View Single Post
Old 06-09-2019, 05:11 PM   #5
Human being with feelings
Join Date: Sep 2010
Posts: 9,003

Yes "shouldn't be glaring"

Upsampling SD program to HD would make it sound more transparently like the HD original on DA converters that run cleaner at HD vs SD. There is no quality added. It's simply making the converter machine not have to anti-alias filter the sampling rate frequency. It's preventing degradation.

That might sound ass backwards...
How could a further conversion - upsampling - prevent degradation?!
The degradation in this example would come about from the AD converter's performance in SD mode. Not from the reduced to SD data. Because even though the data has been reduced to SD, that doesn't touch the actual audio band.

Again, this assumes no one is entertaining the idea that any artifacts above the range of hearing are useful or perceivable. Many people seem to entertain this notion. Some use it as a strawman to state "You can't hear those frequencies so HD is bs!". Trying to suggest that the purpose of HD sampling was to capture above the range of hearing.

Fixed point digital does in fact lose resolution as the volume goes down. Literally losing fidelity as you lower the volume. Put a half volume signal into a 16 bit container and you use 8 bits. That's an 8 bit recording. You need to pay attention to levels with 16 bit. That's still a decent amount of resolution for a finished master. Pop stuff that's squashed dynamically doesn't even hit the edges! Classical or artsy music might get hit a little. This WAS a factor in starting the volume wars. 24 bit gives you an 8 bit "noise floor" bed with 16 bits above that. True 96db dynamic range with pretty full resolution on the very bottom.

Floating point formats preserve a much wider range when the decimal can float and preserve the 'meat' of the numbers instead of padding zeros. Math.

Consumer format is 24 bit fixed... so get your levels in order at the end of your mixing chain!
serr is offline   Reply With Quote