dithering

[Foreverain4]

i am trying to figure out exactly what this does. has anyone heard of the dithering cd from crane song? www.cransong.com does anyone know how to create your own dither track? you should be able to do this with the synthesis function in sound forge right? any info on the matter would be greatly appreciated.

[Nika]

Foreverain,

I do have a copy of the Cranesong DitherCD. I'm not sure exactly what your question is, though? The way you use the CD is you copy the data onto a track in your session and then reduce the level exactly 25dB (from my recollection) and it dithers the track.

Was that what you wanted to know?

Perhaps some additional information on the concept of dither would help? Try reading this - it goes rather quickly:

http://www.cadenzarecording.com/dither.html


Cheers!
Nika.

[Foreverain4]

dude, thanks so much for the info! my question was more how do you create dither noise? i mean, is it just white noise at a certain frequency? how did these people at cranesong come up with this as audio rather than using a pluggin? i read through your article last night, and plan on studying it more in depth. i think in the digital recording world that this concept should be understood. anyway, later!!!!

[Nika]

Lynn,

Dither is white noise, which is random noise at all frequencies, and is created by taking random numbers and assigning them to bit values.

There are many different algorithms that are used, and often times people will take their random numbers and then filter them so that they have a certain behaviour (less
noise in some frequency range and more noise in another, which is no longer white noise). I haven't put DitherCD up on a scope yet to see how they're distributing their noise, but I've looked at other ones, and it becomes a matter of preference as to which you prefer. DitherCD, IDR, POW-r, UV22, and many others are all valid yet different forms of filtered dither noise.


Nika.

[TeeCee]

Foreverain4:
I've been waiting to see where this went before I put my 2 cents in. If you have Sound Forge, and you just want to dither a file, you can use Bit-Depth Converter and just not actually change the bit depth. There are several dithering options in there. You can do this to a silent file created in Sound Forge to generate a dither file. Mixing the result with an existing file will not neccesarily produce the same results as dithering a file directly as you will be adding dither noise (not sure if that's the proper term) to the file where Sound Forge may be dithering the least significant bit of the file.

Nika:
Nice article. The graphs are a great plus. Would you care to comment on the effectiveness of using Sound Forge's dither options? Does it actually alter the lower bits or does it sum the truncated file with generated dither information? How was the dither actually applied in your tests?

[Nika]

TeeCee,

I haven't used SoundForge's dither at all, so I'm a bit useless in that quest. If they do it as everyone else does it (why wouldn't they?) then they add random noise and then truncate.

As for how it was applied in my tests I did it precisely that way - I generated random noise with a noise generator and then mixed it in with the sine waves - then did a bounce to disk at lower bit depths. I'm not sure if this answers your questions?


Nika.

[TeeCee]

Answers some, makes more ;). Did you do this in the digital domain? If so, did this method raise the peak level of the sound file? I would expect that you did this in the digital domain as going through an analog circuit would almost certainly add noise to the file, anyway.

[Nika]

[QUOTE]Originally posted by TeeCee
Answers some, makes more ;). Did you do this in the digital domain?[b]

Yup

[b]If so, did this method raise the peak level of the sound file?

Yup.

Nika.

[TeeCee]

So you truly mixed the two. Interesting. I just tried this in Sound Forge generating a 1kHz sine wave at 44.1kHz 16 bit first with a -12dB peak and then with a -6dB peak. I then "converted" the bit depth to 16dB applying dither and noise shaping, applying dither, and then nothing else. Applying dither increased the peaks by a measured .1dB each time. I use the Scan function under Normalize as I have verified that the Statistics function can differ in results from Scan, Find, and Graphic Dynamics.

This is very important. I typically normalize my 24 bit wave files to something like -.03dB prior to bit depth conversion to 16 bit with dither and noise shaping. According to these results, my final wave file could clip. I thought Sound Forge was replacing the lowest bit with a dithered equivalent to break the truncation pattern. But it seems like it's adding the dither algorithm's output.

[Nika]

Yup. Simply randomizing the last bit is not dither. Dither is ADDING noise. It is important to recognize this. Simply juggling the last bit doesn't ADD anything. Instead it MASKS things and distorts the signal.

Nika.