Wav 32 24 16: the real differences?

Discussion in 'Working with Sound' started by Highdom, Jan 4, 2017.

  1. Highdom

    Highdom Kapellmeister

    Joined:
    Oct 16, 2016
    Messages:
    391
    Likes Received:
    47
    What are real differences processing a sample or track in 32-bit 24-bit 16-bit wav?
    What speaker you need to listen the differences? (if you can listen to)

    My question starts from a simple problem: Logic X doesn't read 32 bit files and i've got lot of sample in 32-bit. Can I convert them to 24-bit without losing ANY quality?

    Good converter 32 to 24 bit wav?
     
  2.  
  3. Best Answer
    The bit size determines the resolution of the audio image.

    There are three forms of 32 bit, float, fixed point integral and true longs. Only true longs actually have 32 bits of usable resolution. Most audio is done using IEEE 32 bit floating point format, which has only 24 bits of resolution...

    [​IMG]

    So, in general, good conversion from 32 to 24 causes no loss of resolution. People use 32 bit because it hides digital overs by exchanging the over for loss of resolution, by truncatiing the lowest order bits. Think scientific numbers taught in most schools. 1 is 10^0, 10 is 10^1, 100 is 10^2 and so on. So, 0.1234 is written 1.234 * 10^-1. 123,456.00 is 1.234 * 10*5, the problem being that the 5 and 6 have been lost.

    In 32 bit float there are 23 bits of resolution plus a sign bit, giving 24 bits. This gets multiplied by the exponent to handle large values (overs) and divided by the exponent to handle very small values. The multiplcation of the mantissa can 'roll-off' the low order bits of the mantissa causing exactly the kind of bit truncation distortion that dithering is designed to hide. If you go to +24dBFS over using 32 bit float, you're recording truncated to 20 bits with no dither, not 24.

    No audio hardware anywhere in this part of the universe can convert more than 21 bits, because -126dBFS is the noise floor of electrons passing through metals.

    Generally, you'll get better more consistent results by using 24 bit wave files, kept in the EBU range of -18dBFS = 0vu, which by a curious coincidence records on exactly 21 bits.

    Try Aiseesoft Audio Converter.
     
    • Like Like x 6
    • Interesting Interesting x 4
    • Useful Useful x 4
    • Winner Winner x 2
    • List
  4. peghead

    peghead Producer

    Joined:
    Jan 18, 2014
    Messages:
    226
    Likes Received:
    145
    WOW, what a fantastic reply THAT was, Fudsey. hat off to you!! :bow:
     
    • Agree Agree x 5
    • Like Like x 1
    • List
  5. ddoctor

    ddoctor Kapellmeister

    Joined:
    Jan 9, 2016
    Messages:
    229
    Likes Received:
    52
    So, by all means, can you hear the difference between 24 and 16 bit?
     
  6. muaB

    muaB Producer

    Joined:
    Nov 17, 2012
    Messages:
    313
    Likes Received:
    97
    i think its audible. because its the actual steps in the sampling of the soundwaves. more steps should lead to a more... natural image of the material. i think it also has something to do with deeper maths and physics to explain this.

    but i dont think its audible for you or me on a laptop with 200 dollar speakers/cheap output converters/ cheap cables / untreated room / untrained ears. hehe just go on and convert it, no one cares.

    and shame on logic for the 32 bit thing...

    I use db Poweramp! solid as hell batch converter and everything you need in converting regards.
     
  7. the_flying

    the_flying Member

    Joined:
    Sep 10, 2012
    Messages:
    36
    Likes Received:
    19
    Isn't the flow of electrons temperature dependent?
     
  8. realitybytez

    realitybytez Audiosexual

    Joined:
    May 29, 2013
    Messages:
    1,456
    Likes Received:
    637
    yes, so you will want to avoid recording at temperatures near absolute zero.
     
