Audacity Forum

Well the subject says it all really. I realise that you don't want it maxing out cos you get clipped audio but what is the ideal place where the meter's should be peaking. I ask because my USB sound card appears to need a much higher input in Linux than it did in windows. In windows I was able to tweak the input level from in windows but i dont think its possible in Linux so I have to do it from my mixer. I want to set the optimum level before I start recording, hence the question.

Cheers

It's a little more exciting than that. You have to figure out where everything in the sound chain overloads. It doesn't do any good to have good comfortable headroom in Audacity when you're beating the prunes out of your microphone amplifier and it's creating fuzz way upstream. If you're doing live performance capture, you should probably never get any closer to zero than -3 to -6. You can tune and compress this later, but the instant you touch zero in either direction, you have crunchy audio and it's difficult or impossible to fix.

You also need to know that the Audacity peak meter only measures the positive cycles of the audio. If you, like that woman announcer on the local FM station have a lot of your voice go negative, you could be cheerfully peaking at -3 and still have crunchy, badly overloaded sound.

Live capture is tricky.

Koz

Ok, thats what I thought, I'm not really capturing 'live', I record mixes from my decks. I was just worried that if I wasn't recording at an 'optimum' level then I might not be getting the best out of the capture (if that makes sense!)

So basically as long as its up around that end of the scale without 'crunching' then I'm still going to get a nice capture (based on my bedroom dj setup and my dodgy DJ skills that is)

As Koz says, you don't want to near the zero otherwise clipping occurs immediately. In this respect digital recording is unlike old analog tape recording, where the physics of the mag tape enabled brief excursions into the "red" without damage to the sound.

BTW you can get a better view of the signal levels by expanding the meter bars, click and dragging on the meter bars will do this - I have mine stretched across the whole screen.

WC

hp1 wrote:Well the subject says it all really. I realise that you don't want it maxing out becos you get clipped audio but what is the ideal place where the meter's should be peaking. I ask because my USB sound card appears to need a much higher input in Linux than it did in windows. In windows I was able to tweak the input level from in windows but i dont think its possible in Linux so I have to do it from my mixer. I want to set the optimum level before I start recording, hence the question.

Cheers

1. geting over 0dB means clipping = severe distortion
2. For 16bit recording, down at -96dB you have digital (quantization) noise.
So this is your range. You must keep stricly below 0dB, and you have to be far above -96dB.
(With lower quality sound card, there actually might be analog noise anywhere above the digital quantization noise,
which makes then range smaller ! Moreover the analog noise might get much higher whan changing the input level by card driver, as you say you did in windows. This specifically applies to 'mic gain boost' options in new laptops internal sound card, which can make the recording unusable, depending on the model.)

If you almost touch 0dB and you have good sound source and sound card, you get what trully can be called CD quality.
If your peaks are at -16dB, the digital noise is actually 'only' 80dB below your signal peaks. You should decide yourself if this is good or not.

If I record sound that I know pretty well, I set peaks to -6dB.
For live recording, I sometimes set peaks to about -18 dB.

The most important is to avoid 0dB clipping and then we have to compromise the danger with how much we know the situation and if later level adjustments are possible. I can imagine once I will preset the level to -40dB if I will have to leave the recording alone, and then I will know that the recorded speech will be audible , at least

Read also post of WC, just above this one.

http://kozco.com/tech/audacity/AudacityPanelFull.jpg

That's my piano solo showing pretty much perfect waveforms and sound meter.

Koz

Your ear works a little funny at volume changes.

Double or half level, 6dB either direction, is a change barely noticeable by most people. 3dB goes right by everybody. This gives you the odd idea that in order to hear a significant difference in your living room, you have to double the size of your sound system, not go from 80 watts to 90 watts. Nobody can hear that. You need to go from 80 watts to 160 watts.

Nobody is going to hear the change from -6 to -3, and nobody is going to hear the change from -3 to 0, but that's the insanely dangerous one. 0dB is a hot stove. Don't touch you should get such a blister.

Most people hear half volume when the sound goes down 20dB, 10 times less. Odd indeed, but this is a survival condition. You can also hear a twig snapping on the other side of the forest which is 1000 times quieter than a person standing next to you talking. That would be the animal trying to eat you.

You can play with this. Are you intent on making your shows "legal" but louder than normal? Try Chris's Compressor and bump the compression ratio up from where it is (.5??) to .77 or higher.

Chris's Compressor
http://pdf23ds.net/software/dynamic-compressor/

This also has the advantage that it will even out the volume variations you put in by accident.

Koz

kozikowski wrote: You can also hear a twig snapping on the other side of the forest which is 1000 times quieter than a person standing next to you talking. That would be the animal trying to eat you.

That's -60 dB, perhaps the person got angry on me.

Is the noise the beast? Or artifacts? Or distortion carefully hiding and moving only during sound peaks ?

Wow, thanks so much for these responses. I have to admit my knowledge of the science of audio is pretty limited so its been great to read these detailed posts.

My question has most definitley been answered. Can I happily record my mixes at -3 or -6 and not have to worry about it being too quiet or missing too much of the dynamic range? Yes.

Thanks

There are tools that let you play with that.

Normalize lets you push your whole performance up (or down) so the highest peak hits a specific level. Many people use "0" dB for that, but knowing the problems that 0dB causes, I would get no closer than -1. The built-in value that Audaciy 1.2 uses is -3dB. 1.3 lets you select the value.

Unfortunately, Normalize affects left and right differently which can lead to stereo direction problems. To get the same effect and fix the relationship between left and right, use the Amplify tool.

Koz