Hi. I'm still kind of a novice at pd, so what I'm trying to might not be a good idea. It doesn't seem to be working any how.

Anyways, I was trying to make a distortion type deal where you select a window in the signal, say .5 to .75 of the absolute value of the signal, or whatever you like, and have it so the signal below that window (say 0 to .5) isn't reduced in amplitude, the signal above that window (.75 to 1) is reduced by a factor of your choice, I'll call it a, a value between 1 and 0, and the signal within the window ramps steadily from no reduction to reduction by a factor of a, so it's smooth.

What I've been doing for this is multiplying the whole signal by a clipped version of the absolute value of the signal that's multiplied and added to such that instead of going from .5 to .75 it goes from 1 to a. So, anything below .5 is multiplied by 1, anything above .75 is multiplied by a, and everything in between is multiplied by something between 1 and a. I thought this would produce a smooth sort of distortion, depending on how much the amplitude is reduced by and the width of the window, but no matter what I do I get a harsher distortion, and there are little rectangular cuts in the peaks that shouldn't be there, unless my logic about all of this is wrong.

This is roughly what's going on.

Signal
|
abs~
|
clip~ .5 .75
|
-~ .5 ( 0 to .25)
|
/~ .25 (0 to 1)
|
*~ -(1 - a) (given a reduction factor of .6 this would be 0 to -.4)
|
+~ 1

And that would result in a range of 1 to .6. And I multiply that by the original signal. I'm just kinda messing around with pd trying to make stuff, so maybe this is just a bad way to do this. Is it just too many operations on the clip to multiply it at the right time?

I can put a patch on this later if you need.

nothing wrong with the timing, pd is doing exactly what you're telling it to do.

i just made the patch you described, and i think the problem is this:

when you multiply the high amplitude parts of the signal by 'a', then with low values of 'a', the multiplication makes this signal drop below the level of the rest of the signal...this produces the rectangular sections you described.

just had a play and couldn't get an answer for you, but you need to rethink your patching logic to make it work properly. you're on the right track.

Your logic may be a little cloudy here. Let's review two things:

A linear system introduces no distortion. If the instantaneous amplitude of the signal input is A and the instantaneous output is B then A=B for all values.

You can view the process as B = f(A). The f is some function that produces B as a result of A.

A non-distorting (linear) effect can be gain, a simple multiply, in which case the function f is a constant, like B = 0.5A to halve the amplitude.

A non-linear system distorts the signal. [clip~] is a non-linear function which clamps B at some level for all A above and below a value. So is [abs~], which changes the sign of negative A values.

A compressor is not an instantaneous function. It changes the average level of B in accordance with the average level of A, yet it preserves the linear instantaneous releationship.

What you seem to be describing is an effect that behaves as a linear compressor up to a threshold, and beyond that it behaves as a distortion.

There are a few ways to go about it, but none can be done as a single flow. You need to introduce a side chain. That means splitting off the signal and performing some averaging/low pass filtering (maybe using [env~]), and then applying the result of that to a linear gain stage. Finally you introduce a separate distortion stage such that once the input passes a threshold some clipped signal results (regardless of the compressor stage).

I hope I have understood your goals.

Hey thanks a lot both of you for your replies. Hardoff, I see what you're saying about the signal dropping below the rest. I guess I didn't give the math enough thought.

Obiwannabe I understand most of what you're saying but it might be a bit before I'll be able to apply it. I've just started reading puckette's book, before I mostly just experimented with things. Would the purpose of low-pass filtering be to smooth the signal and yield a less sporadic amplitude? Is that what you mean by averaging? Also your site's a great resource I plan on going through it when I'm done with the book.

> Would the purpose of low-pass filtering be to smooth the signal and yield a less
> sporadic amplitude? Is that what you mean by averaging?

Yes, that's exactly it.

On a sort of similar note...

How is it possible using a low-pass as an envelope follower to alter the attack and decay response. There was post the other day we this came up i think.....

Something like:

in
|
[hip~ 5] remove DC offset which would give odd values below
|\
[*~] square to ensure signal is always > 0
|
[lop~ 5] filter out the sound to leave the amplitude squared
|
out

might work ok, i haven't tried it yet

EDIT:

for attack/decay, I'd probably use expr and the elementary filters, something like

out from above
|
[expr~ $v1>$f2; if($v1>$f2,$f3,$f4)]
| |
[pd $0-lowpass-filter]
|
new envelope

where $f2 is threshold, $f3 is attack rate and $f4 is decay rate (for the filter)

the [pd $0-lowpass-filter] would use rzero~ rpole~ at a guess, with the rate changing the radius of the pole. No time to figure it out now.