does anybody know what happens mathematically when you layer a sound wave over another one?

I was thinking about this last night when I was trying to cut up some loops that I downloaded.

Heres an example:
-I have a kick that I want to extract out of this loop but it has a hi hat on it that I want gone.
-If, when layered, audio files are multiplied together? then would you be able to remove the hi-hat by layering an inverted wav of the hi-hat over the top?
ie. y(x*(1/x)) = y

havent had the time to try it but would it work? If not is there any way to change this so that it could work?

Trying to take out other sounds by EQing just doesnt do it for me because I lose too much of the sound I want.

Jiddu

Any audio waveform is a single valued function at any instant. Two waveforms are not multiplied but algebraically added. If you multiplied them you would get sidebands and unrelated frequencies. If you had the exact hi-hat waveform you could subtract it (or add the inverse) in phase and it would cancel. But where are you going to get that exact waveform?

neato..

I'm definitely going to try that then.

Where am I going to get the exact waveform?
It'll most likely be the hi-hat I want to rip out so I can get it from somewhere else in the track. I know it won't be identical but it should be close enough right? Or will it create wierd distortion because they wont be in phase and the frequency is so high that will create a new uglier sound?

Probably what you'll get is all sorts of artefacts, little spiky remnants of the original and funny echoes. Worth a try though.

Why don't you just run the kick through a lowpass filter? Lower the cutoff until the hat is gone or does that mess up the kick?

-tek

tek:yeah its sampled from an old soul song so too much of the sound i want is cut out

plus i was thinking I might be able to come up with some new ground breaking loop editing technique

if anyone wants to implement an applet that will take two wav's, layer one inverted over the other then run some algorithm (yet to be determined) to clean up the sound a bit then say.

here is a thought...run it through a high pass filter on one of the .wavs so that only the mid and bass is audible. then place the original over it so that the only frequencies that cancel themselves are the low ones leaving the hat...

or do the oppisite, as i didnt read it correctly.

its worth a shot

i tried a little experiment in wavelab...i see the error in my thought process.

if there was a way to retain the same wave shape while boosting only certain frequencies it might work. but alas i have no idea how to do that.

Jiddu,

There's nothin to it. That's an easy one.
Sound editing...that's my department. I wanted to wait it out until the thread was at rest.
-Tek, pilot and especially mrt1212 were real close

Here is what you do.
1.)Make 2 workable copies
2.)take a copy and run a filter to allow the hit hat only (hi pass*) (the part you don't want)**
3.)take this hi hat only copy and invert it only. do not change it's length
4.)mix paste the inverted hi hat only copy into the other workable copy by about 93-100%and the hi hat will be cancelled out providing it was pasted in the proper place which is essential.
That's it.
*hi pass is a coarse term when editing wavs. Select the frequency spectrum that you need to work and use that amount whatever it is.
**Bass drums have a high frequency "click" to them. Do not select this portion of the wav. and involve this portion with the inverted cancellation proccess of the hi hat. If the hi hat and the "click" sound of the striking attack of the bass drum are played at the same momnet in time, that's ok too but we will have to take it a few steps further. Let me know if it is and I will explain how you do that.

[This message has been edited by Morphamatik (edited 04-26-2003).]