Okay, what's this about? The users sai'ke, RCJacH and SamueleP are working on a JSFX clone of Sonnox Oxford Inflator: RCInflator.
So far the singleband action is ready and nulls against Sonnox.
Here is the latest script revision by sai'ke:

Code:

desc:RCInflator

slider1:0<-6, 12, 0.01>Input (dB)
slider2:0<0, 100, 0.1>Effect (%)
slider3:curve=0<-50, 50, 0.1>Curve
slider4:clip=1<0,1,1{Off,On}>Clip
slider5:0<-12, 0, 0.01>Output (dB)

@init

@slider
in_db = exp(0.11512925464970229 * slider1);
wet = slider2 * 0.01;
dry = 1.0 - wet;
out_db = exp(0.11512925464970229 * slider5);

curvepct = 0.01 * curve;
// 1 + (curve + 50) / 100
curveA = 1.5 + curvepct;
// - curve / 50
curveB = -(curvepct + curvepct);
// (curve - 50) / 100
curveC = curvepct - 0.5;
// 1 / 16 - curve / 400 + curve ^ 2 / (4 * 10 ^ 4)
curveD = 0.0625 - curve * 0.0025 + (curve * curve) * 0.000025;


@sample
function apply_waveshaper(in)
local(s0, s0_2, s0_3, out)
global(curveA, curveB, curveC, curveD,
       clip, dry, in_db, out_db, wet)
(
  in *= in_db;
  s0 = abs(in);
  clip ? (
    s0 = s0 > 1.0 ? 1.0 : s0;
  );
  s0_2 = s0 * s0;
  s0_3 = s0_2 * s0;
  s0 = (s0 >= 2.0) ? (
    0.0
  ) : (s0 > 1.0) ? (
    2.0 * s0 - s0_2
  ) : (
    curveA * s0 + curveB * s0_2 + curveC * s0_3 - curveD * (s0_2 - 2.0 * s0_3 + s0_2 * s0_2)
  );
 
  out = sign(in) * s0 * wet + min(max(in * dry, -2 * dry), 2 * dry);
  out = clip ? max(-1, min(out, 1)) : out;
  out * out_db
);

spl0 = apply_waveshaper(spl0);
spl1 = apply_waveshaper(spl1);

Now it's about the splitband action. And we are completely at a loss. What exactly does Splitband do? The manual states that the signal is divided into 3 bands, but the plugin has 0 spls PDC. So a linear phase filter is probably out of the question. However, we could not detect any phase distortion either, which again makes minimum phase filter very unlikely. Sai'ke says the following:

So we are completely lost. Does anyone here have any information or ideas about what exactly the splitband action actually does?


Knowledge desired!

 

Last edited: Aug 16, 2021

Nice plugin

 

Inflator tutorial

Show And Tell Of Sonnox Oxford Inflator

 

I'm not a coder; but if it's not actually splitting the signal and processing separately regions like a multiband, perhaps they are using either a simple tilt or a fixed frequency emphasis-deemphasis curve that is weighting the processing into 3 basic areas (that they are presenting as "bands".)

Just a guess, if you're not seeing any add'l latency or phase rotation.

 

hmm... that would - to my knowledge - induce either phase distortion or delay. A tilt filter is just that: a filter.

 

There's defo a band splitting in Band Split mode.
When switched off, all freqs in Plugin Doctor have the same output. When switched on, this is no longer the case.
Also, you can easily hear the different shaping between off and on.

 

inflator uses 3 bands with fixed frequencies and if the limiting and/or its gain up do hardly work, you won´t see much distortion of the phase spectrum around the split frequencies.

i bet that that it is a 6 db butterworth (and not a bessel or something), which noticeably works on the phase spectrum when it is not symetrical, and you should see some minor effect when you try more extreme input material, so that the limiter has something to do.

note that white noise or sinewaves not neccessarily the best test tones to find out about transfer functions - and or course not even a spike might show something, when there is no VCAs working behind the bands.

 

Last edited: Aug 16, 2021

now i forgot the most important thing.

of course you can always measure the phase spectrum of a filter even when it does not do anything, so my last sentence was nonsense.

but the frequency dependent processing here most likely is implemented parallel, like in some EQs. so when there is no processing...

 

Of course there is bandsplitting. the question is how is this implemented?

Limiter? Which limiter? Do you mean the Clipper?

 

no i mean the gainup in the plug-in´s main function itself.

if the bandsplit happens parallel, which i assume, you will only notice the change of the phase spectrum by the filters if the bandsplitted signal comes through somewhere. and the filter topology used might still be hard to track when comparing in against out