EQ

EQ

The speakers have parametric EQs organised in containers. Each container holds a fixed number of bands, and each parameter is stored as one array with one entry per band.

Source mark [measured]: see the legend on Home. Tested hardware: KH 120 II firmware 1_7_3, KH 750 DSP firmware 2_1_2, two units of each. Ranges, units and filter types on this page are what the speakers report about themselves via osc/limits; value types and array lengths come from reading every leaf.

Structure

A container lives under an output path, e.g. audio/out/eq2. Below it there are six leaves. Each leaf is an array with one entry per band.

Leaf	Value type	Range / options	Source
enabled	Boolean	–	[measured]
gain	float, dB	-99 … 24	[measured]
boost	float, dB	-99 … 24	[measured]
frequency	float, Hz	10 … 24000	[measured]
q	float	0.1 … 16	[measured]
type	string	PARAMETRIC, LOSHELF, HISHELF, LOWPASS, HIGHPASS, BANDPASS, NOTCH, ALLPASS, HI6DB, LO6DB, INVERSION	[measured]

• All six leaves exist as per-band arrays in every EQ container of both models, with the types above [measured].
• Both models report the same ranges and filter types for every EQ container, including the KH 750 crossover. Each limit carries count (the number of bands) and subscr: true, but no writeable key [measured].
• Every value read was inside the reported range [measured].
• bypass and order do not exist: audio/out/eq2/bypass and audio/out/eq2/order answer 404 address not found on both models [measured].
• All six leaves of audio/out/eq2 accept a set on both models: the test wrote back each leaf's current value, and the device echoed it [measured].

Writing one band

An array of the container's full length with null in every place but one changes only that band; the other bands keep their values. A shorter array answers 406 not acceptable and changes nothing [measured]. Tested on 2026-10-19 with gain on one band of every container of both models (KH 120 II eq2, eq3; KH 750 eq2, eq3, eq1/in1, eq1/in2 on out, out1 and out2): written, read back, other bands untouched, restored [measured].

-> {"audio":{"out":{"eq2":{"gain":[null,null,null,null,null,null,null,null,null,3.0]}}}}
<- {"audio":{"out":{"eq2":{"gain":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,3.0]}}}}

Writing one band this way cannot overwrite a change another client made to a different band in the meantime.

Request size: both models refuse a request of more than 29 JSON tokens with 413 request too long. A set of a 20-band leaf (eq3) is exactly 29 tokens; with an osc/xid it is 33 and refused, so such writes go without the xid. The six leaves of one band in one message are refused for the same reason: a band is always written leaf by leaf [measured]. Details: SSC-Protocol. Asking for a container instead of a leaf does not return its leaves: {"audio":{"out":{"eq2":null}}} answers 404 at the container, on both models [measured]. Error format: SSC-Protocol.

What gain and boost do

Measured on a KH 750 with a 50 Hz sine at input B and the Sub output meter, which sits after the EQ [measured]:

Band on audio/out/eq2	gain +6	boost +6	both +6
PARAMETRIC 50 Hz, Q 1 (on the tone)	+6.0 dB	+6.0 dB	+12.1 dB
PARAMETRIC 500 Hz, Q 4 (far from the tone)	+6.1 dB	0.0 dB	–

• gain lifts or lowers every frequency while the band is on, also for the pass types: a low pass at 200 Hz with gain +6 raised the 50 Hz tone in its pass band by 5.4 dB [measured].
• boost is the height of the filter (peak or shelf) around the band frequency [measured].
• At the band frequency both add up. Setting gain and boost to 0 makes a band flat [measured].

Filter shapes

Measured with full sweeps on a KH 120 II (REW and a calibrated measurement microphone; each setting of one band against the same band switched off, which removes room, microphone and speaker) [measured]:

type	What q means	Shape	Source
PARAMETRIC	cookbook Q	RBJ cookbook peak at the band frequency, height = boost; within 0.03–0.06 dB from 100 Hz to 8 kHz (Q 0.5, 1, 4; ±6 dB)	[measured]
LOSHELF, HISHELF	cookbook shelf slope S, not Q	RBJ cookbook shelf, frequency = midpoint, height = boost; q 1 is the steepest shelf without overshoot, q 2 overshoots slightly. Read as S: a few hundredths of a dB; read as Q: off by up to about 1 dB	[measured]
LOWPASS, HIGHPASS	cookbook Q	2nd order (12 dB/octave); within 0.07–0.10 dB (Q 0.5, 0.707, 1)	[measured]
LO6DB, HI6DB	unused	1st order (6 dB/octave), −3 dB at the band frequency	[measured]
BANDPASS	cookbook Q	constant-skirt form s / (s² + s/Q + 1): centre +20·log10(Q) dB (Q 2: +6.0 dB)	[measured]
NOTCH	cookbook Q	RBJ cookbook notch	[measured]
ALLPASS	cookbook Q	flat level; phase of the 2nd-order all-pass (−180° at the band frequency)	[measured]
INVERSION	unused	flat level; polarity inverted (180°)	[measured]

