I just did my first multi-track project with Audacity and notice that when I playback it, the master volume often hits the ceiling (goes red). From tape-based multitrackers in the past, I’m used to the approach of recording the individual track as strong as possible without clipping, and so I have done now in Audacity. But if I do this, I will have to cut back on the track volume in the mix to avoid clipping. Is this the way to go, or is it preferrable to aim att moderate recording levels, for example never exceeding the relative level 0.5 in the track graph, so that mixing can be done without the need for decreasing the track volume on virtually all the tracks? If you dont record individual tracks as strong as you can (without clipping), I suppose that you will end up (just as in the old analogue case) with more noise than necessary?
Jonas
You should aim for a reasonably high recording level, but there is no need to be as fussy with digital recording as the noise floor of the recording medium (digital data) is incredibly low. Just make sure that the peak level never ever reaches 0.0dB (as that will cause clipping).
For multi-track recording it is best to use 32 bit audio (set this as the default in the Preferences) as this allows you to scale volume levels up and down with virtually zero loss in sound quality.
With the new 1.3.8 version of Audacity there is a new “MixerBoard” feature (“View” menu) that makes this very easy. Note this is a brand new feature and still has some rough edges, but it works pretty well. Just adjust the sliders down for each channel as necessary.
For earlier versions of Audacity you can select all tracks (Ctrl+A) and use the Amplify effect to drop the level of each track by, say 6dB (or whatever amount is necessary). If you are using 32 bit recording there will be virtually no sound quality loss from doing this. You can also use the track volume sliders (this is lossless) or a combination of these two techniques.
thanx for prompt answer, Steve, interesting that you should mention the 32 bit, which I have been wondering about, experiencing that on 1.3.6 the playback of just 4 tracks (obviously entailing implicit realtime mixing) sometimes produces intermittently garbled sound, despite having a rather new pc with quite decent specs. Allegedly, 1.3.8 is more efficient, allowing for more tracks before things start to go awry. I was thinking that if the running goes to heavy on resources, one way to ease up on it should be to decrease to 24 or 16 bit, but then I suppose scaling will not work as well?
Yes! I just saw the mixerboard command. Looks nice, but when looking closer, it seems at little else than a alternate representation of the info already in the track-headers? 
Regarding the noise floor, isnt the analog part of the sound card also part of the equation here?
(Funny enough, I seem to have 2 “soundcards” on my machine, from the factory. One that’s built-in on the motherboard (Realtek) plus a PCI-card from Creative called XB X-fi, while non of them seem to amount to much apart from supporting home-cinema multi-channel systems.)
With a reasonable spec. computer you should be able to get better than that. That’s about on par (or slightly less) to what I get on My 500MHz PIII.
There are a number of places that data bottlenecks can occur - a common one is the hard drive, so plenty of free defragmented space on the C: drive is a must. Also, cutting down on other processes to leave as much resources free as possible, but more commonly, shutting down any programs or processes that will hog resources.
Anti-virus programs can cause problems if they check every .au file as it is being recorded. If you have an anti-virus program running, make sure that it ignores .au files.
Program updates (such as Windows Update) can be a headache, so check that these do not run while you are recording. There are other optimising tips if you search the forum, also some here: http://wiki.audacityteam.org/index.php?title=Tips#Recording_Tips
It’s much easier to adjust levels on 8 tracks when laid out side by side than scrolling up and down the tracks. Also, the individual track meters are nice.
Yes, and that is why you still need to keep the recording level reasonably high, but you will probably notice that if you set the recording level so that the peak is at say -6dB, the noise floor may be at say -76dB (these figures are just for illustration). If you then drop the level so that the peak recording level is -12dB the noise floor will probably drop to -82dB. In both cases the dynamic range is 70dB (the noise floor is 70dB below the peak level). If your hardware is capable of a 70dB dynamic range, then it does not much matter if you have recorded that from -96dB to -26dB or -70.1dB to -0.1dB, the signal to noise ratio (SNR) is the same.
Microphone inputs on inexpensive sound cards are usually very poor quality. The dynamic range for the microphone input on my new laptop (on-board sound card) is around 57dB which is about the same as cassette tape without any noise reduction (not very good) and increasing the recording level raises the noise floor of the microphone pre-amp in the sound card. A good quality sound card should give SNR of well over 90dB making CD quality recordings possible on a home computer.
The dynamic range of 16 bit audio is 96dB, so obviously if you set the recording level very low then the sound quality will suffer. The dynamic range of 32 bit audio is very much greater.
In practice, if you record in 32 bit and keep the recording level reasonable you do not have to worry too much about maximising the recording level.