[mightyduck]

Originally Posted by Greg Malcangi

True, providing you don't use any inserts, any sends or any sub busses and providing your whole mix fits onto 1 DSP chip.

You're trying to redefine the argument. The fact is that the 48 bit mixer plugin, the entire mixer, is 48 bit fixed. The other parts of the data path you mentioned are not part of the double precision plugin. Do you understand that?

Why doesn't it have to be? Everytime the audio goes into and out of an insert, a send or to a sub buss the 48bits gets truncated (or in some cases dithered) to 24bits. Are you really saying that a large number of truncation or dither artefacts are never going to cause an audible degradation of the signal quality?

It gets dithered with the dithered mixer. Dither is not an "artifact". It prevents artifacts. Even truncation at that level is trivial, and, unless your analog noise floor is below -110dBfs or so, inaudible.

What you don't realize, perhaps, is that 32 bit float also has only 24 bits of precision, and it cannot be dithered. So you are stuck with truncation or rounding, and each time the significand changes you throw more quantization error into the calculation. The quantization error in floating point is not limited to the lsb area, it gets written into the file at much higher amplitudes also, and dither is neither practical nor effective there.

"Definitively", really? Show us the proof then.

Look up Andy Moorer's paper on the subject, for starters, if you like. Andy Moorer, who was with Sonic Solutions at the time. Maybe, I'll post some other stuff later, but this is not really controversial, although you seem to want it to be. You could also research some of James Johnston's work if you want to corroborate Moorer's statement.

So why did you start by saying: "I see and hear a lot of spin, from Avid, and from others, about how the "new 32 bit floating point" mixer is an improvement."?

Because that's what they are doing. And its b.s. They did what they did in order to keep compatibility with RTAS in the interim. Otherwise they would have had a mutiny.

Is 32bit FP problematic? Yes but so is the TDM mixer. The reality is more complex than simply saying the TDM mixer is better than the 32bit float mixer. There are plenty of other threads on this subject for you to educate yourself but you appear to have started this thread for an argument rather than for sharing or learning.

No, Its that simple. The TDM mixer is better than 32 bit float, under any reasonable analysis. There is no rational argument to the contrary. If you can't accept the fact that 32 bit float is widely used in audio software for programming ease, then you are simply being naive.


Let me guess. You bought a brand new P.T. 10 system, right? : )