I don't know what a real drum set cymbal crashing looks like on a graph, but I have a strong feeling it isn't flat.
Wait, what?
Think of a shape like a
U
Now imagine a speaker that takes that first line out... "a curved response", similar to a speaker with no bass That speaker gives you a
J
And the J might sound exactly like parts of the U, but like the whole U.
Now think of a bose speaker trying to make that U
L_K
They didn't even try, but some people think it sounds cool when they hear the demo 2 feet away with jurrassic park sounds that are loud
Some speakers get closer though
|J
and closer
(_)
But you just want a speaker that does this, right?
U
That has nothing to do with whether the original U was flat. It just means you're not taking away or adding anything to the original U. It's the ideal speaker only assuming an ideal recording of the letter U. Boy, this feels like sesame street.
Obviously sound is more complex than the letter U, but it's still measured by a mic in a way that can be represented by electrical signals. The two toughest parts are going to be a mic that can represent what was going on and a speaker that can represent what the mic was recording. The rest is all stuff you'd learn in a second year undergrad EE course or something (boooooooringgggg
)
I think what's happening here is a difference between power response (which is what we have of very few speakers but should really have of most as it's a huge indicator of what we hear) and on-axis response (which is basically all we really have of most speakers including the quarts) is being mitigated. There's so much to sound than just a waterfall and a frequency response, although that basically isolates the really bad and the "good enough to bother with", it doesn't mean measurements are getting something wrong. There just isn't enough of them.