Dave and I had another rehearsal on Monday preparing for our inaugural gig as The Djinn (2 May – more details shortly!). As normal, I used my Zoom H2 to capture some recordings to help me do my homework. Unfortunately I had set the recording level too high at the previous rehearsal, creating a very distorted output, and I overcompensated, setting it far too low this time.
Audacity came to the rescue though. While the tracks each needed a boost of about 30dB to become reasonably audible, the results were fine (the Zoom is a wonderfully low-noise recording device). I was impressed at the rescue achieved but wanted to improve the results still further, removing extraneous peaks to let the rest of the track be boosted up to an even fuller volume.
By the end of the session, I had hit on a fairly efficient method for doing this:
- First I would solo the track I wanted to work on and, by listening into portions of the recording, trim it to the best performance (occasionally leaving some spoken comments as well). The shorter the track, the quicker it is to run operations like amplification across it.
- To give a start on improving the levels, I ran the normalise filter. This probably could have been done from the start because it includes the option to make the track as loud as possible without clipping. At this stage I could have used other tools to further compress the difference between loud and soft parts of the track and equalising over-dominant frequencies but, since the overall sound was good, decided to just apply spot fixes.
- I then made a duplicate of the track and amplified it a small way above the point of clipping (about 1.5dB into the red seemed enough). I had turned on clipping highlighting so every point where the volume reached the limit was highlighted with a red line. If 1.5dB produced lots of red lines, I would undo and repeat at a lower level (if it seemed worth it); if hardly any, a higher level could be tried.
- On the master track, I zoomed in close to each red line, selecting a fraction of a second of audio. I then pressed Z to find zero crossing points (where the waveform crosses the central axis – working in this way avoids introducing audible glitches) and used the amplify tool to reduce the volume for the selection by -3dB. This was quicker after the first time as I could just press Cmd-R to repeat the volume reduction with the same setting.
- I also applied the volume reduction to the duplicate track although without such precision (avoiding wasted time). This removed the clipping highlight so I could see what still remained to be done.
- Once completed, the master track could be amplified by the same level as applied to the duplicate but, since all the sections that were too loud had already been dodged down, the unpleasant sound of hard clipping was avoided.
- Finally, I made sure the start and end of each track were neat (using the zero crossing option and also applying a brief fade-out to smooth the end of each one).
For rehearsal recordings – to fuel practise and then be forgotten as the songs evolve – this was undoubably overkill. However, it was a good training ground for building my Audacity-fu for future projects!
Want to add something? Please join the conversation about this posting (nb. Yahoo! account required to log into Flickr).