Setting track bit depth with Edit>Preferences and Track Control Panel > Format

[pato]

Using v3.2.5:
– In Edit>Preferences>Quality I set default sample format to 24 bit but new tracks always come up as 32-bit float. Why?
– For an existing track, Track Control Panel>Format allows setting bit depth to 16, 24, or 32-bit float. If a track is created with bit depth 32-bit float and recording has been done in it, can the bit depth of recorded material in that track be changed to 24-bit simply by changing the track’s Control Panel>Format setting? Risks/issues?

[DVDdoug]

For “technical reasons” Audacity uses floating-point internally. Almost all audio software works in floating point.

If you import a 24-bit file it will be converted to floating point. If you later export the file as 24-bits it’s a lossless process. (Unless you have dither enabled. Dither adds a tiny amount of noise.)

You can check files outside of Audacity with MediaInfoOnline.

[LWinterberg]

– In Edit>Preferences>Quality I set default sample format to 24 bit but new tracks always come up as 32-bit float. Why?

According to the table in Importing Audio - Audacity Manual when you import a 24-bit track with a 24-bit preference, it uses 32-bit instead. I don’t understand why that is either, or if it’s accurate.

If a track is created with bit depth 32-bit float and recording has been done in it, can the bit depth of recorded material in that track be changed to 24-bit simply by changing the track’s Control Panel>Format setting? Risks/issues?

any additional processing you do to the track will be much slower, as for each edit step Audacity would do 24-to-32.conversion → apply effect → 32-to-24-conversion instead of just applying the effect

[steve]

@LWinterberg

According to the table in Importing Audio - Audacity Manual when you import a 24-bit track with a 24-bit preference, it uses 32-bit instead. I don’t understand why that is either, or if it’s accurate.

I believe that is accurate. I (vaguely) recall a discussion about this many years ago (probably a decade ago). If I recall correctly, this was an intentional design decision by Michael Chinen. I think his idea was based on the fact that 32-bit float should normally be preferred to achieve best sound quality, but to make an exception for cases where disk space is an overriding concern (for example when working with extremely long recordings).

@LWinterberg

any additional processing you do to the track will be much slower, as for each edit step Audacity would do 24-to-32.conversion → apply effect → 32-to-24-conversion instead of just applying the effect

Also, because 32 to 24 bit conversion is subject to quantization errors, the audio is dithered during the conversion, which adds noise to the track. Although the added noise is at an extremely low level, it is avoided completely by not converting back to 24-bit, which is one of the big advantages of working with 32-bit float tracks.
Another disadvantage of using 16 or 24 bit tracks is that they are integer formats, which means that the audio data is permanently clipped at 0 dB.