Bezier Mirror Overdrive , simle, light, distortion with fat sounds

Jean louis P
Jean louis P Member Posts: 12 Member
in Building With Reaktor
Bezier miror overdrive 1.png
Bezier miror overdrive 1.png

This is distortion with very simple formula without complex and heavy formula like tanh , and it gives fat sounds like with analog synths. Inspired by Bezier curves but with two segments only , no need more.

It works with absolute values .

It has a normal behavior for values between -1 and +1 , and is designed to manage value up to -4 / +4 by a mirror effect generating a negative bump at the top with a buffer effect.

The core formula for the distortion is just of 3 blocks, the line in the middle, a cross fade between 0 and 1 level. It gives a classic distortion profile between levels 0 and 1 from a linear source with level between -1 and +1.

For the mirror part the values between 1 and 4 , are compressed to 0 - 1.

I use the same formula than for the normal part, multiply the result by 0.5 , then make the mirror work by removing the final value to 1 , for the deeper level of the bump at 0.5. The clipping of values above 4 happens at 0.5.

The final sound with the mirror effect is very fat, like you can hear in the demo.

There is a boost knob between 1 and 4, multiplying the input value.

I have added a delta knob for experimentation , adding a fixed value to the input value between - 0.5 and 0.5. The major part of the demo has deltas values ( we can hear an heavy cyclic effect by the PW ) , removed toward the end making the PW effect more fluid.

bezier mirror 1.zip
Tagged:

Comments

  • colB
    colB Member Posts: 1,068 Guru
    edited May 6

    You can handle inputs with magnitude greater than four:

    image.png
  • Jean louis P
    Jean louis P Member Posts: 12 Member
    https://community.native-instruments.com/discussion/comment/225008#Comment_225008

    yes but i was trying to recreate the behavior of my original model, by replacing a forced distortion of higher level boost, from -1 +1 entry, where for a value multiplication around four i had a top value around 2 and the final value ( for x = 1 ) of 1, by a natural answer to bigger entries with -1 +1 output.

    The purpose is to have a classic like distorted sound between base values ( 0-1 ) , and to have a fat answer for values above 1.

    The clipping at 0.5 solves the problem. In the context i made , at the output of oscillator this one a max effective value of 4 is enough.

    I have not tested on sine wave, but your picture this is the global idea that i had about a value greater than four with mine distortion, with a flat area around the top value.

    I am impressed that you have reworked my work for something new. The formula is different, then the behavior with the same input should be different.

    Where do you get the root square of 2?

    How it sounds compared to mine with sawtooth?

  • colB
    colB Member Posts: 1,068 Guru
    edited May 6
    https://community.native-instruments.com/discussion/comment/225060#Comment_225060

    I am impressed that you have reworked my work for something new. The formula is different, then the behavior with the same input should be different.

    Where do you get the root square of 2?

    It's not different (I think?), it's the same, but with fewer operations.

    EDIT: yes, it's identical apart from some numeric error down at ~+/-1.0e-07

  • colB
    colB Member Posts: 1,068 Guru
    edited May 6
    https://community.native-instruments.com/discussion/comment/225060#Comment_225060

    Where do you get the root square of 2?

    Square root of two comes from removing the *0.5 later in the expression.

    image.png

    Simplifies to:

    image.png

    To remove the later *0.5 (divide by 2), we need to scale the two constants. The 2 becomes square root of 2, and the 0.33333333 becomes sqrt(2)/3.

    (I just faffed about with some very basic algebra)

    so:

    image.png

    becomes:

    image.png

    same with the 0..1 section, where you can fold the addition into the multiplication, so you only need two ops instead of three.

    The downside here is flexibility.

    If you keep the less simplified version, you can directly change that 0.5, and the one for the >4 case, and so you have another parameter you can externalise, scaling the foldback amount…

    I'm sure it's possible to 'do the math' to parameterise the concise version, but time is not an endless resource. :)…

  • Jean louis P
    Jean louis P Member Posts: 12 Member
    https://community.native-instruments.com/discussion/comment/225071#Comment_225071

    woaw it is deep 😀

    good work, congratulation.

    I was wondering why you have done something like this , and not just used my datas with a special management for input above 4. from the first message.

    If i understand you pass by the roots of the values for removing the clipping at 0.5 because the input of the clipping is related to the the square of the input, after your first simplification.

    What is amazing with your simplification, is that for the base part, 0..1 , we have a distortion formula with just two blocks, for a fat sound.

Back To Top

Categories