Digital Room Correction

Has anyone looked at implementing a convolution filter on the empeg? Take a look, especially towards the end of the thread. at what these guys are doing with software on their PCs to *vastly* improve sound-quality in their homes. I was thinking it would be really slick to do the same thing in the car.

http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=283878

If your're referring to using a microphone to sample the audio and then apply an equalizer curve to compensate based on the sampled input, then yes, we have been talking about this sort of thing for a long time.

Do a BBS search on the term "RTA" to see what we've had to say about it.

Never mind, there's no "match whole word only" feature in the search software, so any post containing those three letters will come up. Perhaps somone less sleepy than I can link to some of the recent RTA threads?

I haven't actually read the thread he was talking about, but my guess is that he is interested in correcting in the time domain rather than the frequency domain. i.e. compensating for reflections as opposed to resonances. Kind of like the known bit pattern in the middle of a GSM timeslot, but you would only measure once rather than all the time.

Richard.

This was the last one, IIRC, where genixia gave the pot a good stir.

That convolution filter isn't going to work on the empeg. For those of you who don't want to read through the whole AVS thread, here is a gross simplification of the idea (If I have understood it correctly from my brief skim);

The convolution filter is using a finite impulse response filter to provide eq. One of the side-effects of the FIR is that is can also effect time shifting of it's output. This idea manipulates that effect to provice both equalisation _and_ time shifting on a continuously variable basis across the frequency range. This allows for both frequency-dependant level correction (ie traditional eq) and frequency-dependant phase correction. (phase errors can be introduced by most filters (including crossovers), distance, amps etc).
Obviously both of these can only be done for one sweetspot - a perfect level and phase at one location means non-perfection at another.
Anyway, the method relies on the processing power of modern CPUs to implement many FIR filters using DSP type algorithms. The filters are calculated from a closed loop system using a microphone at the listening spot to provide feedback, and then those calculated filters are run using a separate engine.
The problem is that DSP algorithms are usually highly dependant on floating point math, and these are no exception. And the empegs CPU, aside from being significantly less powerful than P4 or Athlons to begin with, has no floating point hardware. We can emulate floating point mathematics using a library, but it is slow. (Hence the reason why Ogg has only just made it to alpha - the Ogg project relied heavily on floating point math.)
I'll stick to 'simple' RTA. (Thanks for the kick...I've been kinda lazy about it recently.)

This allows for both frequency-dependant level correction (ie traditional eq) and frequency-dependant phase correction.

OH! He means BBE processing!

Yeah, you can get outboard BBE units for the car, no problem, I've seen them. No way to build that into the empeg, I'm afraid, but certainly do-able as an outboard unit.