KVR Audio

itoa · Post by **itoa** » Sat Jul 26, 2014 7:22 am

Any idea how to transform this kind of nonlinear lpf into bilinear form?
y = y + a * (shape(y) - shape(in))

Edit: Of course I mean using bilinear transform instead of backward Euler

karrikuh · Post by **karrikuh** » Sat Jul 26, 2014 2:11 pm

What's a "bilinear form"?

itoa · Post by **itoa** » Sat Jul 26, 2014 6:45 pm

Ok, so I came up to this form

iin =shape( input) - shape(lastOut)
out = buf + f * iin
buf = f * iin + out
lastOut = out

But it needs 2 state variables buf and lastOut, so I thought I don't get something important.

itoa · Post by **itoa** » Sun Jul 27, 2014 3:24 pm

Ok continuing my own private thread

It turned out that using bilinear integrator with additional delay makes this filter as bad as its 'naive' form.
And.. I can do nothing else than TPT and solving higher order equations (shaping). Welcome to hell

Z1202 · Post by **Z1202** » Mon Jul 28, 2014 8:18 am

itoa wrote:Ok continuing my own private thread

It turned out that using bilinear integrator with additional delay makes this filter as bad as its 'naive' form.
And.. I can do nothing else than TPT and solving higher order equations (shaping). Welcome to hell

Try using the "cheap" version (applying the nonlinearities on top, after solving the linear ZDF equation). Or use mystran's method (each approach has pros and cons).

itoa · Post by **itoa** » Mon Jul 28, 2014 8:20 am

Yep tried it.. kind of rape on great TPT idea

. I gonna give hell a chance. Using some sigmoids makes equation analytically solvable.

Bilinear form of nonlinear one pole lpf