"David Kakon" wrote in message
om
Hi:

i am using a VST sampler that can only accept 16 bit files, however
the project and the samples I am using are all at 24 bit. So what I
did was dither the 24 bit samples to 16 bit using the cranesong dither
file (at-25db) and use the new 16 bit samples in my 24bit project.

That would make your files pretty well dithered.

My question is: if this sample is continuously plaing at 0db, will
this be a problem when I dither the project?

Rule of thumb, if you properly dither something once in the 16-24 bit realm,
dithering it again is at worst a waste of time, but audibly nothing gets
hurt.

will I actually be dithering twice?

Probably, but so what?

What if i use multiple dithered samples at the same time?

Proper dithering in the 16-24 bit realm is at worst benign.

i was debating just truncating the sample and dithering the
project, but this didn't feel right.

When in doubt about dither in the 16-24 bit realm, just dither again.

If you were working with 8 bit samples, the dither is so large that you
might actually hear the difference if you unnecessarily dithered twice.

This also applies to another instance where I had to convert a 24 bit
reverb impulse to use with SIR, i had the same dilemma.

No dilemma. If in doubt about dithering in the 24-16 bit realm just dither
again. No audible harm will be done. The noise floor of the rest of your
project is likely to be high enough that the noise added by double dithering
makes a microscopic change.

thanks for your input.

best
david


"Arny Krueger" wrote in message ...
"David Kakon" wrote in message
om
Hi:

i am using a VST sampler that can only accept 16 bit files, however
the project and the samples I am using are all at 24 bit. So what I
did was dither the 24 bit samples to 16 bit using the cranesong dither
file (at-25db) and use the new 16 bit samples in my 24bit project.

That would make your files pretty well dithered.

My question is: if this sample is continuously plaing at 0db, will
this be a problem when I dither the project?

Rule of thumb, if you properly dither something once in the 16-24 bit realm,
dithering it again is at worst a waste of time, but audibly nothing gets
hurt.

will I actually be dithering twice?

Probably, but so what?

What if i use multiple dithered samples at the same time?

Proper dithering in the 16-24 bit realm is at worst benign.

i was debating just truncating the sample and dithering the
project, but this didn't feel right.

When in doubt about dither in the 16-24 bit realm, just dither again.

If you were working with 8 bit samples, the dither is so large that you
might actually hear the difference if you unnecessarily dithered twice.

This also applies to another instance where I had to convert a 24 bit
reverb impulse to use with SIR, i had the same dilemma.

No dilemma. If in doubt about dithering in the 24-16 bit realm just dither
again. No audible harm will be done. The noise floor of the rest of your
project is likely to be high enough that the noise added by double dithering
makes a microscopic change.

In article ,
Arny Krueger wrote:
No dilemma. If in doubt about dithering in the 24-16 bit realm just dither
again. No audible harm will be done. The noise floor of the rest of your
project is likely to be high enough that the noise added by double dithering
makes a microscopic change.

This is good advice. Unless you know a lot about how your signals are
being processed, it's always most safe to dither at each and every
processing step. The risk of not using dither is that you might lose
low level information that didn't have to be lost and your work might
not sound so good. The only downside to dithering is that you'll add
noise. The reality is that the noise you'll add may well be
practically microscopic in level.

While you may be able to 'get away' with not dithering, there isn't
much to lose by doing it at each and every processing step that you
have control over.

My biggest complaint is that some systems don't offer you the
opportunity to dither the processed signal to the processor's output
resolution. Thankfully, this is a lot less common of a situation than
it was a few years ago. But, it is a big problem with digital IMHO,
and if you can choose to use processors that offer high resolution
outputs and dither to that resolution, you'll probably be very happy
with their sound. Dither is not the only parameter important to how a
digital processor will sound, but it is a necessary component of a
high quality processor.

In other words, be glad that you have been offered the opportunity to
dither your processing and take advantage of it. Of course, you want
to keep the resolution between processing steps as high as possible
(e.g. 24 bit PCM or 32 bit floating point) so that the dither will add
negligible noise to your signals, but still, every processing step
should be dithered in the ideal case.

There is no rule against to "multiply dithering" a signal. This
concept is actually a complete fallacy: dither is needed exactly once
for each processing step that results in a wide word that needs to be
truncated to a smaller output word. The fallacy is that just because
you processed a signal once and had to dither it, it doesn't mean that
you don't have to dither any subsequent processing steps. The math of
any subsequent processing really doesn't care about what you did to
the signal before - if that processor is going to generate a wide
internal representation of your signal and truncate that to a smaller
output resolution, then that signal needs to be dithered before the
truncation.

The reality is simply that each and evey processing step that takes a
wide internal word and outputs a narrower word needs to be dithered.
If you dithered before, it doesn't matter. It's actually that simple.


Regards,

Monte McGuire

Monte P McGuire wrote:

There is no rule against to "multiply dithering" a signal. This
concept is actually a complete fallacy: dither is needed exactly once
for each processing step that results in a wide word that needs to be
truncated to a smaller output word. The fallacy is that just because
you processed a signal once and had to dither it, it doesn't mean that
you don't have to dither any subsequent processing steps. The math of
any subsequent processing really doesn't care about what you did to
the signal before - if that processor is going to generate a wide
internal representation of your signal and truncate that to a smaller
output resolution, then that signal needs to be dithered before the
truncation.

The reality is simply that each and evey processing step that takes a
wide internal word and outputs a narrower word needs to be dithered.
If you dithered before, it doesn't matter. It's actually that simple.

Unless you're using noise-shaped dithers, that is. Stick with TPDF or
other simple algorigthms for the inter-process dither.