Audacity Forum

Hi folks... I know a lot of people have questions about sound normalization type stuff, so sorry if this has already been asked. But I'm digitizing these old tapes of lectures, and the lecturers tend to walk away from the mic for minutes at a time. When I'm lucky I can just normalize these sections. But often there are lots of short loud noises that prevent this: saying one word directly into the mic, hitting the mic stand, tape pops, etc. Stuff the click remover won't remove no matter how extremely I set the parameters. I can always find and reduce each noise before normalizing, but that can take a lot of work. Or I can amplify and tell the amplifier to ignore clipping, but the clipping can sound pretty bad on a lot of these noises. So here's what I imagine I need: an effect that would simply amplify all samples by a given amount, except for parts that would clip, which it would compress to avoid bad distortion. Does such a thing exist?

(I would have thought that I could approximate this effect by setting the compressor threshold just above the speech in these quiet sections, and then setting the ratio to maximum. For example, I tried setting the threshold to -24dB and the ratio to 10:1, and clicked the normalize-to-0 box. I figured that anything that was -24dB before would end up no quieter than -2.4dB after. But it just doesn't seem to work that way... the clicks still end up way louder than the speech. Maybe I just don't understand how the compressor works.)

Thanks for your help!
--Allen

The threshold has to be set higher than normal presentation, but lower than the microphone smack. The additional gain reduction (ratio) will only apply to the noise.

Koz

Allen,
I would do the same as I understand you did, and I think it is the same what Koz says.

And the result should be that the loudest sounds will be not 20 dB (or 24) abouve your treshold and normal sound, but only 2dB (or 2.4dB) above it.
This is actually what you write to be the result -- right?
Much louder changes to a bit louder, right?

Might be that the clicks are indeed at most 2.4 above, but because they are abrupt and unwelcommed, you percieve them as louder?

1. It is possible to construct 'negative compressor' that would reverse the volumes. Infact when constructing a compresor, you can end up with that because just confusing few wires (or number in computer). Is that perhaps what you want?
Perhaps some compressor accepts (by mistake) a negative compression ration, or some nyquist compressor might be made to accept it.

2. Usual compressor needs a bit of time to react to high peaks. Cheap one could perhaps let some peaks through, a good one does not but makes very_short smooth distortion when changing the amplification ration.
Computers can look ahead, and there was recently mentioned on the Forum a looking ahead compressor.

Thanks to you both! I'm glad for the confirmation that my thinking is along the right lines. (Actually, I think now that what I was imagining is exactly what the compressor would do with different parameters.) So perhaps I'm just misjudging the average levels of my speech, or the relative loudness of the clicks after compression. Or perhaps it's the looking-ahead issue... Since the compressor had an "attack time" parameter I figured it must be looking ahead, but that was probably a bad assumption since I don't really know what that parameter does. But I did see that mention of the other compressor, and I'll check that out. --Allen