Uli Brueggemann on room correction
From DRC
This has been extracted from a message of Uli Brueggemann sent to the TacT audio user group. Uli Brueggemann is the developer of marcV.
My simple view is the following. If we want to apply a FIR filter to a signal stream we do it by a process called convolution. This is a mathematical procedure. Of course there exist different implementations and the goal is to get it efficiently coded to have a low latency real time application finally. A very good implementation of such a convolution engine is Brutefir written by Anders Torger. The TacT RCS is from this point of view also a convolution engine implemented with some tricks (multi-rate filter). As we all heard another engine is in development by Dynaton. And beside the room correction there are really a lot of different convolution engines on the market, e.g. for reverb generation (SIR by Christian Knufinke). The task of all these units is simply computation. Get a signal and convolve it with a predefined set of filter parameters. By hardware, specialized algorithms or brute force computation (Brutefir). So this is one part. Of course we expect that it is done well and allows us to optimize the second part.
This second part is how to generate the filter parameters. And IMHO this is ART. There are several thousand ways to interprete the measurements and to define the targets. So to get the good sound we all are hunting for we search for the right way to compute the filters. I truly believe Denis will agree that room correction is a compromise. We want a good tonal sound with a detailed stage and transparency. But we also want to move around in the room and we do not like to be nailed at the listening position. So e.g. a simple inversion of the room response to get the filters does not work.
GoodVibrations was a first try to get better parameters for the TacT RCS convolution engine. So beside already existing programs like DRC (Denis Sbragion) or ACXO (Patrick Cazeles) or the TacT or GV software and even selfmade corrections (Matlab, Scilab filters) and upcoming software (Dynaton) I have started to write marcV (multi-amp room correction 2005). My goal is to put in all experiences and knowledge to achieve a good room correction with multi-amp systems (of course single amps work too). And it is a challenging task. We need more than just some simple assumptions. An ear hears different than a microphone. There is a lot of psychoacoustic involved.
To conclude please let me repeat: generating room correction filters is an ART. It should not be limited by the convolution engine anyway. I believe we will get much better listening in the next future.
Note from Denis Sbragion: as expected by Uli, I completely agree with him.
