Re: ToneShiftEQ 0.7.0 Released

Hermann Meyer <brummer--S0/[email protected]> Sat, 1 Aug 2026 17:09:38 +0200
Newsgroups gmane.linux.audio.devel,gmane.linux.audio.users
Message-ID <[email protected]>
Am 01.08.26 um 14:26 schrieb Fons Adriaensen:
> On Sat, Aug 01, 2026 at 09:31:18AM +0200, Hermann Meyer wrote:
>
>> ToneShiftEQ is a modern 12-band parametric equalizer with APO support,
> Some remarks:
>
>> Implement State Variable Filter mode (LV2/ stand-alone only)
> What is the point of having both biquad and state variable modes ?
> Both will produce the same responseif implemented correctly and
> given the same parameters. SV has some advantages in real life,
> and should be the preferred one.
>
> The actual SV code you use avoids high F,Q instability by using
> first order (trapezoidal) integration instead of zero order
> (just adding samples). It's a nice idea, but still an approximation.
> There is no reason to use this if an exact solution is available.
>
> See <http://kokkinizita.linuxaudio.org/papers/digsvfilt.pdf>.
> This is actually used in zita-eq1.
>
>
>> Show accurate Filter-curves in Biquad mode
> Except they are not, also not in SV mode.
>
> Try this two parametrics (in SV or BQ mode):
>
>   500 Hz, -30 dB, Q =3D 10
> 1000 Hz, -30 dB, Q =3D 10
>
> The measured response shows -4.5 dB halfway between
> the two dips (at 710 Hz). Toneshift shows something
> close to -14 dB (difficult to read exactly). Also
> the individual responses of the two filters are wrong.
>
> Errors are smaller for less extreme settings, but that's
> no excuse.
>
Which tool do you use to measure that?

ToneShiftEQ use the same response calculations for the UI and the dsp,

https://github.com/brummer10/ToneShiftEQ/blob/main/EQ12/IRMatch/BiquadResp=
onse.h

https://github.com/brummer10/ToneShiftEQ/blob/main/EQ12/IRMatch/SVFRespons=
e.h

>> APO support.
> This 'Equaliser APO' [1] seems to be one of the most hyped
> things in Windows audio circles. It has 15 or 31 parametric
> bands. There are entire websites and GIT repositories
> offering presets for this thing, each one 'professionally
> optimised' for one particular type of music or 'mood'. As
> if that would make sense. It's hard to find more refined
> pure liquid BS. Which one should I used when listening to
> 'reggae' while in a 'worship' mood ? Both in series ?


There are also profiles out there for speakers and headphones.


>
> * Some more general musings - not related to ToneShiftEQ in
> particular.
>
> Why do 'modern' equalisers need that many sections ?
> I could answer that question, but most readers here
> won't like it, so I'll shut up.

You could still disable any Band you don't need and use just 4 or 3 bands.


> Why do the GUIs of 'modern' equalisers look like the
> dashboard of a cheap Chinese E-car instead of being
> a bit more functional and ergonomic ? Same answer.

That is my taste, I like it this way. Sure, it wont be anyone's darling,=
=20
but, for me it looks nice and gives me the information I need.


> The famous big analog mixing consoles from last century
> (Neve, SSL, Harrison, Cadac, Midas, MCI, ...)  typically
> had 3 or 4 section EQ. Some of them just had a choice of
> 3 frequencies for each band and fixed Q. Still these were
> used to produce some of the most magnificent music recordings
> in history. Of course no FR display, sound engineers didn't
> need it, they just listened.
>
> I remember using Neve, Harrison and SSL, but don't recall
> ever needing more than 3 sections.
>
>
> [1] APO means 'Audio Process Object', sort of a plugin
> standard for the Windows 'system audio' chain. Then why
> is 'Equaliser APO' used as the actual name of one particular
> EQ ?


However, please let me know which tools you use for your measurements,=20
I'll try to reproduce it and find the culprit to fix it.