I have compared two kinds of correction : full-band EQ from listening position vs flat speaker.
I use Neumann KH-120 speakers, in an average room, at a distance of 2.1 meters. Here is the spinorama of the KH-120.
And here are the measurements from my listening position, with no correction, and with my usual correction :
As you can see, above 800 Hz, the speakers are uncorrected. Which means that, according to the spinorama, their frequency response is completely flat.
I am not sitting right in their acoustic axis, but rather 12° aside and 7° above.
This correction sounds perfectly balanced to my ears. There was no target curve for the correction below 800 Hz, though. Everything have been progressively adjusted by ear, week after week, on various musical styles.
Now, here is a full band correction calculated for me by a professional, according to the measurements that I sent him. He chose a target curve according to the type of speakers, the listening distance, the room size, and the decay time measured from the listening position.
We can see that it is not as accurate as my own correction in low frequencies, because this is a FIR filter limited to 6000 taps.
We can see also that above 1600 Hz, it is completely flat from the listening position.
However, the curve that we can see above is too high above 1600 Hz.
Since we are going to compare the two corrections, let's try to match their levels.
We can see that if we ignore the 54 Hz peak, and try to match as well as possible the levels below 1000 Hz, then they clearly mismatch above.
The audible result confirms this : the full band correction sounds brighter than my correction (and we can also hear that the low frequency is uncorrected). Impossible to compare them beyond this. It doesn't sound more or less neutral, it just sounds brighter.
Let's apply a high shelf correction so that the two results have the same overall tilt :
Now we can compare the two methods ! Does it sound better with the speakers completely flat (blue curve), or completely corrected from the listening position (purple curve) ?
I listened to one of my reference tracks, the part Quia Respexit, that begins at 5:28 in this video :
With the level correction, the brightness of the full range correction is eliminated, and they now both sound well-balanced.
My first impression is that the blue correction sounds a bit more realistic than the purple, full band one. However, I have been listening to the blue correction for so long that I might just have become used to it.
I decided to perform an ABX comparison between the two versions.
Before that, I started to ABX an intermediate version of the full band correction, that has a -1.2 dB correction instead of the final -1.6 dB correction. The difference between these two seemed more important, and thus easier to ABX. Then I would ABX the two real corrections.
Unfortunately, I failed this preliminary test !
I nonetheless loaded the two final files in the ABX comparator. Comparing two completely different corrections might, after all be easier than comparing two versions of the same one.
But comparing A and B, the main difference that I can hear between the full-range correction and the flat speaker correction is in the low-mids, where the two corrections differ slightly. Now I don't hear any significant difference in treble !
After all this, I'm not sure anymore that the blue correction sounds more realistic than the purple one. I could try to ABX them relying on the differences that I can hear in the low-mids, but it would be pointless.
My conclusion is that the choice between a full range correction or a correction that makes the speaker's anechoic response flat doesn't make much difference.
Creating a room correction, besides the obvious low frequency problems, the first and hardest difficulty is to find a target curve that sounds right. Choosing to equalize the speaker alone first doesn't help. At best, it can show you the overall tilt from 1000 to 10000 Hz, but here, this tilt was right to begin with in the full range correction.
Balancing the 100 - 800 Hz range vs the 1600 - 8000 Hz range is another matter. It is a very difficult question for whatever kind of correction.
In comparison, choosing between a complete correction from the listening position or a flat speaker uncorrected above 1000 Hz seems quite unimportant... at least with the Neumann KH-120 monitors... 2 meters away... in an average room.
