Support for dithering

[aproc]

Hi James

After Reclock resamples the audio , it does the word length reduction from 32 bit float to 24 bit int (or whatever the output bit depth is selected) without dithering. Can you please implement the standard 2 LSB peak to peak TPDF dithering (with no noise shaping) for this word length reduction?

A properly dithered audio at 24 bit does sound superior to undithered audio. I have tested this in JRiver Media Center by disabling its internal dithering and then using Voxengo Elephant plugin for TPDF dithering. However, JRiver has limitations of its own so I can't always use it. Also, I find Reclock's resampling to be superior.

IgorZep in AVS Forum shared the code for TPDF dithering implementation in one of his posts. I can provide the link of that post if needed.


Thanks
aproc

 

[mindbomb]

You don't have to dither at 24 bit. It is only to improve performance at low bit depths. At high bit depths, noise is very low and inaudible, so the idea of making it sound better with a dither doesn't apply.

 

[aproc]

Perhaps you don't understand dither well. 24 bit dithered audio is mathematically more accurate than undithered audio. Many people are able to hear the sonic differences including me. You may refer to this discussion here where Bob Katz highlighted the importance of dithering when reducing the wordlength to 24 bit:
http://yabb.jriver.com/interact/index.php?topic=76912.0

Chronoptimist in AVSForum explained dither and its importance in very simple terms here:
http://www.avsforum.com/forum/26-ho...ia-center-owner-s-thread-47.html#post28074466

You may also read this to understand dither in more depth: http://www.users.qwest.net/~volt42/cadenzarecording/DitherExplained.pdf

When Reclock allows enabling dithering, certainly it will be an option and those who don't want dither for whatever reason can choose not to enable that.