kbd to filter ...play the filter resoance ..not quit there yet

gentleclockdivider · September 23, 2015, 6:30pm

UNable to play the Analog filter’s frequency chromatically over the whole keyboard range

INstrument filter or effect filter …

Altough the digital filters do not self oscilate , they suffer from the same problem , inaccurate pitch to frequency conversion.

I used a one shot decaying impulse sample , then tweaked

the kbd tracking device to control the filter’s cutoff in the effects section …it’salways a bit off.

For both the comb filter and the ring mod, their highest freq.is 7800 hertz , for the ring mod the lowest frequency is around 8Hz ,the default setting of the kbd (c0-b9) tracks it along the whole key range , but these settings don’t work for the analog filters

So different approach :

Kbd tracking device set to A3=220Hz ,A6=1760HZ …

TRiggering A4 gives us 490 Hz , that’s just ridiculous …see screenshot

And we’re talking about a limited 3 octave range here and allready the scaling is way off

Reducing the octave range A3 to A5 …triggering A4 gives us 470Hz .

Should I say more…

For the polyphonic filter it is even worse

g

For the polyphonic filter setting cutoff to 0.820 = 7800Hz=B9, multiplied by Kbd device default setting , stil No luck.

Check out the patch 'compare ’ left channel is keytracked ring mod , right channel is noise into keytracked poly filter …

The best results I got for the polyphonic filter was a bit more than a one octave range

There’s definately something wrong with the pitch to freq. scaling …

At least a few octaves would be more than normal , too bad these new improved filters can’t even do this, zdf or not .

I would be happy if someone could show me a working playable setup.

Also the frequency read out of the ring mod is wrong …A4 =440Hz and not 220

thanks

https://app.box.com/s/ds96ipzkwextt5ywdb2h9tjuqzdyj3j2

https://app.box.com/s/qfch4muuq1wnwhgb6o7bqvrgamxdq19n

,

TheBellows · September 23, 2015, 7:08pm

It has always been like this i believe, that’s why we had to use the custom LFO to get the correct notes on the various tuned devices. It’s been fixed on the ringmod and combfilter, so i don’t see why it shouldn’t be like this for the other effects too. Would definately be very useful.

I wonder if backwards compatibility is the problem perhaps?

taktik · September 23, 2015, 9:00pm

Yes, the filter’s cutoff parameter is not using a note frequency scaling but uses a custom one, cause it fits better here. Also, if it would, it would be limited to Renoise’s 0-119 MIDI note range, which is ~8 to ~7902 Hz.

gentleclockdivider · September 24, 2015, 7:27pm

so which means no accurate kbd tracking to filter .

Serious ???

OopsIFly · September 24, 2015, 8:20pm

Lol it would be really cool…if the renoise devs would just give out the accurate formulas for every devices frequency or delay parameter range. Hey, this is not open sourcing any secrets, just enabling users to actually use your stuff in better ways. Then, one formula device would be enough to translate from a keytracker, and could even have stuff like real sensitive detune in it. Please, just do, do, do it - and also for the legacy devices and stuff.

What is holding you back? Some sense of humor, that some day one real nerdy geek will reverse assembly and post shit on the forums? Or waiting for some math geek doing number crunching on the parameters to extract forumulas? Or those measley custom lfo buissiness, by lua parsing parameters to-and-fro in inaccurate text string definition? It just sucks. Just give us the formulas. For real.

taktik · September 24, 2015, 9:01pm

Not a secret at all: See https://forum.renoise.com/t/2-7-more-tuned-devices/32280

OopsIFly · September 25, 2015, 9:42am

Those are pre-set LFO-tables with the values for each note “hardcoded”, I’d like to have direct and exact formulas for slider range 0.0…1.0 -> Freq_Hz/Delay_MS to work out metadevices myself (I bet other users would appreciate that, too). You have the code for each dsp device, in the code there’s translation from slider to Hz or Delay (for displaying the parameter value), so it would mean little work for you to just copy&paste the formulas/algorithms for this into the forum, and all will be happy. The LFO tables have their backdraws, and are hard to create. I.e. detuning or bends for frequency based parameters are hard to get exact with LFO’s - with the lots of points in them the LFOs seem to get “steppy”. Just not to say with such formulas you could easily create very custom LFO devices, with custom note ranges and or tunings. The way it was done until now involved some “brave hackery” i.e. a lua script to sample the frequencies of the device and read back a parameter value? This aren’t the true formulas for me.

