's plugin VST's the 24 bit audio goes to 32 bit.
Dropbox - Screenshot 2018-12-18 00.21.20.png
Dropbox - Screenshot 2018-12-18 00.21.56.png
Dropbox - Screenshot 2018-12-18 00.22.49.png

Another question: You mentioned a plugin, that calculates everything in 24 bit version, and has noise shaping. Do you have an example of such a plugin and when exactly you use it? 32 to 24? 32 to 16 also, or dither?
Second thing, maybe would a super-heavy plugin that calculates dither for FP actually worth creating (the question how much heavy it is). Just in order to dither one 32 bit track on a mastering, while all you have is one track anyway.

Last edited by i_magi_nation; 18th December 2018 at 12:55 AM.. Reason: Mistake
18th December 2018
#538
Airwindows

Quote:
Originally Posted by Russell Dawkins
A succinct and useful explanation of the theory, and I found it fairly easy to tolerate the presenter's smug demeanor and slightly condescending attitude, but I take strong exception to his statement regarding the potential spoiling of masters by not employing dither, here:

I can think of one of my masters that was irrevocably spoiled by the lack of dither employed by the mastering engineer back in the day when the folk wisdom was "why add noise?" (about 20 years ago). Both the recording and the monitors were pretty high definition, though—I'm not sure I would have heard the damage under other circumstances.
That's Monty. His opinions on digital audio don't belong on 'High end' (or in a mastering forum) but for all that, Xiph.org's stuff is good for what it is: lossy compression.

But then, JJ's been yelled at for bringing up dither, so what chance do the rest of us have?

I have an interesting cat to put among the pigeons, soon. I've been doing a plugin called Dither Me Timbers that's actually not dither: instead it uses quantization to apply a high-cut filter as hard as possible. It's just making different conditions for which directions to truncate. I added noise shaping and that ended up producing a LOT of unexpected energy, so I've got a version in the works that only quantizes, no noise shaping.

It also quantizes with the opposite intent: brightening as hard as possible, rather than dulling.

It works at 24 bit.

So if people can hear what this really-quite-intense form of EQ is doing, to the point of being able to pick it out blind, they are PROVABLY hearing the 24th bit. (it's also possible to use BitShiftGain and get it to whatever threshold you want, but it's built to work at the 24th bit)

That will be interesting
18th December 2018
#539
Airwindows

Quote:
Originally Posted by i_magi_nation
@chrisj Thanks, for totally amazing explanation. This information should be known to every user, before even start mixing!
First created a track with 64'bit wave file in Reaper and in the end of the chain putted an bit depth analyser of called "Bitter" by Stillwell Audio.
On the Master Bus I putted the same plugin once again. Here what I've got.
In case of Au's it always goes 32 bit, (from 64) you're right. If it's 24 i remains there.
In the case of VST's - the same story with most of the plugins, except most of Your's (not ADClip) and Black Rooster Audio's.
While using your plugins it goes 64 bit on a track. If the project is 64 bit, then on a Master bus also, but if the Project is say 32bit, on the Master it goes to the project's bit depth.
If the audio is 24 bit, still your VST's take it to 64 bit on a track, even if the project is 32 (then it goes back to 32 on a Master). In the case of the rest of the developer's plugin VST's the 24 bit audio goes to 32 bit.
Yes. My stuff uses denormalization code that takes the output to 64 bits (where available) rather than run the CPU extra hard calculating the info between that 64th bit, and zero. I don't know what Black Rooster's doing, but they might be up to the same thing in some way. My stuff actually runs at greater than 64 bit internally, most of the time, but then it noise shapes to either 64 or 32 bit float as needed.

ADClip is special. I'm pretty sure what's going on there is pass-through. You'll find that if the output isn't clipping, and the 'reinforcements' in treble and bass aren't active, it's passing through the input word untouched. I thought it was important to have a safety clipper that really, really got out of the way when not in use, so it bypasses itself and that's why you don't see 64 bit information: when not actively clipping it's like the plugin isn't there at all. I don't remember if I have denormalization in there when clipping: if you used a 32-bit plugin to do a gain boost into ADClip then you could see whether it lights up the 64 bit indicators when actively clipping. I think BitShiftGain would also work, because by definition it must not touch the mantissa at all, only the exponent.
18th December 2018
#540
Here for the gear

@chrisj And once again thank you - it was very important information! It means, that although the Airwindows plugins take the audio to higher bit depth, I don't need to think of dithering, after using your plugin, because it's already noise shaped. What about Reaper on mac, anyway all plugins take it to 32 bits, so as was said, the optimal is probably rendering it 32 bit float before Master. What about ADClip - this plugin totally saved a project, that I was mixing. When the client asked me to increase the volume of the whole concert for about 3-4 db, and Izotope Ozone 8 failed in that operation, after researching lot's of clippers out there, I discovered ADClip.
Here's the result:

Last edited by i_magi_nation; 18th December 2018 at 02:03 PM.. Reason: Mistake
Top Mentioned Products
Topic:

## Welcome to the Gearslutz Pro Audio Community!

###### Registration benefits include:
• The ability to reply to and create new discussions
• Interact with VIP industry experts in our guest Q&As
• Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD \$20/year
• Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.