  9. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    2,683
    Likes Received:
    2,262
    Location:
    Russia
    There are some Adc with higher range, like -154 dB or so, don't remember, but I saw this on gearslutz, it is 32 bit ADC. And I wondered.
    About 32 bit float, Lukin or Chrisj from Airwindows or Bob Ohlsson or someone told that quality is a bit better than 24 bit, a couple of db or less. And I downloaded sine files in 24 bit, 32 bit fixed and 32 bit float, sonograms showed that digital noise in 32 bit float is a bit quieter than 24 bit, maybe it is so because of 24 bit file truncated or so. This noise in 32 bit fixed is much quieter because of 192 dB range. You can find them on src.infinitewave.ca
    Floating point is for to preserve audio from clipping, because of another dynamic range (1640dB?). So if not to clip, 32 bit float has less truncation than 24 bit.
    If you import 24 bit file, you must use 24 bit dithering because of 32 or 64 bit floating-point mixing,mastering,processing. Why dither? Because truncation is worse than dithering. Same with 16 bit.
    I work at possible maximum, 64 bit float, but all depends not on these bits, plugins have their own algorithms, bit depth etc.
     
    Last edited: Jan 4, 2017
  10. the_flying

    the_flying Member

    Joined:
    Sep 10, 2012
    Messages:
    36
    Likes Received:
    19
    Sounds COOL :winker:
     
  11. fiction

    fiction Audiosexual

    Joined:
    Jun 21, 2011
    Messages:
    1,889
    Likes Received:
    687
    That depends on the material and on the level management.
    16bit can go very far, but for music with very high dynamic range (i.e. both loud and very soft parts), the reduced resolution of 16bit can become audible at lower levels, at least if not dithered well.
    With 24bit you simply have more reserves - more headroom without clipping and/or more resolution at lower levels.
    For recording, I would always recommend 24bit, but e.g. for normalized samples, 16bit can often be enough without negative side effects.
    Also, using 24bit, the recording gain level setting is not as critical as with 16bit resolution. You just leave more headroom without soft signals drowning in quantization noise.
     
  12. muse2love

    muse2love Kapellmeister

    Joined:
    Nov 11, 2013
    Messages:
    162
    Likes Received:
    71
    Location:
    Montréal
    Great Post with great answers here...

    Being Aware of this can save your music for the better
    Be Aware

    And thanx all you have input righteously here... nothing to add on my part coz all tips were on the bull's eye !!!

    Stick to this predicament

    That being said... 32 bit float will make you aware and learn the real true notion of headroom
    Wich is a good thing... some can surf the big wave some can't...



    PeaceR
     
    • Interesting Interesting x 1
    • List
  13. G String

    G String Rock Star

    Joined:
    Jun 18, 2011
    Messages:
    974
    Likes Received:
    410
    Xrecode is my fave converter. Just convert 32 to 24 (after making a backup?).

    Simple answer: you won't lose much at all - nothing you will hear. Make some 16 bit conversions too and compare them? Stick with 24 if you can. Or 32. No big deal? Unless you are meticulous like folks above, you won't notice anything at all and it won't make any difference going from 32 to 24.
     
  14. Andrew

    Andrew AudioSEX Maestro Staff Member

    Joined:
    Oct 24, 2011
    Messages:
    1,981
    Likes Received:
    1,201
    Location:
    Between worlds
    Those steps you mentioned are not in fact there and you can't hear them as DAC reconstructs the waveform by interpolation and oversampling.

    32bit float makes sense during production. For playback, 16bit is more than enough to cover 99.999% of musical listening scenarios. Rigorous testing shows that out of 400 people, no-one could hear the difference without cranking up the volume during quiet parts and thus exposing the noise floor of 16bit.

    There are limited benefits from having 24bit "hires" recording, as you can simply dither it to 16bit any way you like (with or without noise shaping). But other than this geeky reason, nothing else really.

    As for sample libraries, 24bit could be beneficial as it gives more headroom to boost quiet parts, but in most cases it's wasted as the samples aren't normalized prior to bit reduction. Usually there's just less than 20bit worth of audio data.
     
    • Interesting Interesting x 1
    • List
  15. Highdom

    Highdom Kapellmeister

    Joined:
    Oct 16, 2016
    Messages:
    391
    Likes Received:
    47
    Yes man I don't find if i can or not notice differeces...but i'm going to send some tracks to a mixing and mastering studio...I would like to avoid the risk of making my records with bad-converted samples!
     
Loading...
Similar Threads - real differences Forum Date
Differences between Real Guitars and Ample Guitars? Software May 7, 2018
Should I really need a condenser mic? Mixing and Mastering Mar 17, 2024
Can you guys recommend the most realistic Spanish guitar vst? Kontakt Mar 2, 2024
RealHeadphones Software Feb 25, 2024
Free & realistic ITB acoustic drums Samplers, Synthesizers Feb 12, 2024
Loading...