How can I get a 'narrow slice' of white noise?

[malcolmcolwell]

I have both Audacity 1.2.6 and 2.0.5, running on XP.

My apologies is this is such a 'newbie' question that I shouldn't be asking it, but what I'm trying to do I have to get exactly right, so I thought it best to ask the experts.

What I need to do is produce a sample of white noise at one very narrow frequency, and another at two very narrow frequencies. Ideally, there should be silence as soon as those frequencies are departed from - a total cutoff.

Can anyone advise me, please?

[kozikowski]

How narrow?

If the sample of noise isn't there long enough to actually make noise, then you get some very oddball Nyquist trash instead of white noise.

You have to have at least three cycles of "tone" to qualify as musical and you need more to stop and start. Three cycles of 20 Hz is well over a tenth of a second, so your sample can never be shorter than that.

Generate gives you durations, amplitudes and types of noise. Isn't that enough?

a sample of white noise at one very narrow frequency,

Then it's not White Noise. White Noise has specific relationships between all its frequencies. You can't pick and choose.

What's the job? We walked in in the middle of the movie and we can't follow the plot. What are you really doing?

"I need a brief blast of noise because....:"

Koz

[Trebor]

What I need to do is produce a sample of white noise at one very narrow frequency ...

Can be done using the equalizer or this "Ultra Narrow" plugin.

[steve]

"White noise" is loosely defined as sound that contains all frequencies within the audio range in equal amounts (constant power spectral density). So, by definition, if you narrow the frequency bandwidth it ceases to be "white noise". To take the narrow bandwidth to the extreme you would have just a single frequency, in other words, a sine wave.

You can generate "narrowband noise" with this "Nyquist script":

Code: Select all

(setq centerfrequency 1000)
(setq bandwidth 100)

(mult (hzosc centerfrequency)
    (lowpass8 (noise) bandwidth))

To use the script, select part of an audio track, then open the "Nyquist Prompt" effect.
Copy and paste the above code into the text box and apply.
The first two lines may be modified as required to set the centre frequency and bandwidth as required.

[malcolmcolwell]

Hi

My apologies for not explaining what I wanted to do and why - can't blame you all for being mystified!

Here goes: I suffer from tinnitus, and while there's no cure, there are lots of remedies that help some people but not others. Trouble is, people try to make as much cash out of you as possible with these treatments, since T sufferers are so desperate.

One of the latest remedies is called 'neuromonics'. With this, you find your T frequency by matching what you hear to a tone generator, then you listen to music that had had 'white noise at this frequency', as they describe it, overlaid onto it. You do this for many hours, then you switch to versions of the same music but without the noise. Miracle results have been reported (but then , they say that about every other T treatment).

The problem is, the people who developed this charge $5,000 for the treatment! And there's no money back guarantee - they won't even refund a portion of the fee.

So lots of T sufferers have been trying to do 'DIY neuromonics', since overlaying music with white noise is a simple matter.

But the big problem is knowing what 'white noise at the frquency of the T' actually means. If your T is a pure tone, of course, it's a misnomer - the 'white noise' will be a pure tone as well. And the neuromomics people won't tell you anything.

So, I've decided on trying white noise of maybe half or quarter of an octave, centred on my T frequency, and broaden or narrow it if that doesn't show any results. But I want to be sure the cutoff is as clean as possible, and that the noise is centred accurately and has exactly the bandwidth I want.

I could mess about with equalisation curves, but maybe there's a surer way, where you can actually type the frequencies you want into a dialogue box.

This is so important to me that I don't want to do it by trial and error if I can get some expert advice on how to do a proper job. Thanks.

[Robert J. H.]

Here's a snippet for the Nyquist prompt that you can try

Code: Select all

(psetq center 4000; Hz
       width 200; Hz
       gain -12); dB, volume of the noise (here, 0.25 linearly)
(sim (areson s center width 1); original without center
     (scale-db gain  (reson (noise) center width 1)))

The audio at the frequency center is replaced with white noise around the same frequency and with the same band-width.
There is never a steep slope (brickwall filter), it is physically impossible.
However, "reson" is not the only filter that can be used for this.
We can also engage a low-/highpass pair with 48 dB roll-off per octave.
If you do not want to eliminate the original audio, then write "s" (without quotes) instead of the areson-part.

It is also thinkable that we use the original envelope for the noise portion.

[steve]

With this, you find your T frequency by matching what you hear to a tone generator, then you listen to music that had had 'white noise at this frequency', as they describe it, overlaid onto it. You do this for many hours,

That is the opposite to what has been previously described by other visitors to the forum that were seeking relief from tinnitus. There is a lengthy discussion here; http://forum.audacityteam.org/viewtopic ... 967#p74967