Working in 32bit Floating Point

Discussion in 'Mixing and Mastering' started by relexted, Sep 23, 2018.

?

Your #1 reason for working in 32bit Floating Point

  1. Infinite headroom

  2. Gain staging is ancient

  3. It sounds better

  4. Because it's possible

  5. Eliminate truncation distortion

  6. Mastering engineers asks for a 32bit file

  7. So I can clip my converter and do magic

  8. Bigger files -> higher quality

  9. Who doesn't love the red light district

  10. Never have to pay attention to the meters

  11. My AD/DA converter operates in 32bit

  12. Other! (please specify in the comments)

Results are only viewable after voting.
  1. relexted

    relexted Kapellmeister

    Joined:
    Jun 18, 2011
    Messages:
    210
    Likes Received:
    45
    Location:
    Amsterdam
    Hi everyone,
    This topic came up recently in other threads.
    I noticed I don't really have a clue why to work in 32bit floating point.
    There are numerous reasons stated, yet can't we avoid these reasons and get the same results in 24bit?
    How do you apply 32bit floating point in your music and what are the real benefits?
    Thanks!
     
    • Useful Useful x 2
    • Like Like x 1
    • List
  2.  
  3. Andrew

    Andrew AudioSEX Maestro Staff Member

    Joined:
    Oct 24, 2011
    Messages:
    1,824
    Likes Received:
    1,012
    Location:
    Between worlds
    First off, I don't work in 32bit FP, but in 64bit FP (dual precision).
    Benefits are easily spotted - as long as FP is implemented throughout the chain, no clipping can occur anywhere.
    Rather noticeable difference when rendering final project- all that's required to be done to the render is to normalize it prior to converting to integer values.
    This saves time, effort and sound quality (no limiter on master track), without having to tinker with mastering level, so that the resulting 24bit WAV won't clip.
     
    • Interesting Interesting x 4
    • Agree Agree x 2
    • Like Like x 1
    • Useful Useful x 1
    • List
  4. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    1,630
    Likes Received:
    1,148
    Location:
    Russia
    What is the point of using 32 bit float, if your DAW internally work in 64 bit float double precision, most of them do. Because 32 bit float is very old (but still in use though) , even 64 bit processing was introduced in iZotope Ozone 3 many years ago. And plugins too, not all but majority in 64 bit float double . Some even work in 80 bit float extended, PSP Master 2, airwindows...
    More headroom, More footroom, less quantisation distortions and so on.
    It is internally, inside your PC only, Digital to Digital etc. There algos work, not your converters. When you do AD or DA even 24 bit converters do not work in 24 bit practically, bit less/worse than that. But in PC processing happens without conversions.
    With float you can avoid clipping.
    Less truncation, More dynamic range.
     
  5. No Avenger

    No Avenger Audiosexual

    Joined:
    Jul 19, 2017
    Messages:
    1,634
    Likes Received:
    1,153
    Location:
    Europe
    Seems I'm the cave man here, still working with 32bit FP. [​IMG]

    The main reasons have already been stated, no clipping above 0dBFS, neither in the channels, nor in the main out. No doubt you can avoid clipping with fixed point, it's just easier this way.
     
    • Like Like x 1
    • Useful Useful x 1
    • List
  6. Baxter

    Baxter Audiosexual

    Joined:
    Jul 20, 2011
    Messages:
    2,374
    Likes Received:
    1,250
    Location:
    Sweden
    I work/record in 24bit (144dB of dynamics), but since my DAW (and many other DAWs) are working in 32bit floating point by default, it doesn't really matter...or makes sense.

    I don't care for 64bit yet. 32bit float internal processing has 1680dB dynamic range, which is weeeeelllll beyond what I need.

    So to answer you: your DAW is already working in 32bit floating point (or deeper).
     
  7. Medrewb

    Medrewb Platinum Record

    Joined:
    Oct 11, 2014
    Messages:
    488
    Likes Received:
    189
    24 bit wav and 16bit wav still cuts the audio at 0db so it will still have digital clipping after 0db. You need to export it in 32 bit float to have no clipping
     
  8. Medrewb

    Medrewb Platinum Record

    Joined:
    Oct 11, 2014
    Messages:
    488
    Likes Received:
    189
    All DAWs operate at 32 bit floating point internally but I think Cubase uses 64 bit floating point. Lots of old men who are analog fan gays in other forum do not understand this. You can't clip a 32 bit floating point (practically). The only stage it will clip is in your AD/DA and analog modeled plugins.
     
