Recording at 24bit appears to be 16 bits padded to 24bit

[eharmonica]

Mitiok posted about this issue in the Russian forum. Here's the gist:

When I record in Audacity at 32bit float or 24bit, a close examination of the wav-form shows it to be 16bits padded to 24bit. How do I (we) know? We use Cool Edit Pro to zoom in each sample, then zoom in on the vertical axis to get to single digits on that axis. According to Mitiok, the vertical axis shows 16bit values regardless of the wav rate. Therefore, anything that is 16bit, will have samples fall exactly on each value of the vertical axis. If the wav is 24bit, there will be samples that fall in-between the vertical axis values.

Leaving aside anything that has been amplified or had DC offset applied, those cases are harder to pin down. But just considering a "pure" sample with nothing done to it other than Save.

This is supposed to be a 24bit sample:

http://i153.photobucket.com/albums/s225 ... ndHand.jpg

look how all the values fall exactly on each line -- the 16bit values.

This si a "true" 24bit sample:

http://i153.photobucket.com/albums/s225 ... SAMPLE.jpg

Notice how the samples fall in-between the 16bit values.

We think that this is a bug in Audacity. Or are we missing something that is painfully obvious?

[kozikowski]

Sorta old news. I think the format modules got away from the developers. I created a modest sound file in one of the newer Audacity test builds that arrives as 16, 24, or 32 depending on which software packages opens it.

The Mac version has always had 24-bit troubles, and I wouldn't be shocked to learn that this may be part of the problem.

Koz

[eharmonica]

Sorta old news. I think the format modules got away from the developers. I created a modest sound file in one of the newer Audacity test builds that arrives as 16, 24, or 32 depending on which software packages opens it.

The Mac version has always had 24-bit troubles, and I wouldn't be shocked to learn that this may be part of the problem.

Koz

Thanks for your response.
"Sorta old news"? If so, Audacity should stop claiming 24bit recording.
Or are you saying that I should use different software to check on the bitrate?

Not using a Mac.

[steve]

When I record in Audacity at 32bit float or 24bit, a close examination of the wav-form shows it to be 16bits padded to 24bit. How do I (we) know? We use Cool Edit Pro to zoom in each sample, then zoom in on the vertical axis to get to single digits on that axis. According to Mitiok, the vertical axis shows 16bit values regardless of the wav rate. Therefore, anything that is 16bit, will have samples fall exactly on each value of the vertical axis. If the wav is 24bit, there will be samples that fall in-between the vertical axis values.

Here's an example that clearly shows samples that are not 16 bit values (vertical scale in dB):

32bit.png (13.54 KiB) Viewed 9159 times

But just considering a "pure" sample with nothing done to it other than Save.

If you save (Export) in a 16 bit format, then the audio will be resampled as 16 bit. The example in my screenshot was Exported as a 32 bit (float) WAV file, then imported back into audacity.

(By the way, I removed the image tags in your post as the pictures were too large for the forum layout and the "vertical scale" that you referred to was hidden)

[steve]

Just had another thought - perhaps the recording was made using a 16 bit sound card.

[thunderstick]

This is serious.

I'd like an answer as to whether Audacity is infact incapable of exporting to 24bit.

If so, this software is junk.

I have dozens of hours of recordings made with v1.3.3, and if all of my 24bit files (exported from 32bit-float) are infact 16bit, I'm going to be very, very angry.

[steve]

This is serious.

I'd like an answer as to whether Audacity is infact incapable of exporting to 24bit.

Read the posts - Audacity IS capable of exporting to 24bit / 32 bit.

If so, this software is junk.

Throw it away then.

[thunderstick]

Thanks for the confirmation.

EDIT: I'll withhold that confirmation until this matter is solved.

[eharmonica]

Thanks for responding, stf.

My soundcard is an Emu 0202 and is 24bit capable.
http://www.emu.com/support/files/storag ... USB-EN.pdf

I am recording 96kHz at 32bit float. I am saving at the same. When I open that wave in Cool Edit Pro and do as I describe above, the samples fall right on the 16bit values. Here is a link to samples taken from throughout my process:
https://download.yousendit.com/Y2o5ckhkdEN6NE94dnc9PQ

When I do the exact same process with a sample recorded with Cool Edit Pro or Spin It Again, the samples fall in-between the 16bit values.

If Audacity is 24bit capable, then we need an answer as to why CEP doesn't show it to be.

I will freely confess that I have no idea what your picture shows. Audacity thinks it's own 32bit float recording is 32bit float, maybe? Open that sample up in CEP and look at it. Or is there a better way to check on the true bit-depth of an unknown sample?

[kozikowski]

I'm tellin' ya. The export modules are scrambled. When the developers figure out my Audacity "Twilight Zone" sound file, we'll all laugh about this.

It is possible the Windows version is dropping the ball because of the odd way Audacity handles the differences between Capture, Import, Open and Export. Sometimes, those are seriously magic.

Koz