Samples 16/44.1 - Project 24/88.2

Discussion in 'Mixing and Mastering' started by ZUK, Feb 23, 2017.

  1. ZUK

    ZUK Rock Star

    Joined:
    Aug 24, 2011
    Messages:
    543
    Likes Received:
    330
    Hi,

    What would be the way to use 16 / 44.1 samples if I work at 24 / 88.2.?
    Should I convert frequencies? bits? both? it is necessary? Best work at 32/88.2 ?

    I say 88.2 because I prefer to work at 88.2 against the 96 for final mix which will finally be 16 / 44.1.

    Explanations, clarifications and recommendations are welcome.:yes:
     
    Last edited: Feb 23, 2017
  2.  
  3. MMJ2017

    MMJ2017 Audiosexual

    Joined:
    Mar 12, 2017
    Messages:
    3,538
    Likes Received:
    1,688
    create your project in 24 88.2 then import the files into your project at end when finished use dithering for 88.2 24bit
     
  4. Bunford

    Bunford Audiosexual

    Joined:
    Jan 17, 2012
    Messages:
    2,170
    Likes Received:
    837
    It doesn't really matter if they are the only samples you have, then just use them in your project and don't worry about it. Sure, get the better quality samples if they are available as they will be higher fidelity and will improve your track quality, but if they are not then just use what you have. Also, keep in mind most music is still released on CDs as the standard format, which is 16 bit and 44.1kHz. Beatport AIFF formats are also 16 bit and 44.1kHz, so an obsession with higher quality is not always that useful. Sure, it may 'future proof' your track for when higher bits and kHz become the norm, but by then your tracks will probably sound outdated (not to be rude) and the higher format will only be useful for archiving. Personally I stick to 24 bit and 48kHz as a balance between quality and sensible CPU and SSD usage.

    Essentially, as soon as a sound file has been encoded in a certain format it is impossible to upgrade it up, for example from 16 bit and 44.1kHz to 24 bit and 88.2kHz. Sure, you can get software that will "do it" and give you the readings that look like it has, but it hasn't really and it's all just smoke and mirrors. As you would be upgrading the resolution, and the sound data is simply non-existent in the original file for the extra resolution blocks, some software will simply have the identical file in the exact same quality but 'stamp' the format type as 24 bit and 88.2kHz (even though it's not really as it will still be 16 and 44.1 quality) or some encoders will attempt to 'fill' the extra gaps the higher resolution has by essentially 'guessing' what the sound is using their algorithm. This will likely do nothing to improve the quality and probably just degrade it further and the 'filling' will likely cause clicks or pops in your sample and make it worse.

    If it's all you have, don't worry about it and use them.

    Also, you should NEVER use dithering unless you are downgrading audio format. For example, you would use highest quality dither if you were exporting your 24 bit and 88.1Hz project to 16 bit and 44.1Hz, but you should never really be using dithering in any other sense, and definitely not as a way to try and 'upgrade' 16 bit and 44.1kHz samples to 24 bit and 88.2kHz ones.

    The above is the info I have for you, though there will bound to be those who disagree with this being the internet! :rofl:
     
    • Agree Agree x 2
    • Like Like x 1
    • List
  5. MMJ2017

    MMJ2017 Audiosexual

    Joined:
    Mar 12, 2017
    Messages:
    3,538
    Likes Received:
    1,688
    i agree with bunford above
     
  6. taskforce

    taskforce Audiosexual

    Joined:
    Jan 27, 2016
    Messages:
    1,957
    Likes Received:
    2,066
    Location:
    Studio 54
    While everything you say it is correct, you omit one certain and very usual situation. The reason to up-sample audio files is because any plugins you will use afterwards will, most likely, benefit from the upsampled quality as most plugins work natively at least at 24bit and many on 32bit floating point as well. Meaning that the quality when you put a 24bit plug on a 16bit file and the same process on the same audio but upsampled to 24 or 32bit fp (depending what plugins you use), will sound better. It is not that the bits when you resample higher an audio are lost. Zooming up to sample point you will see that all the bits are there and again, there are no gaps, but in reality most of times it's adjacent samples copied over.
    Hypothetically speaking, if the upsample algorithm could guess and fill in the digital "holes" correctly with the extra bits of info and not just "binary" info, the difference of the quality would be quite significant. In reality because the upsampled product is a "pseudo" 24 or 32 bit file,as correctly Bunford described, it won't sound as good as a real recorded 24bit file but exactly the same as the 16bit , but can benefit from a post process that requires/supports a higher than 16bit resolution. Such as applying plugins on an audio file and exporting -whatever time you choose- again to 24 or 32bit fp.
     
