Fuse Audio Labs releases the OCELOT Clipper Peak Shaper with Intro Offer
-
- KVRAF
- 2260 posts since 25 Jun, 2008 from Montreal, Canada
Do you plan to add ADAA to the Limiter or it's not necessary when limiting?
-
- KVRAF
- 8705 posts since 24 May, 2002 from Tutukaka, New Zealand
I had a quick read up on ADAA.
https://ccrma.stanford.edu/~jatin/Notebooks/adaa.html
TBH it all went completely over my head - may as well have been written in Nepalese. I gave up at 1st order ADAA, never mind 2nd order.
"2nd-order ADAA
A second option is to use higher-order antiderivatives. Let's say that our nonlinear function y[n]=f(x[n])
has a second antiderivative, which we'll call F2(x)
. Then second-order ADAA can be written as:
y[n]=2x[n]−x[n−2](F2(x[n])−F2(x[n−1])x[n]−x[n−1]−F2(x[n−1])−F2(x[n−2])x[n−1]−x[n−2])"
But it seems it deals with aliasing better for less CPU as long as it's alongside oversampling. Which makes sense for limiting as well as clipping. Limiting will distort as well.
https://ccrma.stanford.edu/~jatin/Notebooks/adaa.html
TBH it all went completely over my head - may as well have been written in Nepalese. I gave up at 1st order ADAA, never mind 2nd order.
"2nd-order ADAA
A second option is to use higher-order antiderivatives. Let's say that our nonlinear function y[n]=f(x[n])
has a second antiderivative, which we'll call F2(x)
. Then second-order ADAA can be written as:
y[n]=2x[n]−x[n−2](F2(x[n])−F2(x[n−1])x[n]−x[n−1]−F2(x[n−1])−F2(x[n−2])x[n−1]−x[n−2])"
But it seems it deals with aliasing better for less CPU as long as it's alongside oversampling. Which makes sense for limiting as well as clipping. Limiting will distort as well.
-
- KVRist
- Topic Starter
- 249 posts since 13 Jan, 2018 from Duesseldorf, Germany
Hi kritikon,
This all comes at a price, e.g. extra logic required to handle limit cases, non-linear phase, high frequency attenuation. The latter can be mitigated using a compensation filter and combining the method with mild, e.g. 2x, 4x oversampling the phase warping can be pushed outwards to non audible regions of the spectrum, too.
There have been various extensions/improvements of the method to stateful systems, as well as the applications of IIR filter kernels to achieve yet superior anti-aliasing performance.
Hope this helps.
Best,
Ray
The idea is pretty straightforward, though: One is using picewise linear interpolation to approximate the continuous time signal to which the non-linearity is applied. Using a short and simple FIR kernel such as a rectangle (1st order) or triangle (2nd order) as a means to achieve anti-aliasing one can resample this intermediate signal back to the discrete time domain. Given the fact that due to the linear interpolation the differential quotient dx/dt is constant between two samples, evaluating the continuous time integral only requires evaluation of the anti-derivative (hence the name) at the sample points, resulting in the expressions you quoted.kritikon wrote: Mon Feb 24, 2025 8:26 am I had a quick read up on ADAA.
https://ccrma.stanford.edu/~jatin/Notebooks/adaa.html
TBH it all went completely over my head - may as well have been written in Nepalese. I gave up at 1st order ADAA, never mind 2nd order.
This all comes at a price, e.g. extra logic required to handle limit cases, non-linear phase, high frequency attenuation. The latter can be mitigated using a compensation filter and combining the method with mild, e.g. 2x, 4x oversampling the phase warping can be pushed outwards to non audible regions of the spectrum, too.
There have been various extensions/improvements of the method to stateful systems, as well as the applications of IIR filter kernels to achieve yet superior anti-aliasing performance.
Hope this helps.
Best,
Ray
-
- KVRAF
- 8705 posts since 24 May, 2002 from Tutukaka, New Zealand
Been otherwise engaged recently so properly tried it out today. Bought it 
.
It can do a lot of clipping before it gets noticeable, and heavily pushed, I quite like the distortion on it. Superb on drum bus, good added bit of grunge and body on 303 type synths. Not at all bad on CPU either with reasonable amount of oversampling. I think this might be a regular in my mixes.
.It can do a lot of clipping before it gets noticeable, and heavily pushed, I quite like the distortion on it. Superb on drum bus, good added bit of grunge and body on 303 type synths. Not at all bad on CPU either with reasonable amount of oversampling. I think this might be a regular in my mixes.
-
- KVRist
- 148 posts since 16 Sep, 2023
Thanks for making an effort to explain this to us curious laymen.
Would it be fair to say that the aliasing represents an unwanted derrirative of the clipping function and that the ADAA is a mathematical function that will make sure, that the unwanted derrirative gets subtracted from the final output, so that the desired clipping in the form of upper harmonics, but not the aliasing (misplaced, undesirable harmonics mirroring from the upper frequency limit) remain?