I've often read that removing DC Offset from an audio file ensures the best dynamic range, etc., when processing. I'm also aware that converting a file to 32-bit floating-point format results in better processing. Does the order of these two operations affect the results? In other words, should I:

1. convert a file to 32-bit format before removing DCOffset?
2. remove the offset before converting, or
3. does it matter?

I'm currently using Plogue Bidule to process audio files. One of the program's authors has confirmed to me that Bidule's Audio File Player automatically converts a 16-bit file to 32-bit floating-point format and simply passes through data from a 32-bit file.

Before loading a file, I always ensure that I remove DCOffset. Would my source file be better if I first converted it to 32-bit format using some utility and then removed DC Offset before loading it into Bidule?

Thanks for your help,
Maes

To be honest I really don't know how those features work but have in mind that when your source file comes from CD you never can get better one, that's my suggestion. I'm not into those editings but this is what I actually have in mind...

Hi and thanks for the reply. Actually many source files copied from CD, even using Exact Audio Copy, have DCOffset. These are not just ones I've ripped but lossless files from Usenet that were ripped using EAC as well. I'm wondering if the offset was present in the digital data on the original CD or if it was introduced by the ripping process.

At any rate, since converting to 32 bits for any serious audio processing is usually recommended, I'm wondering if removing the offset before or after the conversion makes any difference.

AFAIK ripping through EAC secure mode is lossless so there shouldn't be any additional offset introduced by ripping...
So the offset had to be present in the digital data on the original CD...



That's not a question for me... Sorry!

If you have 32bit float capability and want to edit your audio file in any way (remove DC offset, for example):
The very first step should be converting up to 32bit float, the very last step converting again down to 16bit (if needed; dither may be sensible) So choice 1. is best.
This is because 32bit (float or integer) provides many additional possible sample values between the integer sample values in 16bit, better resolution. Every editing process causes less rounding errors. But I doubt the difference is audible at all if you just preform only DC offset adjustment.
If I were you, I wouldn't care about DC offset that much. Even on old CDs with much DC offset, it doesn't affect the sound, since direct current contains no wave information. The worst thing would be two loud pops at the offset's beginning and end.

One of the main reasons to have to remove the DC offset is if your recording equipment would introduced some DC offset during the recording process. Correcting this will make your wav more symettrical. It doesn't really affect the sound if clipping isn't introduced through a DC offset, but I'd recommend if the audio clips because of it, definitely adjust the DC offset.

That is a good reason, of course. But: if the audio clips because of DC offset, it's already too late. The correct thing would be
-A/D conversion into a high resolution format during recording
-DC offset adjustment
-normalizing up to -0,001dB
-converting down to final format with dither
This method uses the maximum dynamics of the final format.

Edit: leave plenty of headroom during recording, at least 20dB

I'm somewhat familiar with Adobe Audition(CEP) and I belive there's a way to adjust your recording signal to a correct DC offset while recording, but I could be wrong.

Yes, that option exists. But how does CEP do that ? In order to be different, or better than the thing I described, the DC adjustment needs to be done analogically, before the actual recording.
I think that option makes it just more simple for non-experienced users.