[chuck-users] heads up for

Ge Wang gewang at CS.Princeton.EDU
Tue Sep 12 06:30:59 EDT 2006


> I do agree that we need a very good look at this now. How many filters 
> do we want in the end? I was thinking that maybe these should be 
> called LPF12, HPF12 and so on (second order so 12db/oct) and later we 
> might also have LPF24 (24 db) and maybe even LPFMoog, LPF303 or 
> LPFPHLin (phase linear) or whatever.

I had originally thought that LPF, HPF, BPF, and BRF should be 
expandable without breaking the api.  For LPF (or whatever it should be 
called), we have so far

     .freq (cutoff)
     .Q (resonance)

This is a 2nd order butterworth.  In the future, if we want our 
24db/oct filter, it might be cool to chuck 4 to:

     .order (filter order)

Furthermore, there could be something to control the type:

     .type (butter, cheb1, cheb2, ellip, more)

(Of course, it's worth remembering that we can already construct any of 
these higher order filter using building blocks such as BiQuad's and 

Now, I am not sure if this is the right factoring.  But organizing it 
by function (low, high, etc.) seems slightly better than a monolithic 
FlexFilter type...

Also, given this strategy, we hopefully should be able to add more 
filter types without inflating the namespace.  Thoughts?

If we want to go all out, someone could implement a chuck library for 
filter design, like the lovely matlab functions such as

     .buttord (great name; specs in, minimum butterworth order required 

as well as analog prototyping functions and bilinear transform??  It's 
probably pointless to try to duplicate matlab in any significant way, 
but the filter package is really fun and can be good for teaching in an 
audio driven manner.  That would give us many different ways to 
approach filter and friends.

now I am talking crazy...  must be dreaming...


More information about the chuck-users mailing list