Qwerty
New member
Do you remember going around and around the mulberry bush when they released Sonar v4? I said it sounded better, you pointed out that different didn't necessarily equal better.
Well, a new version of Sonar is released and it seems another audio engine is being spruiked as sounding sooper-dooper-excellenter than previous releases.
This thread on the Cakewalk forum has been doing laps on this issue -
http://forum.cakewalk.com/tm.asp?m=702782
After five pages of spin and mis-information, Ron Kuper - CTO @ Cakewalk, just weighed in with this example.
In the light of our previous conversations I found it ironic how history was repeating itself... Anyway, the CTO agrees with you!
Ciao!
Q.
Well, a new version of Sonar is released and it seems another audio engine is being spruiked as sounding sooper-dooper-excellenter than previous releases.
This thread on the Cakewalk forum has been doing laps on this issue -
http://forum.cakewalk.com/tm.asp?m=702782
After five pages of spin and mis-information, Ron Kuper - CTO @ Cakewalk, just weighed in with this example.
In the light of our previous conversations I found it ironic how history was repeating itself... Anyway, the CTO agrees with you!
Ciao!
Q.
I imported a 44.1kHz/24-bit mono test tone (sine wave @ 1kHz) into SONAR. I then cloned it 31 more times, creating a 32 track project. I set the first track's gain to 0dB, the second's to -1dB, the third's to -2dB, and so on. So in the end I was summing 32 tracks, from 0dB to -32dB in gain.
(Let's pause a second here. Maybe that's a contrived set up. But I didn't think so when I started. I figured a typical mix will have 32+ tracks, and probably have volumes spread out more or less the way I did it.)
All these tracks were bussed into a main which (obviously) was now overloaded. So I set the master gain to something like -7dB to compensate.
I then exported the mix to stereo 24-bit, using both the 32-bit engine and then the 64-bit double precision engines. (By the way all dithering was turned off in SONAR.)
I then loaded these two mixes into Sound Forge and subtracted one from the other, doing a nulling (A-B) test. If the 2 engines yield the same mix, then (A-B) should have resulted in a flat line signal, they should have nulled out.
The resulting (A-B) signal was not a flat line. Not even close.
Sound Forge's statistics said the minimum sample value in the (A-B) was -2 and the maximum sample value in (A-B) was 276. In other words, there were 9 bits of error introduced! Even with real world physics, noise, etc, this is a sonically significant amount of error.
Does this mean the 64-bit engine sounds "better"? That is subjective, it's up to you. But does it sound different, actually sound different in ways that discerning ears can tell? Yes, without a doubt.