This is not really a specific Audacity issue, but I have a constant internal (to my own head) debate on this issue. And there is another somewhat-related Audacity issue that I have just started seeing.
I ripped all my CD library many years ago, mostly just using Windows Media Player. I recently got a new car that plays music very nicely form a USB stick, so I have been playing around with that a lot. Unfortunately, the car system does not use Replay Gain tags, so I have been noticing the big volume differences between some of my albums. The caused me to start looking at many of them with Audacity, since the initial wave display is an instant indication of peak and average levels.
One thing that has surprised me is that a fair number of my CDs stored in my computer library show a lot of clipping lines in Audacity. Sometimes it is just a peak here or there, and for other CDs, the entire track is filling the display with almost more red than blue. This really surprises me since they are ripped CDs where I have no control of the recording gain, and I never expected a commercial CD to clip.
So now I am re-ripping those offending CDs, using Exact Audio Copy and automatic normalization, but I am a little bugged that the normalization is applied individually by track, with no option to preserve original track levels where the artist may have really wanted that contrast between tracks. I have always chosen to normalize the entire album together for digitized LPs and tapes, even though I realize it makes no difference at all if I only play specific tracks.
So this is not really a big deal for most types of music, but I was wondering if other folks are having this same internal debate? And what should I be doing to avoid those really hot CDs if I do not want to normalize by track?
If the CD is clipped normalizing doesn’t remove the distortion.* It just “hides” the clipping from Audacity. Audacity is not analyzing the wave shape. It’s just checking the peak level. Audacity checks for potential or possible clipping.
but I am a little bugged that the normalization is applied individually by track, with no option to preserve original track levels where the artist may have really wanted that contrast between tracks.
I did not know that! (I’ve never used that feature.)
and I never expected a commercial CD to clip.
That shouldn’t be too common but sometimes they do get too aggressive in the “loudness wars”. Some people call it “clippression”. Clipping is a (bad) kind of dynamic compression.
Of if you made MP3s (or other lossy format) it can be a side-effect of the MP3 compression. MP3 changes the wave shape making some peaks higher and some lower (without affecting the sound of the dynamics). And MP3 can go over 0dB without clipping so it might not actually be clipped. If you play the file at “full digital volume” your DAC will clip at 0dB but I’ve never heard of a case where that clipping was audible.
If you normalize to -1dB before MP3 compression that usually keeps the MP3 below 0dB. (I don’t bother with that and many of my MP3 “show red” when I open them in Audacity.)
Unfortunately, the car system does not use Replay Gain tags, so I have been noticing the big volume differences between some of my albums.
…So now I am re-ripping those offending CDs, using Exact Audio Copy and automatic normalization,
As you probably know, normalizing won’t fix that because normalizing uses the peak levels and “loudness” does not correlate well with peaks. There is WaveGain and MP3Gain which “permanently” adjust the loudness of the file so the player doesn’t have to do anything special.
Or you can use Audacity’s Loudness Normalization to match/adjust loudness. But it’s a little trickier because you have to choose the “correct” target loudness and “manually” check/adjust for clipping.
\
Audacity has a Clip Fix effect but when I tried it a couple of times it just made the waveform look better with no noticeable effect on the sound. And if it’s a commercial CD (as it was in my case) it’s probably not simple-clean digital clipping… It’s probably over-compressed and over-limited in all kinds of terrible ways!
As you probably know, normalizing won’t fix that because normalizing uses the peak levels and “loudness” does not correlate well with peaks. There is WaveGain and MP3Gain which “permanently” adjust the loudness of the file so the player doesn’t have to do anything special.
Yes, I do understand the difference between peak levels and perceived loudness, but this effort is producing some good results nonetheless. I have found quite a few albums where the recording level was set considerably below the max (but apparently fairly compressed anyway, without many high peaks), and normalization is producing noticeable gains in the overall loudness for those.
And for the really hot CDs, normalizing them to 90% on re-rip puts their loudness more in line with the majority of other music. In reality, the overall level probably needs to be reduced to about 80% on those, but I just cannot bring myself to go that far, since I want to leave the EAC settings to work well with any disc without having to change it each time. 90% seems to correlate fairly well with the -1 dB setting in Audacity.
I know that is not correcting anything in the recorded wave forms, just reducing the overall amplification of each track. I have been quite surprised at how many of those I am finding across many genres. For example, several of Eric Clapton’s albums are just driven to absolute max (not his rock stuff, but some of his blues albums), and even one from Faith Hill, with EAC reporting peak levels of 100% on most tracks (and Audacity showing a solid blue wave form with relatively little amplitude variation).
I do not really like Replay Gain tags, and very few players use them anyway, but car playback is one of the very few places where I would have found them useful. And I absolutely am not going to permanently modify the loudness gain for anything in my main library. I have tried using loudness adjustments on tracks I convert to .mp3 for the car, but have not been particularly happy with those results either. I will be content to just normalize them all to the same level and then deal with the perceived loudness just like we all have for the past hundred years or so.
Audacity is not analyzing the wave shape. It’s just checking the peak level. Audacity checks for potential or possible clipping.