Dithering an Organ Recording

[twotoedsloth]

Just recorded an organ recital... With all of the background white noise, the organ blower and the HVAC fans, who needs dither? The noise floor seems to take care of that for itself.

[The Elf]

But what about if you fade out at the end? Without dither it descends into digital grainery...

[Martin Walker]

True, but that spoils the elegance and simplicity of the original post Elf

Martin

[twotoedsloth]

Sorry for my last post, I wasn't clear what information I was looking for.

Would it be detrimental to the recording to add dither noise with the already rather high noise floor? At the moment, it is pretty hard to hear a difference between the dithered and undithered 16 bit file. I think that I can hear a difference between both of those and the original 24 bit file, but maybe I'm kidding myself.

Thanks kindly,

Peter

[Hugh Robjohns]

No, it won't be detrimental, and if you want to include any fade-in/fade-out at the start/end of traks then correct dithering for the chosen word length is essential.

If the noise floor is as high as you suggest there should be no audible difference between the 16 and 24 bit versions since the ambient noise floor will totally swamp the sigital system noise (assuming the same peak levels) but different noise-shaped dithers can introduce very subtly different tonalities because of the wa they pile noise in narrow-ish portions of the HF.

[dmills]

but different noise-shaped dithers can introduce very subtly different tonalities because of the way they pile noise in narrow-ish portions of the HF.

I have always suspected that the major effect here was actually intermod, and that this will vary between different repro chains by enough to make the use of highly noise shaped dithers problematic in terms of how well things translate.
Sure a particular one might sound best in the mastering room with a pair of PMC driven by Brystons or whatever, but putting that much energy up at 20K+ seems to me to be asking for it when played on a lesser system (Worse the effect will be somewhat unpredictable).

But, yea if you are doing fades in the digital domain you should dither the word length reduction.

Regards, Dan.

[Jeraldo]

.....but putting that much energy up at 20K+ seems to me to be asking for it when played on a lesser system......

Lots of energy at the Nyquist point, too, if that matters.

[dmills]

By definition, no energy at that point actually, close is fine, but the maths only works for sampling rate strictly greater then twice the bandwidth..... By definition any energy at that point gets aliased to DC.

The linearity thing is more a problem with downstream repro chain gear then the maths.

Regards, Dan.