thanks to claudius maximus and others for this idea.

it lets you increase the volume of an audio stream without getting nasty digital clipping and buzzing. ..of course sometimes you do want that sound, but more often than not it will add nasty frequencies to your sounds when you push their volume up too high.

you can use it in any situation where you you normally use [clip~ -1 1] ... or the way i prefer to think about it is that you can use it in any situation where a hardware device has an audio output.

so, if you make a synth running through an effect to a mixer, then you would use the tanh function at the outlet of the synth audio stream, then another one at the outlet of the fx stream, and then again at the outlet of the mixer.

the simplest way to do it is this:

[inlet~]
|
[expr~ tanh($v1)]
|
[outlet~]

but [expr~ ] kills your cpu, so it's better to put the function into a table, and then just read the table with [tabread~]

so, here we go:

(have a look at the table you make with this, and you will see the lush tanh curve)

Good call, does it act as a kind of saturator/compressor?

saturater - yes, i think so.

compressor. no,

for compression, you take the low amplitude samples and bring them up, while keeping the high amplitude samples constant. i don't think you can do that with a simple tabread~ patch.

this one just allows you to increase all the amplitudes equally, but rounds the high amplitude samples off so that they cut round instead of square.

so, for example, if the amplitude for samples on an overdriven wave goes like:

0.5, 0.7, 1, 1, 1, 1, 1, 0.7, 0.5,

then, this will smooth it to:

0.5, 0.65, 0.9, 0.95, 1, 0.95, 0.9, 0.65, 0.5

Thanx a lot (again) hardoff... I'm going to look into this one, because i like louder, but then you know i like nasty digital clipping too, so what about louder nasty digital clipping

What about the object [tanh~]?

It is fantastic, thanks hardoff...
I was wondering, doesn't Pd process audio as 32 bit float? If it is right, then it would be better to use tabread4~ instead of tabread~, don't you think?

In fact you'll probably have errors (some noise) to constrict the range of possible values from the [inlet~] signal to a 100000 points table.

anyway this is a very cool method to level volume, pd way of thinking amazes me every time

bye

thanx for the idea,

i made a bitCrusher with it

HK

http://www.pdpatchrepo.info/hurleur/bitCrusher.pd

This might be relevant to the discussion.

I made a nonlinear compressor external using polynomial functions. It's over at puredata.info

http://puredata.info/Members/saturno/saturator-non-linear-compressor
http://puredata.info/Members/saturno/saturator.zip/view

S

it is some years I use the Hardoff's saturator and it sound great
these days I'm working on a project using libpd and xcode and I can't use the expr operator because of the GNU license.
So I was looking for some mathematical function to approximate tanh
finally I used this: y=0.68*atan(1.95*x)
I post the patch. This simple example with atan allows you to change the slope of the curve, by modifying (1.95*x) and create the desired boost for the mid-range level (you have to clip -1 +1 if you increase this value)

It is not the same to tanh of course, I ask you some suggestion to improve it without using expr. I'd like to use the error function but I don't know how to implement in Pd.

cheers

http://www.pdpatchrepo.info/hurleur/atanCompSaturator.pd