ffx · September 25, 2015, 9:50am

I agree here with oops fly, some kind of note-to-freq translator device would be very helpful, so an octave is really an octave on every filter. Gova already built some kind of value grabber-translation-table generator, but it still requires to interpolate in between for the cost of a parameter delay. Or if you self would provide a formula for the more recent filters… Just which curve is used here and there for cutoff.

taktik · September 25, 2015, 10:01am

Well, sure thing.

Most device’s filters frequencies are scaled by:

(pow(10, log10(s)*x) - 1.0) / (s - 1.0), x=0 to 1,s = 100

Plot for s=100:http://www.wolframalpha.com/input/?i=plot%3A+%28pow%2810%2C+log10%28100%29*x%29±+1.0%29+%2F+%28100-1.0%29%2C+x%3D0+to+1

This includes the Filter3 device, all new filter devices in Redux, filters in Chorus, Flanger and old Comb devices.

Comb in Redux & Renoise 3.1 will use note frequency scaling.

LofiDevice uses the same formula withs = 16, old Ringmod device s = 20000.

Again Ringmod in Redux & Renoise 3.1 will use note frequency scaling instead.

So no “LFO tables” will be required for the RingMod or Comb in Renoise 3.1 - Redux. They can be directly connected to a KeyTracker. Same is true for AM filters in Renoise 3.1 and Redux’ instrument section. For filters you will need a map, for the reasons mentioned in my previous post above. DBlue’s LFO map fromhttps://forum.renoise.com/t/2-7-more-tuned-devices/32280 will do the job here.

Let me know if you need any others.

ffx · September 25, 2015, 10:07am

Thanks

OopsIFly · September 25, 2015, 5:56pm

Many big thanks for the formula. How cool is this!

is the formula for the eq points the same as for the filter3device cutoff? Haven’t tried yet, but the graph alwas seemed a little more stretched in the bass freqs, well, I dunno.

Now we can do it like this, put keytracker into a formula device instead of that lfo table, and do stuff like accurate pitch bending filters:

Click to view contents

<?xml version="1.0" encoding="UTF-8"?>

WhatAName

true

true

#Filtertracker

false

1.0

Device only

frq2flt(kt2frq(A, B))

local fromnote = 0.0

local tonote = 119.0

local bendrange = 12.0

local a4 = 57.0

local a4f = 440.0

local ffac = 1.0 / 22050.0

local lgfac = 1.0 / 4.6051701859880913680359829093687284

local trt = 1.0594630943592952645

function frq2flt (inp)

return log(99.0*(inpffac)+1.0) * lgfac

end

function kt2frq (inp, bnd)

local nnbr = inp(tonote-fromnote) + fromnote + (bnd-0.5)bendrange2.0

return a4f * pow(trt, nnbr-a4)

end

Key

Bend

-

true

1

0.277310938

Device only

0.5

Device only

0.0

Device only

-1

Device only

-1

Device only

-1

Device only

ffx · September 25, 2015, 6:51pm

Thx OopslFly! Very helpful.

[SPOILER]

<?xml version="1.0" encoding="UTF-8"?>

true

key tracked filter 2

false

1.0

Device only

50

Device only

Bend

2

0.0

1.0

Linear

46.8000145

Device only

Rez

0

3

0.0

1.0

Linear

0.0

Device only

Inertia

0

5

1.0

0.0

Linear

100

Device only

Key range

1

4

0.5

0.0

Linear

1

5

0.5

1.0

Linear

50

Device only

Key Offs

2

3

0.0

1.0

Linear

0.0

Device only

Key Offs

50

Device only

Macro 7

50

Device only

Macro 8

5

true

Init

true

Init

true

1.0

Device only

2

Device only

0.508457482

Device only

6.55200195

Device only

-60

Device only

0.0078125

Device only

Butterworth 4n

keyfol

false

Init

true

1.0

Device only

-1

Linear

Clamp

0

119

-1

Device only

2

Device only

1.0

Device only

0.0

Device only

1.0

Device only

key2freq

false

note-freq-helper-filter3

true

1.0

Device only

frq2flt(kt2frq(A, B, C))

