jayrope

coming across the cross~ filter object i have a more detailed as well as a broader question:

1. i can't find info on the frequency response of the cross~ object within the tutorials. anyone?

2. more basic: which filters do you prefer for both sound quality, e.g. frequency response, and/or musicality?

personally i tend to rather softly filter audio input/output (or intermediately) with a highpass to decrease effects of unwanted dc offset as well as leave more resolution to the representation of actually music-relevant frequency. 

over the years of traditionally using DAWs of almost any known kind i have found, that highpass filters in audio inputs, single channel strips, and even audio outputs, carefully set, seem to oftenly positively affect rather musical factors, like the representation of the body of the signal, it's positioning in the stereo field, it's openness, generally the clarity and definition of the individual signal as well as the overall sound of the mix, due to the reduction of unnecessary calculation of (at last) electrical energy, the one which turns out irrelevant for the musical result, but hinders a good sound, when resolution, as in all digital sound processing, is limited.

as rather a newbie in max/msp and still evaluating the depths i am quite content with it's sounding quality in the first place myself. nevertheless i would love to hear about your expertise in especially this application-dependent field, both musically and technically, left alone sample rates, converter quality, clock jitter or issues related to outboard equipment, miking or similar.

please excuse my "chain sentences" above. i only stick to the qualities of expression of "Ulysses" here, i suppose... curious to hear your expertise on max/msp-internal sound-relevant (filter) strategies.

personally i tend to rather softly filter audio input/output (or intermediately) with a highpass to decrease effects of unwanted dc offset as well as leave more resolution to the representation of actually music-relevant frequency. 
over the years of traditionally using DAWs of almost any known kind i have found, that highpass filters in audio inputs, single channel strips, and even audio outputs, carefully set, seem to oftenly positively affect rather musical factors, like the representation of the body of the signal, it's positioning in the stereo field, it's openness, generally the clarity and definition of the individual signal as well as the overall sound of the mix, due to the reduction of unnecessary calculation of (at last) electrical energy, the one which turns out irrelevant for the musical result, but hinders a good sound, when resolution, as in all digital sound processing, is limited.

as rather a newbie in max/msp and still evaluating the depths i am quite content with it's sounding quality in the first place myself. nevertheless i would love to hear about your expertise in especially this application-dependent field, both musically and technically, left alone sample rates, converter quality, clock jitter or issues related to outboard equipment, miking or similar. 

4-63-sound-quality-of-standard-filter-objects

short answer: try them all out and see which ones you like.

longer answer: most people complain about the filter externals because they exspect too much from them. i only use two, slide and biquad. slide works nicely as basic dsp building block in many situations. for a highpass just substract slide from the original. biquad is the only filter which musically as well as technically counts as proper audio filter. its drawback is that you first have to learn how to use it. use biquad with filtergraph before you start using it with coeff or custom math code, otherwise you will get lost completely. filtergraph.help will also give you an idea when and why to cascade multiple biquad objects. i recommend to use line(msp) with filtergraph because biquad can explode quickly when frequency input is moved too fast in the bass range. good luck! :)

Quote: Roman Thilenius wrote on Tue, 09 December 2008 04:41

----------------------------------------------------

>> biquad is the only filter which musically as well as technically counts as proper audio filter.

I'm not really sure what you might mean by that. In what way are onepole~/lores~/svf~/reson~ not musically or technically not proper audio filters, or do you simply mean that slide~ is not primarily designed as an audio filter?

For my money svf~ provides me with the kind of results I'm currently looking for (over biquad~), plus the flexibility of four different modes in parallel, but of course it really depends what you're trying to do / what sound you want.

Quote: Roman Thilenius wrote on Tue, 09 December 2008 04:41
----------------------------------------------------

>> biquad is the only filter which musically as well as technically counts as proper audio filter. 

of course i tried most of them already. i was basically trying to ask for other users opinion on the musical & technical qualities of the different filter types. i'll check out biquad again, definately will use filtergraph for it's manual touch (with a wii remote).

alex: i guess the max frequency limit of svf~ doesn't bother you?

i also want to ask again for info on the frequency spectrum available in cross~.

p.s. 110> long time no speak. fired up icq, momentarily doesn't really work. rrrr.

anyone else's opinion? 
i also want to ask again for info on the frequency spectrum available in cross~.

jrp
p.s. 110> long time no speak. fired up icq, momentarily doesn't really work. rrrr.


Quote: jayrope wrote on Tue, 09 December 2008 07:03

> alex: i guess the max frequency limit of svf~ doesn't bother you?

