directional_eq

NODE

Lifts the band the ear reads as height, on the side component only, so the centre keeps its tone.

Parameters

FIELD	WHAT IT DECIDES	RANGE	REQUIRED
depth_cut_db	Frontal depth dip gain on Side channel [dB] (-6.0 to 0.0 dB, 0.0 = bypassed).	-6 〜 0	no
height_freq_hz	Elevation peaking center frequency [Hz] (Blauert height band: 4000 to 14000 Hz).	4,000 〜 14,000	no
height_gain_db	Elevation peaking gain on Side channel [dB] (0.0 = bypassed, up to 12.0 dB).	0 〜 12	no
height_q	Elevation peaking filter Q factor (0.5 to 5.0).	0.5 〜 5	no

{
  "depth_cut_db": -2,
  "height_freq_hz": 8000,
  "height_gain_db": 3,
  "height_q": 2,
  "type": "directional_eq"
}

How it is computed

M = (L + R) / √2                S = (L - R) / √2

1. peaking on S at height_freq_hz, height_gain_db, height_q
   (around 8 kHz, the band associated with perceived elevation)
2. optional dip on S at 3.5 kHz, q = 1.0, by depth_cut_db

L = (M + S) / √2                R = (M - S) / √2

Blauert measured that certain bands make a sound seem to come from above or in front regardless of where it actually is. Applying them to the side component only leaves anything centred, a voice or a lead instrument, with its tone intact.

Using it

It acts on ambience, not on the vocal.
Because only the side component is filtered, what moves is the reverb and the room around the mix, which is usually what you wanted to raise.
The dip trades forwardness for depth.
Cutting around 3.5 kHz softens how far forward the image sits. Used with the lift it reads as a taller, deeper space rather than a brighter one.
It is not loudness matched.
Like the other spatial stages it counts as 0 dB in an A/B comparison, so lifting a band here can make a slot sound louder. Compensate with a gain node when you compare.