  9. Medrewb

    Medrewb Platinum Record

    Joined:
    Oct 11, 2014
    Messages:
    488
    Likes Received:
    189
    • Interesting Interesting x 1
    • List
  10. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    1,630
    Likes Received:
    1,148
    Location:
    Russia
    Majority of DAWs are 64 bit float. Maybe 50/50, but I think more.
    REAPER is 64 bit float since long ago.
    Cubase is 64 bit float since late 2017. Steinberg is so revolutionary, evolutionary and cutting-edge hi-tech, yeah?) Since they became a world standard many years ago, and introduced VST and so on, why they introduced 64 bit float engine only recently, AFTER cheapest REAPER? they should have done it in early 2000s, if not late 90s :D
     
    Last edited: Sep 24, 2018
    • Like Like x 1
    • Funny Funny x 1
    • List
  11. SineWave

    SineWave Audiosexual

    Joined:
    Sep 4, 2011
    Messages:
    2,460
    Likes Received:
    1,365
    Location:
    The darkest pit of my mind
    My main point for making some files 32-bit FP is to keep them editable without any sound degradation [truncation noise]. Nothing else. Not caring much about bits, otherwise. I mean, I do - a great deal, but you have to know why and what you're doing. I'm working with every bit from 8-64. All bits are important when you know what you want to achieve and the effect of using more/less bits. :wink:

    But generally, if you want to preserve the [analogue?ish?] sound quality, keep it at 32-bit FP. As soon as you convert it even to 24-bit, you introduce digital artefacts. It might be bad, it might be good. Depends on what you want... :wink:

    Mastering works best if you get a 32-bit FP file.

    All the stuff I record live is 24-bit. Gets converted to 32-bit FP if I need to edit it. Which is most of the time. Converting back to 24-bit, or converting to 16-bit demands to use dithering to preserve the sound quality. For 24-bit, a simple TPDF dithering will suffice. Hell, I even use TPDF for 16-bit. :) But dithering is usually desirable.

    Cheers!
     
  12. junh1024

    junh1024 Platinum Record

    Joined:
    Jul 28, 2011
    Messages:
    1,119
    Likes Received:
    289
    1. Infinite headroom
    2. Gain staging is ancient
    3. Mastering engineers asks for a 32bit file
    4. Never have to pay attention to the meters
    ALL of these reasons (maybe more) are THE SAME.

    BTW, it's not " Infinite headroom" it's "almost Infinite headroom" because FP32 clips at 300? dB
     
  13. alexbart

    alexbart Newbie

    Joined:
    Mar 17, 2013
    Messages:
    5
    Likes Received:
    1
    To my experience it depends on how is your workflow, if you mix all inside DAW, then internal bit depth gives more headroom, not infinite of course, but for sure modern DAW's allows for more than 32 bit floating point. If your are talking about individual audio samples resolution, it's not that different, if you load a 24 bit sample into a 64 bit DAW then it will be processed at the DAW bit depth.
    It's absolutely not true that professional mastering services asks for 32 bit files, they are ok with 24 bit, they are ok with 16 bit too, what they usually asks for, is a grouped multi track project, for example the mixed all in one drum buss, the vocal buss and so on, so that they can do a specific treatment to each buss before the final mastering. A friend of mine, a popular mastering engineer said to me that it can happen that a mix done on lower quality equipment sounds better than one made on expensive gear. Even if we have more headroom, we have to take care to levels and mix carefully as it's done on analog consoles, usually you can see the difference between a professional mix engineer and a beginner one when it comes to levels, the pro one keeps the levels down on individual channels, the beginner one lower the volume on the master buss.
     
  14. Nana Banana

    Nana Banana Producer

    Joined:
    Oct 3, 2016
    Messages:
    163
    Likes Received:
    134
    Location:
    The 3rd Dimension
    Because it's vintage now! :rofl: (Sorry i couldn't help myself). Actually the only thing 32bit that ends up in my DAW is a bridged VST that never got ported to 64 :wink:
     
  15. Iggy

    Iggy Platinum Record

    Joined:
    Jun 21, 2011
    Messages:
    827
    Likes Received:
    280
    I've done a few things in 32-bit float, but I think you're confused as to what is actually happening when you do a project that way. All of your mixing and plugin processing is happening at 32-bit float (or 64-bit or whatever, depending on the plugin processing your audio and/or the DAW you're using) ... but you're neither recording AD to 32-bit float nor are you outputting from your converter (DA) at 32-bit float unless it is, in fact, a 32-bit converter. And unless you have the aforementioned 32-bit float AD/DA converter, you're also not hearing your audio at 32-bit float, you're just hearing it at plain ol' 24-bits (the rest is truncated by your converter or sound card before it hits your monitors or cans). That doesn't even cover downconverting for the actual master, since there is currently no streaming or consumer format that supports 32-bit float. While you sure get plenty of headroom at 32-bit float, it all manages to go away the moment you convert to 16-bit. And as I just pointed out, recording your mix to an external recording device means your 32-bit float mix is being truncated to 24-bit if it's being converted to analog before it hits your two-track recorder (there might be a DSD recorder out there that can accept a 32-bit float digital input, I guess), unless you have a 32-bit DA converter to output from. I guess what you gain by mixing in the box at 32-bit float can be a looooooong debate, but if it sounds good to you, mix at whatever. It's not actually hurting anybody -- it's not even hurting your mix.

