[LAU] About Algorithms

[Fons Adriaensen]

On Sun, Jul 17, 2011 at 02:03:01AM +0300, Moshe Werner wrote:

> Yet there is one endless discussion between friends and colleges of mine
> that I usually do not participate in due to the lack of knowledge.
> The topic is algorithms of software (DAW) mixers. Some people say that they
> can hear a sound difference between several kinds of software mixers (e.g.
> Cubase, Protools etc.).
> I must say that I never made any serious A\B testing but I didn't notice
> that there is any difference. Although I do work with Pro-tools and Cubase
> (in other studios), most of the time I'm actually using Ardour (and I'm
> loving it).
 
I assume you refer to the basic _mixing_ algorithm, i.e. summing signals,
and not to e.g. effect plugins. For the latter there is usually a zillion
ways to do things, and of course they will sound different. As to the
former, the basic mixing, this is just a lot of nonsense.

Sad facts are that:

- The pro audio world is today infested with the same type of pseudo-
science that hit the hifi market 25 or more years ago and that makes
some people pay $3000 for a piece of cable and two connectors, or makes
them spend money on machines to replace old and tired electrons by fresh
young ones which sound better.
- The same wave of nonsense now hits also the software world.

The reasons are simple: basic problems have been solved, to create a
'competitive edge' you have to add snake oil.


> My questions would be:

> 1. Is it only me that can't here a difference between different DAWs mixing
> algorithms?

Unless some of those algorithms are completely wrong (which would require
quite a high level if ignorance from the designer), nobody will hear any 
difference. Those that claim they can should prove it in a blind test.
I know of no such test that ever demonstrated this.

> 2. To the developers out there, what is your opinion? Is there a
> better/worse algorithm, or is the whole thing another "pay 600$ for this
> software - it has great algorithms!!!" hype?

There isn't much 'algorithm' to speak of, it's just adding. And single
precision floating point provides all the precision you need. There *are*
some issues if you ar mixing thousands of signals - then some ways of doing
it are better than others. But this doesn't occur in normal audio engineering
practice.

> 3. If there is a difference what's the explanation?

See previous point. Explaining this would lead us very far.

> 4. Analog emulation plugins. How does one "emulate" analog waveforms in a
> digital world? That sounds like a paradox to me.

One doesn't emulate 'analog waveforms'. What's done is to reproduce
digitally the defects of some analog equipment (particular types of
distortion in a compressor for example), or the 'look and feel' of
them. There also recently a wave of 'exact digital copies' of e.g.
Neve equalisers. There's no reason why any of these should be better
than one that is not an 'exact digital copy'.

HTH,

-- 
FA