AES: Benefits of 64-bit Mixing

Former Cakewalk CTO Ron Kuper, discusses the benefits of mixing with SONAR's 64-bit double precision audio engine (on 32-bt or 64-bit machines)

Category:

Music

Tags:

• Cakewalk
• SONAR
• Mixing
• 64-bit
• Double
• Precision

[andrewsimper]

This talk missed the point completely. The accumulation in the mixing should be done using doubles, but the audio buffers and gains are completely fine staying as floats: float g1 = 2.0, g2=0.5; float s1 = random(), s2 = random (); double sum = 0.0; sum += g1 * s1; sum += g2 * s2; sum += -g1 * s1; sum /= g2; float output = float(sum); What you get is s2 exactly as you would if you had the gains and buffers as double, but using floats for the gains and buffers is more efficient.

[mortalengines65]

With all due respect to the message.....there was a whole room of ENGINEERS who work with media every day and they couldn't get this right? You do see the irony in this, don't you?

[bwmac]

sorry you feel that way tomahawk360. I use cakewalk sonar-6 producer edition 64 and I watch all of cakewalks videos. My comment was not personal I just call it as i see it.

Bradon, from cakewalk, gives the best cakewalk videos. IMO.

Again no insult intended toward this person.

nuff said :)

[tomahawk360]

That's a really juvenile / ignorant thing to say. This is purely informational, and the information is good. It's YouTube for crying out loud, the audio is compressed no matter what.

[bwmac]

WoW you should have used 64 bit in this video/audio, LOL, Iv seen better quallity from the kids on here.

Come on guys, not very good advertizing.

[maxhodges]

the horror, the horror

(sorry, it's the lighting)

[basspig]

The 24-bit file was recovered, but with some quantizing errors resulting in stairsteps, and the 32-bit file looked perfect. In fact, I was able to keep reducing the 32-bit file until it was at -899dB before normalizing to 0dB brought visible quantizing error to the screen.

Whether moving to 64-bit float brings an audible benefit, well, I suppose we'll know after some time with the product and a lot of listeners have done blind comparisons.

[basspig]

Unquestionably, the more bits used in the calculation of any transform, the greater the accuracy by several orders of magnitude.

I was able to make some observations about the differences between 16- 24 and 32-bit float calculations in SoundForge. My simple test was to create a -90dB sinewave and reduce it by 10dB and then normalize the whole thing to 0dB. The 16-bit file became a no-signal file.

[genericaudioperson]

great discussion, despite no lightbulbs for the guy. he should have a zillion of these mini discussions, talking about the mix buss, summing, etc.