I’m editing a live music show of 2 hours, the current project I’m working on most songs from the master source are already clipped eg. peak volume of the show +0.36/+0.08dB so I am NOT using normalize.
The DC Offset for the whole 2 hour show says -4.242% / -0.925%.
My question is if I correct the DC Offset when the volume is already clipped will applying DC Offset cause any audio to be lost / cut off?, for example the WAV file I have applied DC Offset to you can see the top channel now has a gap at the bottom where it has moved up, so considering the top part already looked clipped and it’s moved up even higher since DC Offset has been applied does that mean the top part of the audio has been cut off?
In this case is it better to leave as is and not fix DC Offset?
Also in the case of a 2 hour live show should I use DC Offset by applying it to the whole show as 1 single large 2 hour file?, or should I do it as separate individual tracks/songs?
By default, Audacity uses 32-bit float format, and always uses 32-bit float for processing (sample values are 32-bit floating point numbers).
One of the main benefits of 32-bit float over ‘normal’ formats (such as 16 or 24-bit integer) is that values beyond 0 dB are supported. To demonstrate this (assuming that Audacity’s default Quality settings have not been changed:
- Launch Audacity
- Generate a sine tone, frequency 440 Hz, amplitude 0.8
- Apply the Amplify effect and set the “New peak amplitude” to 6 dB (+6). Note that you will need to temporarily enable “Allow clipping”.
Note that the peaks and troughs of the waveform go way off-screen.
If you play the sound, it has a “buzz” quality rather than a smooth mid-pitched “hum”. This is because sound cards work in integer format and cannot produce peaks over 0 dB.
- Open the Amplify effect again, un-check the “Allow clipping” checkbox, and set the “New peak amplitude” to -1 dB (minus 1), and apply
Note that the peaks and troughs are fully restored, and if you play the sound you will hear a mid-pitched “hum”. By default, Audacity works with 32-bit float tracks, so although the sound after step 3 was badly clipped when played, the actual audio data in Audacity was undamaged.
What you should do with your recordings is to apply the Normalize effect to set the peak level no more than 0 dB (preferably a little below 0 dB) and to correct the DC offset. This will ensure that all of your audio is within the “valid” range for integer audio (always used for playback by all known sound cards), and does not have DC offset. See: http://manual.audacityteam.org/man/normalize.html
Yes correcting the offset can possibly make the clipping worse, but not if you normalize. And of course, you can’t repair the existing clipping. But, hold-on…
If you are over 0dB, you’re NOT (digitally) clipped because digital clipping is caused by the 0dB integer limit. I’m guessing you have MP3s? MP3s can go over 0dB without clipping, but your DAC will clip if you play them at full digital volume (because your DAC takes integers as input). Audacity can also go over 0dB temporarily/internally. The red in waveform indicates potential clipping.
…so I am NOT using normalize.
It’s OK to normalize! FYI Normalize will change the volume up OR down as necessary. Amplify will default to the same thing.
Note that MP3 is lossy compression and the wave shape changes… Some peaks get higher and some peaks get lower. Sometimes when you normalize to 0dB and then make an MP3, the peaks go over 0dB can you can get potential clipping. If you rip lots of CDs to MP3, many tracks will go over 0dB and show potential clipping in Audacity.
Also in the case of a 2 hour live show should I use DC Offset by applying it to the whole show as 1 single large 2 hour file?
Apply it to the whole file. If you correct/change the offset in the middle of the file (especially during quiet/silent parts) you’ll get a glitch (a “tick” or “click”) where the DC silence level suddenly changes.
Thanks to all for the info.
Not using MP3’s, it’s WAV, not sure why there are peaks over 0dB as the recording originates from an unknown person, who knows what they’ve done to it!?!!!
OK so I think I understand correctly that even though there are peaks above 0dB eg. loudest peak is +0.36dB and quite a lot more over 0dB, if I Normalize to -0.01dB (which I normally do with low volumes) then Audacity will correctly fix DC Offset even though it’s normalizing down from a peak over 0dB…?, I just wasn’t sure about Normalizing DOWN seeing as it has peaks over 0dB, I do know that the existing clipping cannot be fixed though.
Thanks for the info everyone.