    This, however, will definitely hurt your mix:

    Complete bullshit. I spent years mixing (and failing) with all my audio tracks normalized to -0 dB. I assumed you could do it this way, except that almost no plugin is designed to process 0 dB audio, especially compressors. I've never mixed hot, but my actual mixing levels, or even keeping an eye on my RMS (my mixes usually end up in the mid-Nineties' range of -14 dBFS RMS) didn't matter, because my individual tracks were all the way to -0.3 dB and my insert processors just made everything sound like shit. This guy has it all backwards -- your DAW may or may not be able to "handle it", but all those spiffy high-end processor plugs (and DAWs) were all designed with those "old analog guys" in mind, and therefore process audio at "old analog guy" levels. I also don't see that changing any time soon, either, as the "old analog guys" are the ones willing to pay thousands for Waves Complete or buying a $6000 PT HD bundle ... and all the young mixers are either using the "old analog guy" shit like multitrack tape recorders and homemade plate reverbs and thinks that anyone who records on a computer should be shot, or they're using whatever their heroes (the "old analog guys") are using and recording like their heroes are recording.
     
    Last edited: Sep 24, 2018
    • Like Like x 1
    • Interesting Interesting x 1
    • List
  16. The-RoBoT

    The-RoBoT Rock Star

    Joined:
    Jun 16, 2011
    Messages:
    819
    Likes Received:
    331
    Location:
    Plan 9
    This might help clarify some misunderstood things about the subject. :D
    https://www.steinberg.net/forums/viewtopic.php?t=126778

    "Any critical process is executed within double precision (64bit). That was so within the old engine, and is still so in the new.
    Plugins handling critical processes also upsample to double precision.
    Old VST2 standard was 32bit in- and out.
    VST 3 is 64bit in- and out.

    Making the engine full double precision eliminates the need for upsampling and truncating before and after each insert slot and from each channel to bus/bus/output. So since the plugins are (or can be) 64bit nowadays, there is no reason at all for "converting" before and after the insert slots.

    So what happens is that by having a 64 bit engine from start to end, a lot of unnecessary processing is removed from the audio engine.
    Does this make a difference in sound quality: Nope. Only under exotic laboratory conditions you would be able to expose the "gain in quality".
    It does simplify the piping/processing and programming throughout the audio engine?
    But indeed, the majority of people still thinks that "more is better", so even for that reason alone, the change is justified."

     
    Last edited: Sep 24, 2018
    • Like Like x 3
    • Interesting Interesting x 1
    • List
  17. Medrewb

    Medrewb Platinum Record

    Joined:
    Oct 11, 2014
    Messages:
    488
    Likes Received:
    189
    I think you should read carefully what the article meant to say. He put exceptions on analog modeled plugins too. For me, I am always above 0 in individual channel for years and just lower it in the master channel, and I never gain stage.
     
  18. m9cao

    m9cao Producer

    Joined:
    Oct 3, 2012
    Messages:
    378
    Likes Received:
    106
    i think yours have mislead the topic, we were already know about daw floating point processing, but it almost useless,

    if you have top end ad/da hardware, you could get more sound detail from ad, and da for playback usually has better quality than ad, and more protocol higher precision in hifi system, you can hear much much much more, so you can push the preamp/effect/synth parameters or even soft synth through loopback to max

    we should know more about quality, human ears doesnt has better quality but the recording port has, and more about instrumentation and sound design, most of things cant makes you to hear the difference of quality
     
    Last edited: Sep 24, 2018
    • Like Like x 1
    • Agree Agree x 1
    • List
  19. Iggy

    Iggy Platinum Record

    Joined:
    Jun 21, 2011
    Messages:
    827
    Likes Received:
    280
    It's not just analog-modeled plugs, though -- it's all plugins, and a vast majority of DAWs. Despite what you believe your computer or DAW can do, most of the digital audio stuff out there is designed with old farts and analog gear in mind, engineers and gear in which 0VU equals -24 dBFS. I'm not arguing that you can't max everything out, as I've done it for years. It just means that the resulting mix will most likely sound like warm shit. Hell, for every article out there espousing the virtues of not giving a fuck about gain staging, there's one that argues just as fervently that you should definitely give a fuck about it. Like this one.
     
  20. Medrewb

    Medrewb Platinum Record

    Joined:
    Oct 11, 2014
    Messages:
    488
    Likes Received:
    189
    I got your point but I think most of the vst plugins that does not emulate any gear does not need gain staging tho (but like you said except for those compressors and what-nots.) Take any stock eq for example, it doesn't matter how much gain you put into it or less, sounds the same.
    Maybe for newbies but, that guy makes his living tutoring and mixing and mixed lots of hits like Many Moore, S Club 7, etc.
    I personally do not find it misleading but thats me.
     
  21. Recoil

    Recoil Audiosexual

    Joined:
    Nov 9, 2017
    Messages:
    506
    Likes Received:
    548
     
Loading...
Loading...