somehow testable by starting on top and bottom:
I hear height diference. and the model is not made for me, so I this can perhaps work
It works for moving sound that is already there, by chancing the frequency distribution of this sound while playing. You hear the sound, then fr. is changed, you hear it moves up.
It does not work to move the perceived location of the sound source from one speaker. There is no reference for the sound playing, so changing its frequency distribution before anything is playing does not move it up or down.
Perceived location in height from a speaker depends on reflections from nearby surfaces, its location and its reproduction properties. When the speaker itself has the inherent property of not being localized, and nearby reflections are attenuated sufficiently, other cues like seeing a person on the screen talking places the sound firmly on the screen, even if the speaker playing is placed at a different height.