Topic: The great dither shootout...

[RebelRebel]

http://www.24-96.net/dither/index.htm

[china_rider]

It would be interestring to hear the source file for this.  I happened to pick the MBIT dither that I use but I have never encountered the type of hiss that is in these samples in my finished mix.

[it-goes-to-eleven]

Is there any info on the methodology of the test?

[RebelRebel]

you can email the author
dither@24-96.net

[china_rider]

Is it just my ears? It sounds like all the samples come from a 8th gen+ cassette.     If my dithered shows came out anywhere close to any of these samples I would not even bother with 24bit.  I'm far from a dither expert but not one sample in this test comes anywhere close the quality I get with a MBIT+ dither.

[scb]

it's the 54 db gain....documented on the bottom of the page

>>All files have been dithered to 16 bits from the same original 24 bit file. 54 dB of gain has been applied after dithering. The original files have been kindly supplied by Philipe Goutier of Steinberg. The samples using Megabitmax, Sound Forge, CoolEdit Pro and pow-r dither have been produced by Alexey Lukin. The Airwindows Mastering Tools samples have been produced by Chris Johnson. The samples using Wavelab, UV22 and Waves dither have been redone from the original versions by Philipe Goutier by Robin Schmidt in order to provide a consistent boost of gain over all samples. The files below are presented in alphabetical order. It should be pointed out that great care has been taken in the preparation of these samples and no dither product has been favoured above another. For questions or suggestions please contact dither@24-96.net.

[bhtoque]

it's the 54 db gain....documented on the bottom of the page

WTF?? why bother to do a comp on such a poor sample.

I'd have loaned them a mic pre if they needed one that badly.

JAson

[fozzy]

I was getting crappy D/L speeds, here is a yousendit link for the files

http://download.yousendit.com/47F4533373D4D588

[scb]

it's the 54 db gain....documented on the bottom of the page

WTF?? why bother to do a comp on such a poor sample.

I'd have loaned them a mic pre if they needed one that badly.

JAson

i was under the assumption that by adding all that gain, they thought they were really showing the audible noise each dither scheme adds

maybe?

[china_rider]

My browser was not showing the bottom of the page either....

They mention in the instructions to listen to the samples so the noise was hardly noticable.  It's so harsh that I don't think that is possibile.

[BC]

it's the 54 db gain....documented on the bottom of the page

WTF?? why bother to do a comp on such a poor sample.

I'd have loaned them a mic pre if they needed one that badly.

JAson

i was under the assumption that by adding all that gain, they thought they were really showing the audible noise each dither scheme adds

maybe?

Seems reasonable to me.

[RebelRebel]

Im sure he had a rhyme and a reason . He is a very well respected ME here.

He is my neighbor(literally, neighboring city)so I called and left a message on his machine to call me back.

if you have any specific questions, I am all ears. Speak up and I will relay them to him.

teddy

[BobW]

I have nothing solid to say about the dither test.
I just joined the thread to be in good company  +s all around.

FWIW, the differences in the dither seem to affect the least heard parts of the music.
Although it ceratinly isn't a moot point, as we want to hear as best and as much as possible,
it isn't nearly as important as location, mic selection, gain structure, pres, and beer temperature   

I've heard a few pros say that they like POW-R best of all.
I'm not sure what they heard to justify that.

[RebelRebel]

POW-R is favored by many professionals because it is a product of Weiss laboratories.  Weiss boxes are as high end as they come. POW-R dither is also a native scheme in some HIgh-End mastering DAWs(Sequoia, Samplitude, SAW?) Bob Katz uses Weiss products, and lauds them in his book.
As far as SRC Conversion, nothing tops the R8 Brain pro plugin, to my ears. (Will be trying weiss saracon soon...well see. )




a

I have nothing solid to say about the dither test.
I just joined the thread to be in good company  +s all around.

FWIW, the differences in the dither seem to affect the least heard parts of the music.
Although it ceratinly isn't a moot point, as we want to hear as best and as much as possible,
it isn't nearly as important as location, mic selection, gain structure, pres, and beer temperature   

I've heard a few pros say that they like POW-R best of all.
I'm not sure what they heard to justify that.

[china_rider]

I asked a few questions about the shootout... here is the response I got... makes more sense now....

The noise you are heard IS the dither. Here is how the files were prepared:

A file was recorded at very low level in 24 bit. So low, that when just cutting the lower 8 bits off, the dynamic range would introduce high distortion because the dynamic range wouldn't be large enough. Then, the different dither algos were applied and the file was rescaled to 16 bit.
After that, 54 dB of gain was applied, to make sure that you can hear exactly what the dither is doing, without the own noise of the playback equipment getting in the way (we amplified the dithered fade out digitally, so that the testees won't have to do that with equipment that we have no control over).

So, the noise you are hearing is the dither and little else, but the dither.
BUT, you wouldn't normally hear the dither noise so loud, nor would you hear the music fade out at such a high volume. So in order to conduct a proper test, it is advisable to turn the volume down low, just until you can barely hear the dithering noise and then listen, how long and how clearly you can hear the music play, while it fades out. The point of all these algorithms is to enable a high audible dynamic range without quantization distortion while producing the least amount of audible artefacts (noise).

I.e. in order to prevent the distortion that just cuts off the music and produces harsh harmonis, noise has been added to make sure, the total level (music plus noise) is always in a range that a 16 bit recording can faithfully recreate. That's the jist of dither.

To understand this better, first listen to the original file, then the one, that's been truncated to 16 bit without dither (you will hear the signal coherent distortion) and then listen to how different dithering algorithm manage to prevent this distortion from occurring. Notice, how the dither noise also makes the music last longer during the fade out, i.e. how the dither extends the dynamic range ... at the cost of noise.


Links to the original files (I think they're also on the site somewhere):

http://www.24-96.net/dither/Samples/24-96_Dither_Shootout_-_original_24bit_recording_plus_54_dB.wav

http://www.24-96.net/dither/Samples/24-96_Dither_Shootout_-_original_truncated_to_16bit_without_dither.wav

I hope I managed to explain reasonably well what dither is and how we prepared the samples. If you understand dithering, then the only missing link is that we applied 54 dB of gain after the dithering.

If you want another resourse on the idea of dithering, try Dithering on www.wikipedia.com. Understanding the concept of dither becomes easier when looking at the pictures (dithering is used in imaging too). Plus, you get to see a cute cat... when you compare the dithered cat to the undithered cat, you see that the dithered cat image has more noise but closer resembles the original, uncompressed picture.