Skip to main content

Lab · Biquad design

Computed

High shelf filters, corner by corner

Lifts or drops everything above f0 by a fixed amount and leaves the bottom flat. Every ISO third-octave centre from 20 Hz to 20 kHz, solved at 48 kHz and seven other sample rates.

Every corner, with what changes across the sweep

The pole radius column is the one worth reading down. It rises toward 1 as the corner drops relative to the sample rate, and it is what decides the last column: the 5 corners at and below 50 Hz do not survive 16-bit fixed point at 48 kHz.

CornerAt f₀Pole radius @ 48 kHz16-bit fixed point
20 Hz+3.00 dB0.997802diverges
25 Hz+3.00 dB0.997254diverges
31.5 Hz+3.00 dB0.996541diverges
40 Hz+3.00 dB0.995609diverges
50 Hz+3.00 dB0.994515diverges
60 Hz+3.00 dB0.993421holds · ≤ 1.303 dB
80 Hz+3.00 dB0.991238holds · ≤ 0.732 dB
100 Hz+3.00 dB0.989060holds · ≤ 0.053 dB
125 Hz+3.00 dB0.986343holds · ≤ 2.484 dB
160 Hz+3.00 dB0.982553holds · ≤ 1.923 dB
200 Hz+3.00 dB0.978239holds · ≤ 1.157 dB
250 Hz+3.00 dB0.972874holds · ≤ 0.754 dB
315 Hz+3.00 dB0.965943holds · ≤ 0.011 dB
400 Hz+3.00 dB0.956955holds · ≤ 0.285 dB
500 Hz+3.00 dB0.946489holds · ≤ 0.004 dB
630 Hz+3.00 dB0.933057holds · ≤ 0.025 dB
800 Hz+3.00 dB0.915786holds · ≤ 0.011 dB
1 kHz+3.00 dB0.895887holds · ≤ 0.049 dB
1.25 kHz+3.00 dB0.871640holds · ≤ 0.003 dB
1.6 kHz+3.00 dB0.838843holds · ≤ 0.020 dB
2 kHz+3.00 dB0.802961holds · ≤ 0.001 dB
2.5 kHz+3.00 dB0.760439holds · ≤ 0.002 dB
3.15 kHz+3.00 dB0.708912holds · ≤ 0.006 dB
4 kHz+3.00 dB0.647725holds · ≤ 0.004 dB
5 kHz+3.00 dB0.584520holds · ≤ 0.000 dB
6.3 kHz+3.00 dB0.516530holds · ≤ 0.000 dB
8 kHz+3.00 dB0.452761holds · ≤ 0.001 dB
10 kHz+3.00 dB0.416689holds · ≤ 0.000 dB
12.5 kHz+3.00 dB0.430322holds · ≤ 0.000 dB
16 kHz+3.00 dB0.534310holds · ≤ 0.001 dB
20 kHz+3.00 dB0.731048holds · ≤ 0.001 dB

The other seven types

Method. RBJ Audio EQ Cookbook coefficients at Q = 0.7071 and +6 dB, normalised to a₀ = 1. Pole radii are the roots of the denominator quadratic, not √|a₂| — the two disagree exactly when coefficient rounding pushes the poles onto the real axis, which is the case this table's last column is about. Every figure is solved; nothing is fetched.