• Slope S and cookbook Q of a shelf: 1/Q² = (A + 1/A)(1/S − 1) + 2 with A = 10^(boost/40). For a steep slope at a large height the right side turns negative and the cookbook form ends; what the speakers do there is not measured.
• The EQ runs at 48 kHz: near the top (a peak at 14 kHz, a high shelf at 12 kHz) the digital cookbook biquad at 48 kHz matches clearly better than the analog shape — a peak there is narrower than its analog form [measured].
• A second session with the microphone moved repeated a shelf of the first within 0.03 dB [measured].
• On the KH 750's own woofer a peak at 50 Hz (Q 2, ±6 dB) matched the same cookbook shape, and an earlier single-tone check (50 Hz sine, Sub output meter, 21 cases) agreed within about 0.8 dB [measured].

Clicks

By ear on the KH 750 (50 Hz sine, low level) [measured]:

• No click when switching a band on or off, for gain or boost steps of 0.5 and 3 dB, a frequency step 50 → 60 Hz or Q 1 → 2.
• A slight click for a type change PARAMETRIC → LOSHELF.
• audio/out/mute on clicked slightly to clearly, mute off clearly to loudly: muting around EQ changes would add clicks, not remove them.
• Writing several leaves one after another can pass through large level jumps (for example a high pass still at its old frequency) and click loudly.

Layout per model

KH 120 II

Container	Bands	Source
audio/out/eq2	10	[measured]
audio/out/eq3	20	[measured]

The KH 120 II has no eq1: audio/out/eq1/enabled is rejected, and the device's schema lists only eq2 and eq3 [measured].

KH 750 DSP (reports itself as KH 750), firmware 2_1_2

Output	Drives	Container	Bands	Source
out	the KH 750's own woofer	audio/out/eq1/in1, audio/out/eq1/in2 (crossover)	2 each	[measured]
out	the KH 750's own woofer	audio/out/eq2, audio/out/eq3	10 each	[measured]
out1	output socket A	audio/out1/eq1/in1, audio/out1/eq1/in2 (crossover)	2 each	[measured]
out1	output socket A	audio/out1/eq2, audio/out1/eq3	10 each	[measured]
out2	output socket B	audio/out2/eq1/in1, audio/out2/eq1/in2 (crossover)	2 each	[measured]
out2	output socket B	audio/out2/eq2, audio/out2/eq3	10 each	[measured]

• Both KH 750 units gave the same results [measured].

• What the outputs drive [measured]: audio/out is the KH 750's own woofer, not the sum of all outputs: a flat +6 dB (gain) on a band of out/eq2 moved the Sub output meter by +6.1 dB and the meters A and B by 0.2–0.3 dB. out1 is output socket A and out2 socket B: +3 dB on out1/eq2 moved meter A by +3.1 dB, on out2/eq2 meter B by +3.1 dB, the other meters by at most 0.1 dB.

• eq1 is the crossover, one level deeper. eq1 has no leaves of its own: audio/out/eq1/enabled answers 404 [measured]. Its leaves sit under eq1/in1 and eq1/in2, 2 bands each, all six leaves [measured].

• eq1/in1 acts on input A, eq1/in2 on input B: +6 dB gain on an in2 band moved a tone on input B by +5.7 dB, on in1 by +0.5 dB [measured].

• Crossover settings read on both units (every band of in1 and in2 identical):

  Output	type	frequency	q	enabled	gain, boost	Source
  out	LOWPASS	80 Hz	0.7	true	0	[measured]
  out1, out2	HIGHPASS	80 Hz	0.7	true	0	[measured]

  These are the settings read on 2026-10-17; on 2026-10-21 the measured unit's input-B crossover had both bands switched off by its owner.

• Crossover shape [measured] (KH 750 woofer, input B fed with a receiver's LFE, channel B mode Ext.BM LFE (fullrange), rear bass management switch DISABLED): with both low-pass bands on at 80 Hz the response is −6.3 dB at 80 Hz, with one band −3.3 dB; the same at 40 Hz and 120 Hz. So the two 2nd-order bands with Q 0.7 in a row make a crossover of about 24 dB per octave (Linkwitz-Riley), as the cookbook shapes predict. It acts in this mode once its bands are on; with both bands off its frequency changes nothing. Far above the crossover frequency the measurement reached its floor (about −14 to −21 dB), so the deep stop band is not measured. When a receiver's bass management already filters the subwoofer feed, the KH 750's low pass comes on top of it.

• The crossover is write-tested (one band, restored; see Writing one band) [measured].

• KH 80 DSP, KH 150 and other KH 750 firmware versions: not tested on hardware.

See also: SSC-Paths · Device-Behaviour · Limits-and-Untested