Skip to main content
Topic: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders? (Read 715 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

I have 32 bit audio in my audio editor. I want to encode it with nero aac encoder.

First I have to export it to a wav file, and then I can use that wav file as input for nero aac encoder. Because in nero aac encoder docs, it says that the input file must be in Microsoft WAV format and contain PCM data.

When I export it to a wav file, the audio editor gives me 3 choices: 32 bit wav, 24 bit wav, and 16 bit wav.

Question 1: if I don't do further editing to the wav file, and if I encode it to aac, then should I export 32 bit wav or 16 bit wav?

Question 2: an aac file encoded from 32 bit wav input, has higher quality than an aac file encoded from 16 bit wav input?

Question 3: an aac file encoded from 32 bit wav input, has the same quality as an aac file encoded from 16 bit wav input?

Question 4: the 32 to 16 bit conversion made by an audio editor (such as Audition, Audacity, SoX), has higher quality than the 32 to 16 bit conversion made by a lossy audio encoder (such as nero aac or lame mp3)?

Question 5: if input is 32 bit, there is one conversion to 16 bit made by the audio encoder. If input is 16 bit, there are two conversions to 16 bit, one made by the audio editor, and another one made by the audio encoder. Is one conversion to 16 bit better than two conversions to 16 bit? One conversion to 16 bit means less quality loss than two conversions to 16 bit? Or two conversions to 16 bit result in the same quality as one conversion to 16 bit?

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #1
1. If you feed a encoder >16 bit (given the encoder in question can handle it), then you are trusting the encoder built in resampler to do the job. I'd rather use a dedicated resampler/ downmixer like SoX than using the encoder built in one. Will you hear any difference? Probably not. But to me it makes sense to use dedicated tools where possible.

2. Source bit depth alone should not affect the output result. Unless it's less than 16 bit off course.

3. Maybe, if the encoder built in downsampler is worse than a dedicated one, like SoX. But probably not.

4. Probably, maybe. Can you hear the difference? Probably not.

5. AFAIK, "conversion to 16 bit" only happens once in that theoretical chain.

To get true answers to these questions, the best thing to do is ABX blind testing. Foobar2000 + "ABX Comparator" is great for that: https://www.foobar2000.org/components/view/foo_abx.
Whenver you are unsure about how something will affect output quality: ABX it.

(I might not be correct on every question here, long time since I played with audio conversion. I got lazy after Spotify came I guess.)

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #2
Question 4: the 32 to 16 bit conversion made by an audio editor (such as Audition, Audacity, SoX), has higher quality than the 32 to 16 bit conversion made by a lossy audio encoder (such as nero aac or lame mp3)?
If encoder can accept 32 bit as input, it doesn't convert to 16 bit. Lossy-encoded  files have no fixed bit depth.

If you feed a encoder >16 bit (given the encoder in question can handle it), then you are trusting the encoder built in resampler to do the job
What resampler has to do with bit depth?

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #3
Question 4: the 32 to 16 bit conversion made by an audio editor (such as Audition, Audacity, SoX), has higher quality than the 32 to 16 bit conversion made by a lossy audio encoder (such as nero aac or lame mp3)?
If encoder can accept 32 bit as input, it doesn't convert to 16 bit. Lossy-encoded  files have no fixed bit depth.

If you feed a encoder >16 bit (given the encoder in question can handle it), then you are trusting the encoder built in resampler to do the job
What resampler has to do with bit depth?

Yep, thought I'd probably not be 100% correct here, not on terminology either. What's the correct term for such tools? Downsamplers?

The ABX part still stands though: If you can't hear a difference, then it's just up to preference and ease.

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #4
As editing digital audio is doing all kind of calculations, 32 bit samples are used by audio editors to avoid quantization errors.
It has nothing to do with the actual dynamic range.
If you have a very quit recording chain you might capture 20/21 bits of resolution.
Don’t want to start another highres debate but in general 16 bits will do for the final product.
TheWellTemperedComputer.com

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #5
For what it's worth my understadning is that the BBC uses a 24/48 input to its AAC coders. I have wondered whether the output would be the same as the same file taking he file downsampling it to 16/44 and before inputting into the codec and if not why not.

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #6
For what it's worth my understadning is that the BBC uses a 24/48 input to its AAC coders. I have wondered whether the output would be the same as the same file taking he file downsampling it to 16/44 and before inputting into the codec and if not why not.

It won't be the same since the sampling rate is different.  It won't be much different though.

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #7
I still don't know what to choose. So the audio in the audio editor is 32 bit. And nero aac encoder accepts 32 bit wav as input.

So I have 2 choices:

Choice 1: from audio editor, export 16 bit wav, then encode with nero aac encoder.

Choice 2: from audio editor, export 32 bit wav, then encode with nero aac encoder.

If I make choice 1, I loose quality. Because the 32 to 16 bit conversion causes rounding errors (quantisation errors).

If I make choice 2, I don't know if I loose quality.

So I have to ask: I know aac doesn't have a fixed bitdepth, but if nero aac encoder receives 32 bit wav as input, before it starts creating the lossy aac file, does it convert the 32 bit wav to a 16 bit wav on the fly?

I'm trying to avoid loosing quality, where possible,

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #8
So I have to ask: I know aac doesn't have a fixed bitdepth, but if nero aac encoder receives 32 bit wav as input, before it starts creating the lossy aac file, does it convert the 32 bit wav to a 16 bit wav on the fly?

Neither, internally it is all converted to floating point.  Since files with even 16 bit dynamic range are rare, I doubt it makes any difference what you pick. 

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #9
I also doubt you'll hear any difference and you've already decided to go lossy so I wouldn't loose sleep over this...

Quote
I have 32 bit audio in my audio editor. I want to encode it with nero aac encoder.
Your audio editor can't export to AAC?

Quote
before it starts creating the lossy aac file, does it convert the 32 bit wav to a 16 bit wav on the fly?
I don't know, but try this:

Reduce the volume by 100dB.  (You can't go below -96dB with 16-bits).
Export to 32-bits and then convert to AAC.
Re-open the file and amplify by 100 to see if there's anything there.    (If you get your audio back it's not getting converted to 16-bits.  (If you get silence it doesn't "prove" it's getting converted to 16 bits because the lossy compression may be throwing-away those "tiny details" anyway.)

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #10
I don't know, but try this:
I tried it. I got the audio back. And there was no noise after amplification.

If nero aac encoder converted the input from 32 to 16 bit, I couldn't get the audio back. And the only thing that would amplify, is the noise (which is probably dithering to randomize quantization error).

Therefore, I conclude that: aac file encoded from 32 bit wav file, is higher quality than aac file encoded from 16 bit wav file, because it doesn't have that noise. I know I can't hear that noise, but I want to avoid adding it anyway.

Re: Is 32 bit wav better than 16 bit wav, as input for lossy audio encoders?

Reply #11
Your audio editor can't export to AAC?
It can. But based on the listening tests, nero and apple aac encoders could be better than the aac encoder in the audio editor.

 
SimplePortal 1.0.0 RC1 © 2008-2020