way to loop 1 second of noise but sound like its not looped

[[email protected]]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

 

[Greg Neill]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

Can you play it backwards and forwards repeatedly, rather than
restarting at the beginning?

 

[[email protected]]

funny you should mention this! i tried playing it back and then
reversing it and it sounds much better and there is a very large
reduction in the loop frequency "beat". however there is still an
audible "heartbeat" although much reduced

 

[markp]

funny you should mention this! i tried playing it back and then
reversing it and it sounds much better and there is a very large
reduction in the loop frequency "beat". however there is still an
audible "heartbeat" although much reduced

How about cutting the sample into smaller samples, say 16, then randomly
re-ordering them on playback?

Mark.

 

[[email protected]]

i should also say that I can manipulate the original noise file in
anyway I like using windows audio utilities, and then dump it into the
playback device. i have a 10 second sample of the noise, and I am
cutting out a 1 second chunk of it for the device. but I can cut splice
modify those ten seconds however I want to come up with the magic 1
second clip

 

[[email protected]]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

Does it help at all to crop the wave file such that at the end of the
wave, the wave function is as close as possible to Y=0 (with a positive
slope - that is, approaching Y=0 from just below the X axis)? And also
crop the wave file at the beginning, so that it also begins at or near
Y=0, and also with a positive slope? That way, at least the wave file
would appear continuous.

 

[John Popelish]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

I think I would perform an FFT transform of the 1 second burst,
discard all components below some minimum inaudible cutoff and then do
an inverse transform of what's left to produce the output waveform.
The beginning and end should splice perfectly, though it will contain
only integer Hz frequency components.

 

[Luhan]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

You don't say what your 'device' is.

If its hardware, substitute a feedback shift register and generate the
noise.

If is a micro, just generate random numbers and output them.

Luhan

 

[Ben Bradley]

You don't say what your 'device' is.

If its hardware, substitute a feedback shift register and generate the
noise.

If is a micro, just generate random numbers and output them.

I was going to suggest this. A small linear-congruence PRNG, either
in code or in hardware, can generate a sequence of random numbers with
a period of minutes, hours or days. It's much better than using 44k
words of memory to hold one second's worth of random numbers.
I suspect a second's worth of looped random noise will always sound
looped, and there's no manipulation you can do to fix it.

 

[Robert Baer]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

Look at the waveform at the start; maybe the first 100mSec, and then
look near the end, say the last 200mSec to find something that "matches"
in general shape / pattern, and most especially slope and polarity.
Those places are where the splice should go, after applying slow
attenuation or increase of gain (if necessary) to one end to match
amplitude.
Use the highest sampling rate possible for the signal of interest as
the source for this work; one can always "dumb it down" later if need.

 

[Luhan]

I suspect a second's worth of looped random noise will always sound
looped, and there's no manipulation you can do to fix it.

Yes, this is what I found out some time ago. Over a period of a second
or less, your ear will always here the cycle.

Luhan

 

[BobG]

Looping is an art. Sampling synthesizer guys and game sound designers
do it all the time. There are programs that have several looping
algorithms... some sound better than others depending on the wave being
looped... obviously, the beginning and end need to splice together
smoothly. If the sound being looped is periodic, like a trumpet note,
or an idling engine, you need to make the splice at the end of a cycle.
The frequency components at the beginning and end need to be steady...
you cant loop the engine if its speeding up... that loop would sound
like a sawtooth... speedup, jump back down. The loudness needs to be
the same at the beg and end of the loop, or you get a beat at the loop
freq... sounds like a washing machine. Cooledit (now adobe audtion)
lets you preview the loop and move the loop point around. Best results
are by trial and error in my experience.

 

[[email protected]]

I have a digital sample of noise, its not really pink or white, it has
various peaks and valleys, and I'm trying to find a way to loop it
without it sounding like its looped. My device can only play back 1
second, or possibly 2 seconds of sound, (if I reduce the sampling rate
from 44100 to 22500). Is there a way to do this? Its all subjective and
the idea is to make it sound good to a person, its not for an
experiment.

why not amplify the noise from a zener - that will be good... what do
you want it for anyway>