I hadn't paid much attention to that to be totally honest, although I dimly remember reading that part of the help file at some point. But still, for what I'm doing now it's not a major issue, as I don't need to filter up that high. I suppose if I was really bothered I might consider upsampling in a poly~ by a factor of two which would give the same range as the audio outside the poly~, or by implementing some other form of oversampling. Thanks for bringing it to my attention.

FYI: 2up.svf~ oversamples by a factor of two, and hence is not quite so cpu efficient, but I believe will remain more stable at higher frequencies. However, after messing around with it for a while, and recompiling it with various modifications (and for some rather boring technical reasons), I was unable to get it sounding like the msp version (which I preferred). I didn't manage to figure out why this was, but I solved my problem a different way, and so went back to the msp version.

ANyway, it does really depend what kind of filtering you want to do. If i wanted a general purpose eq type filter I wouldn't use svf~, but it suits the stuff I'm interested in at the moment with sweeping effects. My main point was that for musicality your mileage may vary. Biquad~ is flexible, but you might have an application where a finite impulse response (using buffir~, or an fft method) would be more appropriate, or one of the other msp filter objects. If you can both listen, and also read up on filter design to learn what the characteristics of different filters are you should be able to better work out what you need for the sounds/music you want to make.

Quote: jayrope wrote on Tue, 09 December 2008 07:03
----------------------------------------------------

One of my main concerns about svf~ was it's 12025 Hz limit in particular, because it could actually kill the "air" in my signal, even if used a highpass outlet from it. That is unconfirmed, but also not answered back. So without any more msp-indepth knowledege about this i kind of refrained from using it, even if it cost more cpu.

So i was fond of finding the cross~ filter in the first place, as it accomplished my momentarily most needed tasks of delivering high- and lowpass filters on the same signal at the same time. Unfortunately i still can't find frequency response info about this yet. Will have to make a listening test later, sorry, should have done that already.

Your remakrs bout different filter designs are definately interesting, without more than reference-based knowledge i am content i found cascade, as well as reson~ yet.

Quote: jayrope wrote on Tue, 09 December 2008 10:34

> Alex, this was very detailed, thank you.

> One of my main concerns about svf~ was it's 12025 Hz limit in particular, because it could actually kill the "air" in my signal, even if used a highpass outlet from it....

I think that the situation here is that the parameter is limited, not the audio range (which I would have hoped I would've noticed if it were the case). In other words you may not be allowed high filter frequencies, as for technical reasons that make the filter unstable - this does not mean that the audio going through svf~ is limited in frequency range to fs/4. You can confirm this by passing a sine wave at 14kHz or higher (sample rate 44.1kHz) through svf~ and listening to the highpass output. So the issue is not the air in the signal, but whether you wish to have a filter frequency above fs/4, which is fairly high for the sweepy stuff I'm up to at the moment...

Here's a patch (in the old school 4 format) so you can see what I mean - careful with the volume, it's a bit piercing:

Quote: jayrope wrote on Tue, 09 December 2008 10:34
----------------------------------------------------
> Alex, this was very detailed, thank you.
>
> One of my main concerns about svf~ was it's 12025 Hz limit in particular, because it could actually kill the "air" in my signal, even if used a highpass outlet from it....

Thanx Stefan, no, i couldn't find info regarding cross~ freq response in the 4.63 help files...

And yes, i use filters for resonances, too, but generally it is more about a general purpose situation here momentarily, trying to split fish from meat. Erm. Probably looking for the egg-laying woolmilkpig again. And maybe i just go lowres~ in the end, as that one sounds very juicy despite it's obvious quality limits in absolute technical terms.

Looking forward to see you next week! Max/msp has got me caught now.

jayrope wrote on Tue, 09 December 2008 07:03

the 110 religion exspects me to have an "afk year" every 25 years - sorry world, i know how you must haved missed me.

btw for superfast, non-exploding highpass/lowpass try FIR. there should be some buffir example in the examples folder which should give you a clue.

those do not too sound good at only 32 samples lenght but you can change frequency without any interpolation.

(will send you some building blocks when i am officially back online.)

jayrope wrote on Tue, 09 December 2008 07:03110:

jrp
p.s. 110> long time no speak. fired up icq, momentarily doesn't really work. rrrr.

the 110 religion exspects me to have an "afk year" every 25 years - sorry world, i know how you must haved missed me. 

btw for superfast, non-exploding highpass/lowpass try FIR. there should be some buffir example in the examples folder which should give you a clue.
those do not too sound good at only 32 samples lenght but you can change frequency without any interpolation.
(will send you some building blocks when i am officially back online.)

[4.63] - sound quality of standard filter objects