    Last edited: Mar 13, 2017
    • Disagree Disagree x 1
    • Interesting Interesting x 1
    • List
  7. junh1024

    junh1024 Rock Star

    Joined:
    Jul 28, 2011
    Messages:
    1,396
    Likes Received:
    432
    No upscampling algo I know of smooths out the jaggies in 16 to make a smoooth 24 bit, apart from perhaps, denoisers, which ofc make it sound more shit.


    ALSO, dither is not needed. EVen when going from 24->16,
    http://wiki.hydrogenaud.io/index.php?title=Dither

    I've done a few listening tests myself, I can't tell. In fact, you can go as low as 14bit, put on a few heavy FX without dithering, and you won't notice, if you're using sensible levels.

    You can't hear above ~20k, so higher rates than ~48k are useless.

    Higher rates may be harmful, https://www.gearslutz.com/board/mas...houghts-high-resolution-audio-processing.html

    If you're smashing your dynamics , and you want to avoid inter sample clipping by using a higher rate (the ONLY possible benefit), you can do that at lower rates simply by setting a lower ceiling/threshold.

    some VSTs may have bugs at >44k, and alot of VSTs work best at 44k.

    My DAW works in at least 32bit (mixing) (but your VSTs may not), and I work in 48/44k. I export at 16bit.
     
    • Interesting Interesting x 1
    • List
  8. MMJ2017

    MMJ2017 Audiosexual

    Joined:
    Mar 12, 2017
    Messages:
    3,538
    Likes Received:
    1,688
    "You can't hear above ~20k, so higher rates than ~48k are useless."


    even if you cannot hear it it impacts the harmonic structure and transients below it

    think about it a lot of people cannot hear the fine grain detail difference between a pro level amazing mix and a mediocre one they trying to get right could be ears could be room could be playback system could be lot of things but no-one that has anything to do with the fine detail that actually exists and the actual difference between mediocre mix and great one.


    that said nothing wrong with working in 44.1 24bit the entire process if that is going to be your delivery format

    as far as dithering
    http://wavefrontmastering.com/sound-basics/dithering-format-faq/



    "Higher rates may be harmful, https://www.gearslutz.com/board/mas...houghts-high-resolution-audio-processing.html"


    a thread with a dude talking about sine waves with a specific context and specific converters is not exactly compelling


    it seems bizarre that you think chopping off bits is harmless and fine but chopping of sample rate with sine wave bleeps is dangerous? you know that bits are the most important thing right?


    "If you're smashing your dynamics , and you want to avoid inter sample clipping by using a higher rate (the ONLY possible benefit), you can do that at lower rates simply by setting a lower ceiling/threshold."


    we all have to smash the dynamics it is mandatory as no human being alive has final mixes at -20db rms are you suggesting that you yourself do not smash your dynamics therefore have no need for the higher sampling rate? ( that is harmful to us?)
     
    Last edited: Mar 13, 2017
    • Like Like x 1
    • Disagree Disagree x 1
    • List
  9. MMJ2017

    MMJ2017 Audiosexual

    Joined:
    Mar 12, 2017
    Messages:
    3,538
    Likes Received:
    1,688
    "No upscampling algo I know of smooths out the jaggies in 16 to make a smoooth 24 bit, apart from perhaps, denoisers, which ofc make it sound more shit."


    thats right think of a tiny picture you try to blowup to poster size, although if your going to "rebuild" the image in photoshop you could make a high res version based n the concept of the tiny one but that is not what he was suggesting.
     
  10. Iggy

    Iggy Rock Star

    Joined:
    Jun 21, 2011
    Messages:
    1,092
    Likes Received:
    434
    Location:
    The stage, man
    You're talking about using a sample library in Kontakt (or similar sampler) that's 16 bit/44.1 kHz in a 24-bit/88.2 kHz DAW project? There shouldn't be any problems. If you bounce the sampler's audio track, you're still recording at 24-bit/88.2 kHz. You can't improve the fidelity of your samples by upconverting them, and as VIs like Kontakt, old-school EZDrummer and SampleTank use 16-bit/44.1 kHz samples all the time in projects with higher sample rates and bit depths with no real noticeable problems.
     