local fromnote = 0.0

local tonote = 119.0

local bendrange = 12.0

local a4 = 57.0

local a4f = 440.0

local ffac = 1.0 / 22050.0

local lgfac = 1.0 / 4.6051701859880913680359829093687284

local trt = 1.0594630943592952645

function frq2flt (inp)

return log(99.0*(inp*ffac)+1.0) * lgfac

end

function kt2frq (inp, bnd, offset)

local nnbr = (inp+offset-0.5)*(tonote-fromnote) + fromnote + (bnd-0.5)bendrange2.0

return a4f * pow(trt, nnbr-a4)

end

Key

Bend

Key Offset

true

1

0.70588237

Device only

0.5

Device only

0.5

Device only

-1

Device only

0.0

Device only

2

Device only

[/SPOILER]

Akiz · September 26, 2015, 8:27am

Can somebody create REDUX example preset please?

OopsIFly · January 1, 2016, 3:08pm

Hello!

Now that 3.1 is out, I wish to resume my work on making formula device keytrackers for the new filters etc. - Yay, the formula device is official now?

I also want to try to document my stuff in this forum, and share the formulas for each device.

There are some points I’m still missing, maybe @Taktik can help clearing up these formulas.

The new analog/digital filter has a different scaling than the old filter device. It also seems to have a different frequency range. Please share the formulas for it, exp the new analog deems to bloom when keytracked, when self-oscillating! -

hlrf · January 1, 2016, 7:38pm

I did some measurement of key-tracking the new analogue filters (by playing a white noise sample thru instrument filters, with a single keytrack device), the resonant peaks seem to be about 3 octave higher than the note played.

roughly,

C2>D#5

C4>B6

C6>E8

OopsIFly · January 1, 2016, 9:46pm

Ah, sorry for my post. Now I feel just as dumb again as when I was first presented with the formula. It’s been a while since I found how to use the formula properly, and I lost my original document and just kept the resulting formula device code. I just had forgotten - you need to adjust the formula for the frequency range of the plugin, damn…

Now I have repeated the steps and found it works perfectly with the new analog filter device. Will post the corrected formulas in the other thread soon.

gentleclockdivider · January 7, 2016, 9:00am

well the formula device is of no use when we want to play the POLYPHONIC FILTER’s frequency , that’s the filter in the instrument 's tab guys !

OopsIFly · January 7, 2016, 5:23pm

Oh, not mentioned yet, here a Thread with lots of devices: http://forum.renoise.com/index.php/topic/45617-renoiseredux-slider-translation-formulas/

@ gentleclockdivider, I haven’t investigated if some known formula for keytracking the modulation filters could be valid. There aren’t all formulas know yet, anyways, and we have to wait for Taktik to disclose them to know for sure. Or try to do some real math reverse engineering via datasets to come close. So far I have only looked into the DSP devices. When I find some time & peace of mind I’ll experiment with the known formulas and the modulation filters.

Controlling the modulation filters via a keytracked formula device has some downsides tho I think. The tracking won’t be polyphonic, I think this couldn’t be possible so far. You could only track a single offset for all currently played notes, but this could be independet of the notes pitches via some tricks, so it might be useful after all. As the keytracker isn’t polyphonic, and the formula device would also have no means to distinguish between the notes, there can only one value be sent to the modulation filters. Polyphonic accurate keytracking would need either a note-correct frequency control range like with the new comb and ringmod, or a polyphonic formula device in the modulation tab.

You can access some of the modulation filters as well via the chorus/flanger devices though. Just use them without any modulation and 100% wet. Most of the new (3.1) filters’ frequency formulas are known, look at my other thread linked in this post.

OopsIFly · January 9, 2016, 2:14pm

I’ve looked at it, the new 3.1 modulation filters respond to the same known formula as the new analog filter / chorus filters. But the limitations apply - sadly no way for polyphonic keytracked action.

toimp · May 14, 2020, 4:50pm

It seems this is still the case, right?

Its nice that the filter is polyphonic, but its out of tune, not sure how i can fix this

I tried to replicate this tip in Renoise: