KVR Audio

Sundr0wn · Post by **Sundr0wn** » Thu Aug 05, 2021 6:50 am

Should i dither when simply resampling the Sample rate of a wav file?
So for example i open a 16bit 96kHz file and resample it to 48kHz, then export it again to 16bit. Should i dither on export? Since it is also reducing the amount of information does it also result in quantization errors?
What if i do this edit in 32bit float (audacity for example), should i dither then on export since i am processing the audio once at 32bit?

BertKoor · Post by **BertKoor** » Thu Aug 05, 2021 7:37 am

Sundr0wn wrote: ↑Thu Aug 05, 2021 6:50 am Should i dither when simply resampling the Sample rate of a wav file?

Nope, You only dither when reducing the bit depth (eg from 32 bits to 16)
Bit depth is the vertical axis, sampling rate is the horizontal axis. These are orthogonal. You can resample while staying in 16 bits (not convert to 32 bits) without audible issues.

Sundr0wn wrote: ↑Thu Aug 05, 2021 6:50 amthen export it again to 16bit.

Yes, that is the point where you should dither.
It should be an option in the dialog that performs that action in any audio editor worth its salt.

Sundr0wn · Post by **Sundr0wn** » Thu Aug 05, 2021 7:53 am

BertKoor wrote: ↑Thu Aug 05, 2021 7:37 am Yes, that is the point where you should dither.
It should be an option in the dialog that performs that action in any audio editor worth its salt.

Even if the input and output file are both 16bit? Just the resampling happens at 32bit float.
If yes is it still preferable to resample at 32bit float, or am i better of to resample that file in 16bit and not introduce additional dither?

I am aware there is no audible difference but when working for a client with the goal to keep as close to the original file as possible while getting the technical best result.

BertKoor · Post by **BertKoor** » Thu Aug 05, 2021 8:07 am

I'd stop worrying about it by using this free utility:
https://www.voxengo.com/product/r8brain/
created by one of the best DSP gurus I'm aware of...

Sundr0wn · Post by **Sundr0wn** » Thu Aug 05, 2021 11:32 am

Spent the last three hours comparing sample rate converters. I went with finalCD, seems to be the best one out there and the SRC comparisson results are even slightly better than r8brain pro. I just hope i get it working right since there are no settings or manual.
Got the frontend, but seems like it's using an old version of the resampler?

BertKoor · Post by **BertKoor** » Thu Aug 05, 2021 11:37 am

Well, you either go for a tested & complete all-in-one tool that "just works" or you go for a "slightly better" one that takes a day to figure out how it works and if/how to install stuff it depends on. You totally have the right to waste your own time

Sundr0wn · Post by **Sundr0wn** » Thu Aug 05, 2021 11:46 am

My man, you have no idea how much time i've already wasted and am still going to waste

At this point i'm happy to spend that extra time and figure out the most 'perfect' way. At least then once i'm good i'll be good for good (i hope). If i was worried about time i'd just stick to my current audacity set up and keep the around 1800 files i have to redo anyway if i switch my converter now.

BertKoor · Post by **BertKoor** » Thu Aug 05, 2021 11:58 am

Hmm, have you closely studied the comparison between FinalCD and R8brain?
This is the most clear picture I found:

Screen Shot 2021-08-05 at 13.55.31.png

(from https://src.infinitewave.ca) and I see the difference is only in frequencies beyond 21.8kHz. Both are close to theoretically perfect, humanly inaudible.

And watch out you don't set the filter in FinalCD to something else than "sharp" otherwise you'll get very bad results:

Screen Shot 2021-08-05 at 14.00.59.png

Sundr0wn · Post by **Sundr0wn** » Thu Aug 05, 2021 5:14 pm

I did set it to sharp dont worry

Well to me FinalCD looks to be much closer to the 'ideal filter' line in Transition and Passband, even if it overshoots just a few pixels at the very bottom. Obviously that difference is next to null, however the difference between 120$ and Free isnt

And r8brain free does, at least on src check, perform worse than either pro or finalCD.

Hermetech Mastering · Post by **Hermetech Mastering** » Sun Aug 15, 2021 9:55 am

I've been using R8BrainPro, FinalCD and RX Advanced SRC all for over a decade now. They all do the job extremely well, and I doubt anyone could tell the difference in a 96 to 44 blind A/B. The null test results are below the threshold of noticeability.

I've recently quit all iZotope products so don't use RX any more, but I'll use FinalCD for "best practise" and R8Brain Pro for convenience and speed. The FinalCD frontend is older than the current command line utility, so it doesn't present you with all the latest options, like upsampling etc. I hate command lines so do find myself using R8Brain Pro more and more, and FinalCD less and less. Wish he'd update the frontend...

The free version of R8Brain isn't nearly in the same league (check the InifineteWave site), in that case I'd always use FinalCD.

Dither on Sample Rate conversion?