A visual comparison between the 2010 UMG Exile remaster and two of its predecessors
With all the talk about the UMG remasters being louder and more compressed than the Virgins, I decided to rip a few tracks to my HD to compare. I don't actually have the full Virgin version of Exile on CD, but I do have two different sources for Tumbling Dice - Jump Back, which I bought in 1995, and Forty Licks, which I bought in 2002. Expecting these two versions to be identical except for volume normalization, I was planning to post only the results of my comparison between Forty Licks and the UMG remaster. Things didn't entirely go as planned, however...
First of all, let me show you what Tumbling Dice looks like on Jump Back:
[
www.inter.nl.net]
As you can see, the dynamic range is okay. I wouldn't call it huge, but this isn't classical music anyway. For all intents and purposes, the dynamic range is fine (although one might wonder if it could've been better still). Anyway, here are the numbers:
Channel Average level Peak level
Left -13.2538 dBFS +2.8141 dBFS
Right -12.8093 dBFS +2.8127 dBFS
It's hard to judge if any of the peaks are clipped. Some of them might be, but I can't tell for certain. Even if they are, the clipping is limited to a few isolated instances.
Well then... It's time to move on to Forty Licks:
[
www.inter.nl.net]
This looks a little troubling, as the dynamic range seems to have been reduced. In fact, the song is also audibly louder than on Jump Back. Dynamic range compression was certainly employed here. I'm not exactly sure what's going on here, because I used to think the second CD of Forty Licks simply used the 1994 remasters. The differences here go beyond simple normalization, as is evidenced by the numbers:
Channel Average level Peak level
Left -10.5933 dBFS +3.0010 dBFS
Right -10.7323 dBFS +3.0010 dBFS
Suspicious things are going on, especially with regard to the channel balance. On Jump Back, the right channel was louder than the left channel, but now the left channel has become louder than the right channel. Furthermore, both the minimum value and the maximum value are equal for the left and right channel, and the minimum (i.e. most negative) value and the maximum value are each other's bitwise inverse. Believe me, that's never going to happen if you just perform a straight transfer from analog tape to a digital medium! It's not a very bold statement to make that some peaks must've been clipped. By how much? We don't know, as that information is lost forever... Well, it's not really lost forever as the original files are presumably still stored somewhere, but for the *listener* it is. One could try to line up both versions of the song to investigate this, which should be possible if they came from the same digital transfer, but I'll leave that for another occasion.
It gets worse. Obviously, I'd almost say. Lo and behold, the UMG remaster:
[
www.inter.nl.net]
OUCH! Everything seems to be clipping against an invisible wall! Also, where did our dynamic range go? And once again, the song is audibly louder than on both Forty Licks and Jump Back. Will the numbers confirm our hearings and sightings? Of course they will:
Channel Average level Peak level
Left -7.8538 dBFS +2.7014 dBFS
Right -8.2445 dBFS +2.7014 dBFS
How's that for a change? The UMG remaster has the lowest peak level, but the highest average level - by an impressive margin. Congratulations UMG, you've just won the loudness war! Sadly, this inherently means you've also lost the fidelity war. As you can all see in the picture, the song was first compressed and limited and then scaled *back* as a means of volume normalization. This shameless butchering is the hallmark of "modern" mastering techniques. Let me tell you a little secret: the maximum peak level attainable on a CD (for negative values) is actually +3.0106 dBFS! This means that UMG could easily have made this track nearly a third of a dB louder than it already is without any further loss of fidelity (in fact, it would improve fidelity - although I'd call that a moot point). Of course, it'd ruin the volume balance between songs, but that could've been fixed by a better dynamic range compression approach. It always saddens me to see that good music is first pumped up to the max to sound as loud as possible and then scaled back to perform volume normalization, leaving part of the medium's dynamic range unused. Sadly, the easy approach seems to be preferred these days. I'd rather have them not perform any dynamic range compression during the mastering phase at all, but if you're going to do it, at least do it properly and don't clip peaks any more than necessary. By the way, did you notice that the left channel is now far louder than the right channel? And no, I didn't accidentally swap them, in case you were wondering.
I'm not saying only UMG is guilty of this - everybody seems to be doing it these days. I've also kind of given up hope that the tide can be turned, but it never hurts to point out that this is in fact occurring. The music we all love is being ruined at the mastering desk.
One remark I have to make is that I didn't compensate for the different lenghts of the pregaps. If I had done that, the most significant difference would be that the average level for Jump Back and Forty Licks would've been slightly higher relative to the UMG remaster. Between the two of them, they are pretty close in terms of length, although there's a time shift - but that doesn't affect that average level.
You can also look at the file sizes of the images, just for kicks.
Jump Back: 146414 bytes
Forty Licks: 143634 bytes
Exile UMG: 97405 bytes
All the clipping makes the waveform look so simple that the PNG compressor needs far less space to store it. Does the file size of these PNGs have a one-to-one correspondence with fidelity? Of course not, but it's still telling.
Edited 1 time(s). Last edit at 2010-05-20 17:36 by FreeBird.