So .... the empeg has a microphone input and lots of fancy DACs and DSPs and pretty graphics ...

A while back everyone was on a kick of flattening their audio profile in their cars with RTA software and fancy mic's.

What if you plugged in a decent mic' into the empeg and had it self-correct the EQ for your car. It KNOWS what it should sound like because it's playing it, right?

I'm sure it wouldn't be great, but it might get people *close* to overcoming some drawbacks in their car's environment.

Greg

I've brought this up before and was basically told that the mic input has a limited sampling range so it won't be able to do a good job. I still think it'd be neat for doing at least a limited range, but it seems like the RTA software is the best bet. (as long as they keep offering free demos)

Having actually done this now, I can say that the most critical corrections (at least on my system) were in the lower frequency ranges, and could in theory be handled by the mic input. Okay, one would have to still work on the highest stuff by hand, but it would certainly be a great start.

Since Genixia's already done some work with sending commands to the equalizer, I could see this being made to work. You could run a userland app to have it do the analysis and it could spit out a set of EQ settings that you could transcribe into the real equalizer.

There would be two huge stumbling blocks in the code however:

1) The analysis algorithm. Having worked with RTA software now, I can see that a given frequency doesn't simply "sit there", it jitters quite a bit and you need to do either averaging or peak-holding to get a decent reading. I found I got the flattest and most accurate results with peak-holding on my system.

2) The algorithm that decides what corrections to make to the EQ. With only X number of bands to use, it has to make compromises, it can't correct every tiny little thing. The trick is coding it to "know" what a given correction will do to the waveform, then do the best-fit algorithm. This is not going to be an easy task. Take this example frequency graph:



You tell me how to code an algorithm to flatten that using only 8 parametric EQ bands.

What if you travel down that frequency chart and average from your neighbors to remove the regional peaks. You'd still have the overall flow, but without the "noise" ... from here, bump the EQ to bring that flow up to more of a flat line (well as close as 8 bands and Q will get you).

Comments?

Greg

The whole point of an eq is to tweak the sound so that it sounds good (as opposed to look good) and what sounds good to one person's ears might not sound good to someone else's.
For example, older people cannot hear higher pitched sounds as well,
so they would tend to raise the eq settings for the higher frequencies.
Or if they had advanced hearing loss,
they would not be able to hear the highs, no matter how high the treble...
A younger person might turn on an older person's radio and say, "jeez where's all the treble?"
and raise the treble, while the old man who's car it is percieves no difference in the sound.

Well anyways, that's my 2 cents...

RTA is usually only the first step in EQing a system. The point is to get odd peaks and resonances corrected first, THEN adjust to taste.

Having done this, I can say that it's very useful. By correcting tiny problems in the EQ, I was able to get overall response sounding better because I was no longer killing the entire midrange just to get rid of one tiny little audible peak in the midrange.

I'm sure that's true for some ... but for me, I've got perfect hearing from 1hz to 100mhz ...

I've personally got to run my motherboard a little hot, as the 100mhz bus keeps me up at night with all that screeching. I had a Pentium that made my ears bleed a few years ago.

I always try to avoid listening to music around 6hz though ... causes some major bowel release.

Greg

Uh... Yeah. Right, what you said.