"analog" phase filter/EQ ?

bootsy · 3 Feb 2011

dasdeck wrote: hi there! i wonder how would i got about changing the phase response characterristics of filters. can i just use RBJ filters and alter/add some parts? i'd like to experiment a bit.

thanks!

jm

You need to change the order of the filters then. If just for experiments then maybe just start by adding an allpass to change the overall phase response. Dunno what your specific goal is ...

dasdeck · 4 Feb 2011

i played with some eqs that offer different phase modes (digital/minimum/linear/analog/etc.) and i really liked the "analog" one and the "minimum" one for high boosting on bus and master for example. it seems they shift the phase on high frequnzy content (?), as it sounds less glassy (delayd high frequencys?). the frequency response seems indetical in different modes, or at least it looked like it, lookign at the analyzer in reaper.

i'll try some stuff with allpass now, thanks.

jm

edit: ok, ive analysed the phase in the analyser. the analog phase thing looks similar like the one i get from an oversampled equalizer (x4, linear phase resampling filter). the phase curve is more open towards/above the nyquist, similar to the magnitude curve. i gueus i'll have to generate the FIR coefficients from the iir eq shape i want to reproduce. can you name some methods of doing it? is (de)convolution a good way to do go? are there free implemtentations to copy or learn from?

thanks 🙂

jm

MegaHurtz · 4 Feb 2011

They probbably used different filters, the liniar ones can be a bit cold. The best ones to give you those analog feelings are the Moog and the Pi filter, since these have a curved phase response. Else I guess its back to the stacking differentials to match your definition of ^T. Im currently getting kinda sweet and perfect results using approx 140 sines stacked together to make up the spectrum and the filter. Ofcurse 24% CPU.

dasdeck · 4 Feb 2011

i'm not shure i want the analog sound. i just want what i heard in this eq 🙂. it seems to me its like a biquad (bell) unwarped at the nyquist, since there is no indication an oversampling filter at work.

jm

edit: i should say that i have almost no clue what you are talking about 🙂. some hints to some documentation would be great 🙂.

cheers!

jm

MegaHurtz · 5 Feb 2011

one word.. biquad 🙂 two words.. Its very simple in code, so fun to alter!

dasdeck · 5 Feb 2011

so the "Moog and the Pi filter" are biquad coefficient calculation formulas? i use biquads all the time (with rbj coeff calculations). what i would like to know is, if/how i can unwrap the bell shapes in the high treble region (near nyquist) without oversampling, so that bells look (and sound) more symetric (like the one i described in the first post). i'm fine oversampling thought, as i'll probably couple the eq with some shapers, i'm just curious on how to approach this. maybe the other approach i analysed (form the OP) is with oversampling too, but one can not see the typical oversampling lowpass near nyquist (even the toolkit halfband filter has a slight roll off). i use a linear phase fir @ 20000 to oversample, as i find those very nice sounding. on th econtrary i find the halfband sounding "hard" in itself somehow. it may be because of thezerostuffing/upsampling side of filtering, i'll have to test this one day (and use hybrid up/downsampling for example).

thanks!

jm

MegaHurtz · 6 Feb 2011

I would dare go as far to say a biquad could do both (ish), Though in the SFV and Moog are different in principle. Where the Moog has a slightly better Q factor. Notably less leaking going on there. And the SFV has fuller resonance and lows. But the Q factor is terrible Especially at lower frequency`s. Paradoxically thats where it gets its "warmer" sound from. Come to think about it the resonace is due to leakage aswell. 🙂

Nowadays its popular practice to get a good impulse response going, maybe also the best way to do your eq. Bit like a liquid mix. I think these sound incredible for digital reproductions.