32 bit vs 16 bit

Discussion in 'Audio Hardware' started by Jason Smith, Jan 3, 2003.

Thread Status:
Not open for further replies.
  1. Jason Smith

    Jason Smith Senior Member Thread Starter

    Location:
    Chicago
    I know processing a wav file in 32 bit sounds better for things like reverb and declicking, but what about simple edits? For example, If I had a 16 bit wav file with 10 seconds of silence at the end and I wanted to trim the silence, would it degrade the sound to keep it at 16 bit? Would it just be a waste of time to convert to 32 bit for simple edits? Also, would doing simple fades (in or out) in 16 bit degrade the sound? Thanks.
     
  2. rontokyo

    rontokyo Senior Member

    Location:
    Tokyo, Japan
    Good question, Jason. Grant will respond after a while, but until then . . . . For simple edits like you mention I leave the file in 16 bit as it is rather time consuming to convert back and forth to 32. Besides, if you're just trimming silence it doesn't seem like you're applying anything to the wave file itself.
     
  3. Jamie Tate

    Jamie Tate New Member

    Location:
    Nashville
    Well, any processing, like eq, compression and volume changes are better the higher the bit rate. There's just more numbers to work with. Every function (especially a 16 bit file) produces more bits than were there originally. Think about dividing something, there's usually a remainder and if you do an edit at 16 bits those remainder numbers will just be chopped off instead of dithered down properly to 16 bits later.

    Edits are mini volume changes. What the computer does is make a small fade up and a small fade out and overlaps them so you hear it as seamless. 32bit is the way to go.
     
  4. Grant

    Grant Life is a rock, but the radio rolled me!

    Ron and Jamie are both correct.

    Personally, if it's just a small amount of editing, I just leave things at 16-bit. Processing at 32-bit is mandatory! Now, lately, I have been doing less important projects wholly in 16-bit, but I keep my processes to a minimum. But for serious stuff, the higher bit-depths are the way to go.

    One thing, if you must process an already dithered 16-bit file, do NOT use dither when saving again. It will add even more noise and dull the file very noticably. In this case, once the file has been dithered once it should not be dithered again. It may sound slightly brighter or grainer, but it way preferable to dithering twice. To save in Cool Edit without dither, you must first go to settings and disable it under the data tab. After saving, just remember to re-enable it afterwards so you don't screw up anything else you do.

    If you dither higher bit-bepths, noise shaping is an option to shift the dither noise around, but it's hard to configure to get just the right type for your file in question, and there are sonic tradeoffs, sometimes to the point of changing the toneality! As far as software goes, Cool Edit offers one of the best dither and noise shaping tools. Ozone plug-ins are the best for the PC that I know of. One thing I discovered, Sound Forge has the WORST dithering and noise shaping in the industry! Try them yourself!

    This is a reason it is best to make backups of any higher-bit depth files you have. In case something needs work after the fact, you can just go back to your high-bit files and make the adjustments there, and then produce a new 16-bit file to replace the old one.

    Dither. It is the necesary evil in digital. It is the main reason why Steve does his work in analog up to the last minute and then goes directly to 16-bit. 16-bit isn't bad, it's that nowadays most engineers use DAWs for all their work. There's nothing wrong with DAW of high bit-depths, they're just not the choice for everyone.
     
  5. sgraham

    sgraham New Member

    Location:
    Michigan
    Grant, a couple of questions.

    First off, do you have any specifically recommended dither settings for starting points at least? -- there are so many options!

    Second, why does dithering an already dithered file make it dull? Dither shouldn't affect the frequency balance. I would only expect it to get a tiny bit noisier.

    And if you take a 16-bit file, convert to 32 bit, edit and process then go back down to 16-bits without dithering, won't the edited pieces be truncated in a potentially ugly fashion?

    (And aren't virtually all recordings dithered one way or another anyhow?)
     
  6. lukpac

    lukpac Senior Member

    Location:
    Milwaukee, WI
    I suppose it might depend on your software, but if you're just doing simple edits (even in the middle of a song), I'd say you'll hurt things a lot less if you just stay in 16-bit. Any editing will only affect the samples just before and just after the edit. Nothing else will be changed. I don't know about you, but I don't really think those few samples are that big of a deal, as long as they line up and don't cause a click...

    Along those same lines, I'd say it's a bit silly to do de-clicking in 32-bit as well. Again, you're only modifying a handful of samples, and the result is just a guess anyway.

    Yes, things like noise reduction, volume changes, etc can take advantage of 32-bit, but if you're only working on short bursts (as you are with editing and de-clicking), it seems like a better idea to leave the rest of the file untouched by any conversions and work strictly in 16-bit.
     
  7. Grant

    Grant Life is a rock, but the radio rolled me!

    UOTE

    For stuff with a high frequency content, I tend to favor the C2 with a dither depth of .7 bits. It is reasonably quiet and really retains the crispiness and clarity of the highs. The drawback is that it dulls out the bass a bit. For stuff with low, or mellow high frequency content, I use the 44.1, 16-bit noise shape with a 1 bit dither depth. It retains that bass very nicely, but dulls or smears the transients a bit.

    I use a p.d.f. "Shaped Triangular". I once tried "Gaussian".:shake:Terrible!

    There isn't a lot written about noise shaping because there isn't much to tell. All it is is moving the dither noise around so it is either less noticable or doesn't ruin the music. If you use noise shaping you can get away with using less dither if the music will allow.

    If you add white noise to a band of frequencies, you will alter the perception of the sound and amplitude of those frequencies. That's how dither changes the sound. By not dithering a second time, you balance out truncation distortion with what's already been dithered so it should come out sounding fairly even. A fair tradeoff. The music may sound slightly more brittle anyway, but it will be preferable to the damage double dither will cause. Try your own experiments and see what you find. You may find a different way, or a new way. That's what's cool about all of this, there are really no hard, iron-clad rules. What i've presented are generally accepted truths found by experimentation and experience.


    Yup.

    No.
     
  8. Grant

    Grant Life is a rock, but the radio rolled me!

    Sometimes, no, frequently, I use a declicker tool on a whole file. Sometimes, it these cases, it may be benificial to convert to 32-bit first.

    You are correct in that doing very little spot declicking at 16-bit will not really produce noticable changes in sound. Same with edits. But if you do a lot of editing, your software may decide enough bits were added to require resaving the whole file, or worse yet, flush and save. Then you really will alter the sound
     
  9. lukpac

    lukpac Senior Member

    Location:
    Milwaukee, WI
    But the point is, the de-clicking is only taking place where there are clicks. The music itself isn't getting touched. Since the clicks aren't part of the music, and the recreation of the sound wave in those spots is nothing but an approximation anyway, it's pretty pointless to go to 32-bit.

    I guess it depends on how your software works. In ProTools, for example, all of your original audio files remain intact. If you do an edit, the software simply jumps from one spot in the file to another, without actually deleting the stuff in between (you can easily get it back later, for example). Thus, I could make 100 edits, and my original audio file wouldn't be changed. The only changes in output would be the couple of samples on either side of the edits.
     
  10. mne563

    mne563 Senior Member

    Location:
    DFW, Texas
    Grant- What causes Cool Edit to flush and re-save a file? I get this sometimes, usually it will ask me something like "Do you want to save changes" or whatever. Anyway to get around flushing the file?

    FWIW, I record almost everything in 32-bit. I try to avoid changing anything, but almost always will bring up the volume closer to zero. Forget eq for the most part. My rule is the less change of that file the better. I've made some awesome sounding vinyl transfers, even if there is an occasional pop or click.

    I remember hearing about a blind test where a digital music file was played, and then overlaid with (barley audible) pops and clicks from an old vinyl lp. The result was most people preferred the sound with the pops and clicks!

    Michael
     
  11. Grant

    Grant Life is a rock, but the radio rolled me!

    If you use a diclicking algorhythm, it touches all of the music scanning for what it considers to be clicks. The transients are affected. If you do this stuff, listen. It is virtually impossible to find a declicker that will not touch transients. Don't be fooled. They all do it. Even Pro Tools. Some are just better at it than others. This is why many of our fellow members will only manually declick.

    I guess it depends on how your software works. In ProTools, for example, all of your original audio files remain intact. If you do an edit, the software simply jumps from one spot in the file to another, without actually deleting the stuff in between (you can easily get it back later, for example). Thus, I could make 100 edits, and my original audio file wouldn't be changed. The only changes in output would be the couple of samples on either side of the edits. [/B][/QUOTE]Anytime you make *any* significant changes to a file you are adding bits. The software must get rid of those extra bits if it saves changes to 16-bit. That's why the sound of the file *will* change. To save, most quality software will, by default, dither the file to avoid distortion. If you do this to a 32-bit floating point file, there is no risk of changing the sound by extensive editing.
     
  12. Grant

    Grant Life is a rock, but the radio rolled me!

    They didn't test Michael or me! They could have tested a bunch of audiophiles biased toward vinyl.
     
  13. lukpac

    lukpac Senior Member

    Location:
    Milwaukee, WI
    Grant, the plug-in/program I use for de-clicking (which isn't ProTools, BTW) lets you listen to exactly what it's taking out. If you have it set incorrectly, yeah, it can affect transients. But if you use it correctly, it won't. When I hear it taking out a click once every few seconds, I'm not exactly worried about it hurting transients.

    So yeah - it *scans* everything, but it only *changes* what it sees as clicks. And if you set it correctly, you don't have anything to worry about.

    No, it won't. I can make 100 edits to a file, and the sound of it won't change at all except at the samples immediately surrounding the edits. I'm not sure what you're talking about by "adding bits" - any software that makes changes to samples other than what you tell it to (ie, on either side of the edit) is crap. If making an edit (or many edits) affects the file in places other than the edit, well, something is *seriously* wrong.
     
  14. mne563

    mne563 Senior Member

    Location:
    DFW, Texas
    I'm talking out of mostly ignorance here so forgive me, but I seem to remember that each time a digital wave is saved, the wordlength expands, and thus the sample is changed.

    Some good reading here

    Michael
     
  15. lukpac

    lukpac Senior Member

    Location:
    Milwaukee, WI
    That might be the case with some audio programs, but a good one shouldn't do that. I know I can save and re-save multiple times in ProTools and the file won't change at all. In fact, it won't do that in any of the programs I use.
     
  16. lukpac

    lukpac Senior Member

    Location:
    Milwaukee, WI
    Indeed. This is a helpful bit, in particular:

    How does this relate to the topic at hand? Well, with the programs I use, the null test passes, except of course at the exact points of edit and de-click. That is, if I take a file, de-click it with a plug-in, then perform the null test on the original and de-clicked files, the result will be total silence except in the particular spots where the clicks were (in those places you'll hear the clicks themselves). Same thing if I make 100 edits...

    Some programs may not work like this - I dont' know. But I *do* know that both ProTools and Spark on the Mac don't change things you don't want them to.
     
Thread Status:
Not open for further replies.

Share This Page

molar-endocrine