Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[blueboy]

If a 44.1/16 audio file that has been dithered to 16bits is then upsampled to 48/16, what is the effect on the dither?

As the intent of dither is to remove the correlation between the signal and the noise from quantization errors, and upsampling will affect levels in the newly created audio file, will the benefit of dither be negated, or will the impact be minimal?

For some background on why I am asking this, read here:

http://recforums.prosoundweb.com/index.php/t/7841/11771/?SQ= a4d174223251e166079352196d139fb7

I find it hard to believe that this would be done to a mastered audio file, but I am just curious as to technically what would happen.

Thanks.

JL

[bobkatz]

blueboy wrote on Sun, 06 November 2005 17:42

If a 44.1/16 audio file that has been dithered to 16bits is then upsampled to 48/16, what is the effect on the dither?

The dither in the original is no longer effective, it just becomes an additional noise floor. You need to upsample to the longest wordlength of the upsampling processor. Yes, the original dither's effects will be "negated"---but not exactly negated, just that the original dither did its job and you will eventually need another dither when reducing the wordlength again later on.

For example if you upsample to 48K with the SRC built into your DAW, which is likely operating at 32 bit float, then you need to upsample to 48/32.

Technically what would happen is the sound quality goes downhill very fast, which is why a mastered, 16 bit file is just not designed to be further processed.

BK

[danlavry]

blueboy wrote on Sun, 06 November 2005 22:42

If a 44.1/16 audio file that has been dithered to 16bits is then upsampled to 48/16, what is the effect on the dither? As the intent of dither is to remove the correlation between the signal and the noise from quantization errors, and upsampling will affect levels in the newly created audio file, will the benefit of dither be negated, or will the impact be minimal? For some background on why I am asking this, read here:  http://recforums.prosoundweb.com/index.php/t/7841/11771/?SQ= a4d174223251e166079352196d139fb7 I find it hard to believe that this would be done to a mastered audio file, but I am just curious as to technically what would happen. Thanks. JL

I would dither a second time, when taking the output of the SRC down to 16 bits. But this time, the need for dither, as well as it's effectiveness will be less then the first time you dithered.

Why? because, the noise floor of the original signal is a bit higher then the noise floor of a raw unditherd signal. The idea of using dither requires that the signal be accurate to better then the final word length.

An extreme example: say you have -60dB noise floor or so. The noise is so high that it does not matter what you do at bit 16.

Using dither on dither is not as extreme as -60dB dynamics, but you are "somewhere in between" needing (or not) and benefiting (or not) from more dither.

I will test it and get back with more solid statement. My best guess is to dither a second time. Most of the trad off is already done, and more dither will not add too much noise... The noise degradation second time is a lot lower...

Regards
Dan Lavry
www.lavryengineering.com  



 

[blueboy]

Thanks very much for the replies Bob and Dan.

I am constantly amazed at the wealth of knowledge, and the willingness to share it on these forums.

The DualDisc I am referring to sounds to me like it was upsampled on the DVD side. The highs sound like they went through SRC, and a lot of the low level detail seems to be "dirtier", leaving it with a flat sound.

It's not terrible, it just sounds less natural and delicate overall. Hopefully I won't find out that the DVD side was actually mastered at 48k separately and it's my converters that sound bad at 48k vs. 44.1k.

Regards,

JL

[Ronny]

danlavry wrote on Mon, 07 November 2005 15:53

blueboy wrote on Sun, 06 November 2005 22:42 If a 44.1/16 audio file that has been dithered to 16bits is then upsampled to 48/16, what is the effect on the dither? As the intent of dither is to remove the correlation between the signal and the noise from quantization errors, and upsampling will affect levels in the newly created audio file, will the benefit of dither be negated, or will the impact be minimal? For some background on why I am asking this, read here:   http://recforums.prosoundweb.com/index.php/t/7841/11771/?SQ= a4d174223251e166079352196d139fb7 I find it hard to believe that this would be done to a mastered audio file, but I am just curious as to technically what would happen. Thanks. JL I would dither a second time, when taking the output of the SRC down to 16 bits. But this time, the need for dither, as well as it's effectiveness will be less then the first time you dithered. Why? because, the noise floor of the original signal is a bit higher then the noise floor of a raw unditherd signal. The idea of using dither requires that the signal be accurate to better then the final word length. An extreme example: say you have -60dB noise floor or so. The noise is so high that it does not matter what you do at bit 16. Using dither on dither is not as extreme as -60dB dynamics, but you are "somewhere in between" needing (or not) and benefiting (or not) from more dither. I will test it and get back with more solid statement. My best guess is to dither a second time. Most of the trad off is already done, and more dither will not add too much noise... The noise degradation second time is a lot lower... Regards Dan Lavry www.lavryengineering.com

Please do test this, Dan, but don't be surprised if the double dither has much less detrimental effect than not adding dither during a quantization, even when source was word increased from dithered 16 bit, than processed and dithered again on the reduction back to 16.