What is happening when I change a file's accuracy?

I have a lot of recordings in 32bit-float 44.1. I need to get some of these to the 16/44.1 standard. I don’t need to process any of the files, only convert them. I need to maintain as much fidelity as possible.

I used the “accuracy” function to convert a test file. Was this the best process to use if my goal is to maintain as much fidelity as possible? What is happening when I change a file’s accuracy?

When you use that function, the result file is the same as if you had rendered to 44.1KHz 16 bit without dither. So it wouldn’t be recommended, and a dithered file would generally be considered better quality, and the proper way to do it.

The manual has curious wording for this function, which might make somebody think that the result would possibly be higher quality.

“Changing the Audio Properties
You can change the declared sample rate and sample accuracy of audio files. Changing these values does not process the audio file in any way (in contrast to using Save as).”

But a “Save as” or a render forced to 16 bit would give you exactly the same result as this function (which can be confirmed with the File Comparator), so there’s no advantage converting to 16 bit with it, and the disadvantage is there’s no dither used.

The term “accuracy” has always seemed strange to me in the way it’s used within WaveLab.

As Bob said, I would certainly dither if you are changing a file from 32-bit float to 16-bit.

As far as what it’s doing, it would appear it’s just truncating from 32 float to 16 bit fixed: removing everything below the 16th bit (removing everything below -96db), and flattening any peaks you have above 0db down to 0db. But it’s leaving the area between 0 and -96db exactly the same, so the 32 and 16 bit files are identical from 0 to -96db, but that’s normal in any non-dither digital copy from 32f to 16. so the accuracy isn’t any special process, from what I can tell. The results are the same as doing the same thing any other way in Wavelab.

“Accuracy” is used here to mean what I’d call “precision”, or simply “bit depth”.