  11. taskforce

    taskforce Audiosexual

    Joined:
    Jan 27, 2016
    Messages:
    1,957
    Likes Received:
    2,066
    Location:
    Studio 54
    @junh1024
    You do what your ears imply, and you are justified to do so. To my ears a song on Blue-ray (24bit) sounds better than the same song on CD (16bit). Especially when it includes a lot of real instruments, like jazz or classical. SACD is even better. Surely this represents just a few machines and listeners but they are there. Most know digital conversion to analog sound recreation has noise artifacts. At 44.1 khz, noise starts dispersing @ 22.05 Khz which theoretically is too close to the human hearing and in action also reported to produce harmonic artifacts to audible frequencies, combined with 16bit, usually is identified by the listeners as "harshness".
    On the other hand, the Nyquist-Shannon theorem states "upsampling to a sampling rate which is a multiple of the signal's sampling rate would allow for a perfect reconstruction of the signal". But although this is the rule,yes, upscaling in reality may cause odd order harmonics to appear, interpolation noise and in fact may ruin a file, because of not perfect algorithms. So you are quite correct although to my defense in near perfect conversion or re-recording to 24bit sensitive audio material could be further enhanced easier or more efficiently.
     
  12. Andrew

    Andrew AudioSEX Maestro Staff Member

    Joined:
    Oct 24, 2011
    Messages:
    1,981
    Likes Received:
    1,201
    Location:
    Between worlds
    If you're going to work exclusively with 44.1/16 samples, there's no reason to bump the sample rate to 88.2kHz.
    Set your project to 44.1kHz and either 24bit or floating-point and that's it.

    24bit cannot sound "better", the only difference between that and 16bit is the noise floor.
    Over 550 tests were conducted in AES test in 2007 which concluded that the only way to differentiate between 16bit and 24bit in final product is to crank the volume much higher during quiet parts of the songs (past safe limits).
    http://www.aes.org/e-lib/browse.cfm?elib=14195

    Source: http://nwavguy.blogspot.com/2011/05/subjective-vs-objective-debate.html
     
    • Like Like x 2
    • Agree Agree x 2
    • List
  13. SineWave

    SineWave Audiosexual

    Joined:
    Sep 4, 2011
    Messages:
    4,273
    Likes Received:
    3,371
    Location:
    Where the sun doesn't shine.
    We've been working with professional digital audio for more than 20 years and people still believe a lot of nonsense about it. :wink:

    Benefit of upsampling is none. Except what Taskforce said that processing a higher bit file will produce slightly better sounding result when processed through plugins. One other benefit of upsampling is that you can get better sounding samples if you upsample them to the frequency and 24bits before using them in your DAW because your DAW might be using crappy sample rate conversion. Lots of DAWs and editors have not the best SRC. Use freeware Voxengo R8Brain for sample rate conversion without any fuss and best results.

    Regarding dithering, if you're satisfied with how your track sounds without it, don't use it. However, if you care about the quality of sound [2.5 people these days and I'm one of them] always use dither even when going from 32 to 24 bits. In that case regular TPDF dither will suffice. Actually it is better to use TPDF dither when converting from 24bit to 16bit than not using any dither. People who know how undithered audio sounds will never skip dithering, trust me. :wink: But yes - upsampling with dithering is nonsense.

    Cheers!
     
    Last edited: Mar 13, 2017
    • Like Like x 2
    • Interesting Interesting x 1
    • List
  14. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    2,683
    Likes Received:
    2,262
    Location:
    Russia
    There is nothing wrong with 96 kHz for final 44.1
    It is 2017, and all the problems are gone with src algos. Create sine, convert it as you want in rx. For example from 44.1 to 96 to 48 to 44.1 to 192 etc. Only you will get is super fast inaudible pre and post ringing at 22.05, you will not be able to hear it even if you use steepest filters with linear phase for cutting at 16-18 kHz.
    So aliasing in some src is below -175db, so it is enough for 32 bit int files, but for 24 bit converted your lowest point will be about -144 db or so..
    What I do is arranging and choosing samples for project first, then if they fit for project, I convert them offline with ultra quality src, instead of using shitty ones from DAWs, because top quality src are much better than in daws. Then replace original ones with converted ones. If your source sample is already peaking at 0dB, gain it down a bit, then convert. Or use 32 bit float to prevent clipping. I always use 32 bit float or fixed, but when you render project as 24 bit or/and 16 bit , you must use dither as 24 bit dithering or 16 bit dithering respectively. With or without noise shaping.
    Don't dither for converting to 32 bit float or fixed, it will not solve anything, and dithering for float doesn't exist.
    And you should know, dithering with noise shaping for each sample rate is different. Rx as example.

    And, Truncation is always worse than dithering. Read Bob Ohlsson posts about this on gearslutz.
    Work as 32 bit float/int or higher..

    For final render formats dithering is always after conversion.
    If you do vice versa it will be like washing legs before a step on shit

    The best src are finalcd with sharp mode, rx, sox, dbpoweramp, as well as not so excellent but very good are voxengo r8brain pro or free, Weiss saracon and some others
    Go to src.infinitewave.ca
    Dithering are mbit by izotope, Airwindows free ones, Toneboosters tb dither (make your own dither as you want), psp x-dither, and some others.
     
    Last edited: Mar 13, 2017
    • Like Like x 2
    • Agree Agree x 1
    • List
  15. MMJ2017

    MMJ2017 Audiosexual

    Joined:
    Mar 12, 2017
    Messages:
    3,538
    Likes Received:
    1,688
    good points people
     
  16. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    2,683
    Likes Received:
    2,262
    Location:
    Russia
    Dithering is better than truncation. Noise shaping masks dither. So, it's too lazy to add dither, because nobody pays attention?
    So, maybe fuck all soundcards? 16 bit is enough to digitize? Mp3 samples or rendered tracking in mp3 maybe enough for you, because nobody hears difference between wav? Maybe you use old synthedit synths, because nobody hears difference of good emulation or hardware moog?
    It is so easy to add dithering / noiseshaping because it is better. Truncation is not only noise, it's harmonical distortion, and sounds like "sshrshchsrrshsr s sr sh tr sh", your quietest places are cut, silent...and truncation is louder than tape-like dither "shshshshshs" and much louder than ultrasonically noiseshaped "sssssssss". And with dithering you preserve more quiet passages, than without.
    Try to make 8-bit dithered, noiseshaped and truncated tracks, try to make them quieter and you will see. It is simple. From reading true professional info (not blogs made by 15 y.o. teen amateurs, or guys without monitoring and acoustic treatment) to making your own tests.
    Watch long videos by Airwindows who made some dithering plugins, read Bob Ohlsson on gearslutz. Do your own tests.
    Not just "this guy said fuck this because he doesn't know what it is about, so I say fuck this too".
    Don't know why do you work at 32 etc, then render undithered 16. So you avoiding truncation, but in final render you add this.
    For mp3 files dithering does not work properly, works but too much dirtier. It is better to render mp3 directly from 44.1 / 24(32), but dithering does something good with quiet passages, but worse than in wav. It is so because lossy conversions kill dithering noise and it becomes just noise, but preserves some places. So you decide, truncation in mp3, or truncation with noise and some preserved quiet passages.
     
    Last edited: Mar 13, 2017
  17. Adamdog

    Adamdog Platinum Record

    Joined:
    Jan 18, 2016
    Messages:
    389
    Likes Received:
    204
    Location:
    Saturn
    this is very interesting. usually I record and mix small projects without samples at 88.2 kHz, or I convert 88.2 kHz takes to mix at 44.1 kHz
    RME ADI 8 DS converters

    if I got it well, reading the Gearslutz thread, in both cases I have the 4 kHz harmonic issue that is audible
    we can t hear 24 kHz harmonics but 4 kHz... is there up in your face

    so, what to do? all at 44.1 kHz? 48?
    why do they sell higher sample rates converters?
     
  18. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    2,683
    Likes Received:
    2,262
    Location:
    Russia
    Internally convert your files from low sample rate to higher.
     
  19. mild pump milk

    mild pump milk Russian Milk Drunkard

    Joined:
    Mar 29, 2014
    Messages:
    2,683
    Likes Received:
    2,262
    Location:
    Russia
    And guys, for final master- src first to 44.1, then limiter with ceiling, then dither.
    Src after limiter may cause peaks overs, true peak overs and may cause more clipping. Even with linear phase src filter
     
  20. SineWave

    SineWave Audiosexual

    Joined:
    Sep 4, 2011
    Messages:
    4,273
    Likes Received:
    3,371
    Location:
    Where the sun doesn't shine.
    :rofl: Great analogy! :hahaha:
     
  21. ZUK

    ZUK Rock Star

    Joined:
    Aug 24, 2011
    Messages:
    543
    Likes Received:
    330
    In the end it has become interesting. :mates:
     
Loading...
Loading...