Audio Engine?

[chagzuki]

I thought that SX uses 64bit doubles or whatever it's called. And a handful of other hosts.
So in that case they should be identical in sound quality. ?

[nuffink]

Dumbest thread ever

I cant agree more.

Not even close to being the dumbest thread today.

[Robert Randolph]

Nobody said it was the primary problem, in fact I've said MANY times it was one of the least significant, but most often denied problems.

One again, ive given a basic outline of the test, if you dont wanna do it yourself, f**k off. If you want to go do some reading and learn the math/tech behind it, go ahead... otherwise f**k off.

It's not alchemy.

And cubase does use the full 80-bit fp register, which basically means it accumulating into a double. summing with no dither it will probabaly sound similiar to samplitude, and logic (maybe?)... but not anything like paris, sawstudio or protools TDM

AS for all this, when comparing the MIXING engine, differences are greater than summing. Id love to do a mixing engine test, that would be something interesting

Im agreeing now, dumbest thread ever. I will make some time to design a test for mixing engines (which of course includes summing portion of the app), mark my words now SOME HOSTS WILL HAVE SIMILIAR or EXACTLY THE SAME RESULTS, not most, but some.

-R bert

[chagzuki]

Tracktion is also to have 64 bit double doo-dah mixing, so it seems that it's becoming the norm.

I don't think this thread is stupid. The non-programmers here like me want to get the facts straight on these issues, and concensus appears to have been reached.

[nuffink]

Nobody said it was the primary problem, in fact I've said MANY times it was one of the least significant, but most often denied problems.

One again, ive given a basic outline of the test, if you dont wanna do it yourself, f**k off. If you want to go do some reading and learn the math/tech behind it, go ahead... otherwise f**k off.

It's not alchemy.

And cubase does use the full 80-bit fp register, which basically means it accumulating into a double. summing with no dither it will probabaly sound similiar to samplitude, and logic (maybe?)... but not anything like paris, sawstudio or protools TDM

AS for all this, when comparing the MIXING engine, differences are greater than summing. Id love to do a mixing engine test, that would be something interesting

Im agreeing now, dumbest thread ever. I will make some time to design a test for mixing engines (which of course includes summing portion of the app), mark my words now SOME HOSTS WILL HAVE SIMILIAR or EXACTLY THE SAME RESULTS, not most, but some.

-R bert

Since you're apparently not willing to put up (and this is the dumbest thread ever) can we assume you'll now shut up?

[kilroy]

The most disagreeable "audio engine" issues originate from the tune.

[Robert Randolph]

Nobody said it was the primary problem, in fact I've said MANY times it was one of the least significant, but most often denied problems.

One again, ive given a basic outline of the test, if you dont wanna do it yourself, f**k off. If you want to go do some reading and learn the math/tech behind it, go ahead... otherwise f**k off.

It's not alchemy.

And cubase does use the full 80-bit fp register, which basically means it accumulating into a double. summing with no dither it will probabaly sound similiar to samplitude, and logic (maybe?)... but not anything like paris, sawstudio or protools TDM

AS for all this, when comparing the MIXING engine, differences are greater than summing. Id love to do a mixing engine test, that would be something interesting

Im agreeing now, dumbest thread ever. I will make some time to design a test for mixing engines (which of course includes summing portion of the app), mark my words now SOME HOSTS WILL HAVE SIMILIAR or EXACTLY THE SAME RESULTS, not most, but some.

-R bert

Since you're apparently not willing to put up (and this is the dumbest thread ever) can we assume you'll now shut up?

Did you not read my entire post? read it again, particularly the end, and then f**k off.

-Robert

[nuffink]

Yeah I read it. I look forward to the test.

b.t.w. Do you mind awfully if I don't f**k off? I enjoy following your attempts to buy some talent so much.

[nuffink]

Oh and I'm sure you'll appreciate a nudge about those upcoming tests every now and again.

[Robert Randolph]

Oh and I'm sure you'll appreciate a nudge about those upcoming tests every now and again.

In all seriousness (sarcasm aside), please do. I am prone to start working on things then forget. If you wanna be a bit more friendly (as I) i do need at least 5-6 people to help out providing test hosts, all I have currently is sonar, tracktion and samp soon. All else has been sold... (or banned from my computer forever).

and buying talent... hrm, buying happiness or contentment is more like it.

-R bert

[tony tony chopper]

Using 64-bit doubles for mixing and problem gone

Well it's not gone, it's just less important.

What most here fail to understand is that digital is always limited. You could mix using 1024bit floats, there would still be a loss. So which loss is acceptable? The one that comes with 32bit float mixing in a sequencer is already wayyy beyond what anyone could notice. If not, you can count on robert to come up with some tests

Tracktion is also to have 64 bit double doo-dah mixing, so it seems that it's becoming the norm.

yeah, funny, but VST(i)'s still process from/into 32bit buffers anyway.
It's just marketing, people think that their mixes sound like shit because it's done in 32bit float, someone turns all of his single floats into doubles, & claims to have a better audio engine. Of course at the end the user is the one f**ked, his mix still sounds like the same shit, but the CPU usage grew, for the same results.

[Spe3D]

This is where the 64-bit chip question comes into play again

Regards,

Spe3d

[Mr. Slater's Parrot]

Well, I'm siding with gol on this -- at least until someone can show evidence that different hosts get different results when they add the same IEEE 754 32-bit floats together. Or until someone demonstrates that 16-bit or 24-bit recorded audio combined with 32-bit VSTi output samples mixed in a typical pop/rock song (because that's what I happen to be interested in) lose anything that can be discerned reliably when handled in a 32-bit floating point based system compared to any system using some representation with more precision. And even then, I probably would be happy sticking with 32-bit floats if the alternative has a significant performance penalty.

[jibnut]

Stupid thread, perfect example of brain masterbation.

1. There isnt a sound card made that can even represent the rounding errors from a 32 bit float. Even if you accumlate enough errors from summing of enough signals the resultant sound would be far to dense to hear it.

2. How the f**k do you record audio with a noise floor low enough that the rounding errors make any differance?

3. Even if you load a sample from a sample CD thers 256x more error from the 16 bit sample that you get when its mixed in the audio engine at 32bit float.

Oh well...

chris

[IIRs]

I'm not going to try to explain why, but Cubase VST doesn't sound as good as Tracktion on my machine.

VST sounds hard and flat, while Tracktion sounds warm & fluid, with much better stereo imaging.

You can argue as much as you like, I know what I hear

Can't comment on SX of course, haven't bought it..