Is EQing a dithered file a no-no?

After dithering and noise shaping what I felt was the final version of a file, I'm finding I want a little more warmth in it, which requires a bit of EQ.  What happens if I modify that dithered file... do I get anything resembling even slight damage to the sound?  What if I change to 32-bit first?  Does modifying undo the work the dithering did?

If you can process a pre-recorded cd I don't see any reason why you can't eq it.

[Does modifying undo the work the dithering did?]

After dithering and noise shaping what I felt was the final version of a file, I'm finding I want a little more warmth in it, which requires a bit of EQ.  What happens if I modify that dithered file... do I get anything resembling even slight damage to the sound?  What if I change to 32-bit first?  Does modifying undo the work the dithering did?

Since you've noise shaped the dither, yes it does alter it somewhat. Strictly speaking, you should go back to the undithered mix to make the alterations, and then do the final 16-bit conversion with the shaped dither again. If you alter the EQ, you alter the shape of the dither.

Whether you would actually hear a difference to the dither performance is debatable, of course - but altering anything at all about a dithered version will alter the absolute amount of dither, and altering EQ will affect the noise shaping by the same amount in dB.

If you can process a pre-recorded cd I don't see any reason why you can't eq it.

I can't see the relevance of this to the question at all - it is self-evident that you can actually do it - but the implications of doing it are perhaps not so self-evident, and that's what the question is about.

Strictly speaking, you should go back to the undithered mix to make the alterations, and then do the final 16-bit conversion with the shaped dither again. If you alter the EQ, you alter the shape of the dither.

Obviously the rule would be to keep a copy of the file undithered... at least until I've lived with it long enough to know there will be no further changes.  But I'm also understanding that a single change to a dithered file may not be too traumatic a situation... certainly not enough to radically alter the noise placement accomplished in the shaping.

But I'm also understanding that a single change to a dithered file may not be too traumatic a situation... certainly not enough to radically alter the noise placement accomplished in the shaping.

It depends how big this change is, and where it's situated, I think. If you boosted everything at 16k by 20dB, for instance, all bets are off...

It depends how big this change is, and where it's situated, I think. If you boosted everything at 16k by 20dB, for instance, all bets are off...

Using 48K shaping I see a slight peak at 21K (dogs may get a slight buzz out of that, heh heh), and my change would be a .5dB increase with a Q of .5 at about 300Hz -- just to warm it up.  I'd think I'm safe, thanks to the reassurance in your earlier explanation.

I'd think I'm safe, thanks to the reassurance in your earlier explanation.

Hehe! I think you are too!

I can't see the relevance of this to the question at all

He's merely commenting that processing of already dithered material seems to work.

The first CD I had released was edited fairly heavily, using CE96.  As I recall, each edit step was dithered by the program, as the results were held in 16 bits each time - as a result some tracks were probably dithered six or seven times.  Some reviewers praised the recording (of a solo piano) for its delicacy and transparency; so I don't think that extra layers of dither need be the biggest worry!

Paul

so I don't think that extra layers of dither need be the biggest worry!

Multiple dithered saves could be a further extension of my question. Is it a worry at all?  What's happening to the math results if you dither and noise shape the same file several times?

What's happening to the math results if you dither and noise shape the same file several times?

Well, the theory says that they will algebraically add, so you will get a higher level of the same thing. But whether that holds good at the one-bit level is somewhat debatable - the only real thing to do is to try this on 16-bit silence and see what happens to the dither alone. This would be an easy experiment for anybody to carry out, but you'd probably need to do it several times on different dither noise profiles to get a clear picture of the effects.

Well, the theory says that they will algebraically add, so you will get a higher level of the same thing.

But because of the (pseudo-) random nature of dither, the addition will not in practice lead to the level of accumulation that a simplistic thought might lead one to expect.

Paul

I can't see the relevance of this to the question at all

He's merely commenting that processing of already dithered material seems to work.
Paul

Thanks pwhodges, Im glad you understand me even If I don't express myself properly sometimes.  

Thanks pwhodges, Im glad you understand me even If I don't express myself properly sometimes.  

It was a perfectly proper statement of the blindingly obvious...
 

But because of the (pseudo-) random nature of dither, the addition will not in practice lead to the level of accumulation that a simplistic thought might lead one to expect.

That's why I qualified the statement - I don't think that the normal dB addition formula applies here.