Digital filters?

[Z1202]

But when you say digital filter alias the high end, isn't over sampling (192kHz) the simplest and only answer to that?

No, it's not the only way

...

You can make your own transform to bridge the gap, it's only a matter of precision used.

What I was wondering is what is more efficient - oversampling or a better transform of equal quality.

Best regards,
Dr Sync

[Urs]

Oversampling is a good idea anyway. The phase mess usually starts at Nyquist/2...

I also like the approach of the double sampled state variable filter, as described at musicdsp.org - simple and works as expected.

Cheers,

Urs

[antti]

DrSync had it right, modulation is not covered in textbooks, and that's where the hard part resides, modulation is continuous in analog filters, whereas it can only be faked to a certain extent in DSPs, you get steps. all right.

You can make (easily too) a digital filter with continuous modulation. The steps people talk about here are probably from changing the cutoff with mouse or midi CC (as stefancrs pointed out). This really has nothing to do with the filter itself and is a matter of smoothing out the user control. A not trivial problem since for "proper" smoothing, you'd have to know the future values in advance.

Concerning tubes, you'll find a good collection of models around, saying the amplitude, order and offset of each harmonic they create. But concerning compression it's a whole different problem in its own, hard to figure out....

The "compression" is a property of the circuit, not the tubes themselves. A tube just distorts, and you can model a normal preamp tube completely with two 2d tables. Depending on the circuit the tube is in, you can get away using a simpler model even (see my Saro plugin at http://antti.smartelectronix.com)

[spaceman]

in impOSCar.. if you move the mouse away far enough from the control, the value increments are very small. I'm sure a trained ear can still hear some sort of "stepping" but there's no way in hell your average listener would hear it.. and definitely not in a mix

this is the same nonsense as the aliasing crap .. who f**king cares

like a bunch of nerds tossing off on some well written Perl code..

you make music for real people you know, not bloody mastering engineers.. if it sounds good, it sounds good, with or without aliasing or any other crap

[monsterbeetle]

[quote]like a bunch of nerds tossing off on some well written Perl code.. [/quote]

Well, the dudes who actually decided one day to code EQs for VST hosts had to figure the shit out so that it sounds good, so it's no nerdy wanking in talking a little bit bout the techie side of the prob

[spaceman]

like a bunch of nerds tossing off on some well written Perl code..

Well, the dudes who actually decided one day to code EQs for VST hosts had to figure the shit out so that it sounds good, so it's no nerdy wanking in talking a little bit bout the techie side of the prob

I love the techie side of it, I'm reading this with great interest.. but I think some people are taking this too far into the theoretical, trying to get the best possible maths to do the job without keeping an eye on the important bit.. the ears

with current computer power there's no way you're going to get it modelled very close to its analogue couterpart where you come to a point you can call it a real "model".. not on your average machine anyway, running something like VSTs

so concentrate on making them sound good instead


and sorry.. my previous post did sound a bit harsh..

I blame the aliasing-police

[citizenchunk]

I love the techie side of it, I'm reading this with great interest.. but I think some people are taking this too far into the theoretical, trying to get the best possible maths to do the job without keeping an eye on the important bit.. the ears

i think that you are overlooking the importance of theory. any halfway decent programmer can grab some biquad filter code from the musicdsp archives and create an EQ plugin, without any understanding of the bilinear transform. (i'm a perfect example.) but what makes a really good EQ plugin is the attention to detail--most of which is theory. all of that fussing over aliasing helps to avoid some pretty nasty side effects. i think if you have ever heard them, you would agree that with me that the programmer should have paid more attention to aliasing.

[stefancrs]

this is the same nonsense as the aliasing crap .. who f**king cares

Erm. You should. It's _oh so_ audible if you don't use higher sampling rates than one really should have to. Vanguard for instance really do sound totally crappy at higher notes, and I'm talking about normal notes here. Notes you want to use but can't since they sound too shitty on some synths. Even if you don't exactly hear it in a mix they for sure messes the mix up. Sometimes I want that aliasing effect, but that is usually because I want aliasing due to ugly resampling, hence I use elite reducer or the like. My choice :)

[texture]

you make music for real people you know, not bloody mastering engineers.. if it sounds good, it sounds good, with or without aliasing or any other crap

You can make music that sounds good, but to make music that sounds amazing, then you really do need to pay attention to the detail and try to keep everything sounding good at every step in the production process.
Errors quickly escalate, so if you minimise them all the time the results will sound significantly better to the untrained ear, even though the difference at each small step may have been totally unnoticable to them.

[mdsp]

just a note on the bilinear transform, so that newbie don't get confused...

there's NO alisaing involved here. Everything is strictly linear, you'll never here aliasing caused by a biquad filter using the bilinear transform.

someone mention distorsion in the HF, it only means that the analog frequency range 0:Inf is mapped to the discrete one 0:nyquist with an atan mapping.

p -> A * (z-1)/(z+1)
=>
wa = A * tan(wd/A)

When using A = 2/T, this transform is a good approximation in the LF but the error grow as you increase the frequency.
So if you had a 10kHz cutoff in the analog world it will be slightly distorded to say 9.5kHz.



With other values of A, you can choose the frequency range around which the transformation will be accurate.

cheers

remy

[snooky]

try this:

A=0(-2/5)
Z*Z - Z*X (2=4)
P-P-Z=(2*2)

2*4=z/p

4%2-&"6))-%
75-&7?-7!$
$$£

Try this math example and you will get a filter that sounds exactly analouge and your PC will get +2ghz processing power.

[spaceman]

i think that you are overlooking the importance of theory. any halfway decent programmer can grab some biquad filter code from the musicdsp archives and create an EQ plugin, without any understanding of the bilinear transform. (i'm a perfect example.) but what makes a really good EQ plugin is the attention to detail--most of which is theory. all of that fussing over aliasing helps to avoid some pretty nasty side effects. i think if you have ever heard them, you would agree that with me that the programmer should have paid more attention to aliasing.

do you think it's possible someone might come up with a completely different mathematical approach to the problem at some point or are developers pretty much stuck with what's available now?