Removing Hiss

[rainmaker]

Let me know what other info you may need. I laid it out pretty clearly in my last post. But if you have specific questions, let me know and I'll try to answer them for you.

[axxeman]

Just a quick follow up to this...

Eventually I used the Waves X-Noise that I already had, (having tried a couple of others without success). Having set a noise profile, I rendered the three offending tracks and, with judicious use of the threshold setting, managed to clean them up very nicely without compromising the audio that I was trying to preserve.

Many thanks to all who helped in the thread and a (very) belated Happy New Year to all fellow 'Slutz'.

Axxe

[Nonlinear]

Originally Posted by wado1942

Hiss is a random noise. If you sample hiss and mix it back with the noisy track, it doesn't matter if the polarity is matched or inverted, you will always wind up boosting the hiss by 3dB.

This is not the same situation as trying to cancel tape or electronic hiss. If the noise is part of the sampled waveform it is no longer "random" - it's now part of the waveform data - and can be inverted and canceled. (Take a track of pure noise, copy it, invert it and mix it back. It will null. It HAS to because it's simply adding inverted numbers.)

The problem with that approach is that you only want to cancel the hiss at low levels - otherwise you just create a high-cut filter. That would require some kind of noise gate on the inverted track that only opens below a threshold. I can't think of any noise gates that do that (but you could invert the gating action using a third track...)

Just use Izotope RX2 - works great!

[axxeman]

Originally Posted by Nonlinear

This is not the same situation as trying to cancel tape or electronic hiss. If the noise is part of the sampled waveform it is no longer "random" - it's now part of the waveform data - and can be inverted and canceled. (Take a track of pure noise, copy it, invert it and mix it back. It will null. It HAS to because it's simply adding inverted numbers.)

The problem with that approach is that you only want to cancel the hiss at low levels - otherwise you just create a high-cut filter. That would require some kind of noise gate on the inverted track that only opens below a threshold. I can't think of any noise gates that do that (but you could invert the gating action using a third track...)

Just use Izotope RX2 - works great!

The stand alone trial version has save disabled and the plug in was just too complex for me to understand given that I wanted to get quick results. I daresay, like most software, when you have used it 5 or 6 times it becomes second nature, but I couldn't justify spending half a day trying to work out the parameters. Time's precious at my age!

Thank you for your input though.

[The_K_Man]

All sizes | Giant Snake in Malaysia? | Flickr - Photo Sharing!

[haberdasher]

Originally Posted by Nonlinear

This is not the same situation as trying to cancel tape or electronic hiss. If the noise is part of the sampled waveform it is no longer "random" - it's now part of the waveform data - and can be inverted and canceled. (Take a track of pure noise, copy it, invert it and mix it back. It will null. It HAS to because it's simply adding inverted numbers.)

Sure it will get rid of the noise, but it's a crazy thing to do. You will remove all noise and signal, in which case why not just mute the channel?

There are several more focussed ways of doing this, all already touched upon:

1) dedicated noise removal software

2) editing - find a similar bit and copy it there, or loop a fade out with more attenuation each time

3) low pass filter/other EQ

4) expander/noise gate

5) low pass filter/other EQ AND noise gate

6) multiband expander/noise gate

7) fade out quicker

8) ask a friendly mastering engineer

[Nonlinear]

Originally Posted by haberdasher

Sure it will get rid of the noise, but it's a crazy thing to do. You will remove all noise and signal, in which case why not just mute the channel?

Right. Now read the second paragraph of that post.

[haberdasher]

Originally Posted by Nonlinear

Right. Now read the second paragraph of that post.

The point is, that inverting a signal against itself is no different from muting a channel. So why go to the effort of creating a new track and inverting it? It's nothing to do with a high-cut filter, it attenuates *everything*.

[Nonlinear]

Originally Posted by haberdasher

The point is, that inverting a signal against itself is no different from muting a channel. So why go to the effort of creating a new track and inverting it? It's nothing to do with a high-cut filter, it attenuates *everything*.

I see the confusion - I didn't type everything I was thinking. The copied and inverted track needs to be high pass filtered (using delay compensated, linear phase EQ). That will cause the mix of the two tracks to null only in the high frequencies where the hiss resides. Furthermore, the inverted track needs to be gated so the nulling only occurs at low levels when the desired signal is not playing. This requires a "reverse gate/expander" that opens at LOW levels.

Yes, it is a long way around but it does work (I've used this approach for de-essing and it works great). I was just trying to point out that random noise is no longer random once it is part of a sampled waveform and can be canceled out.

But for noise reduction purposes a dedicate plugin/stand alone is definitely the way to go. Agreed!

[totalgear]

Cedar DNS 3000

[TBush]

I put a 'high cut' filter in Logic, and then automate it down as the note fades. Start at 20k and bring down to around 1.5k as needed- works great.