Submarine remastering: The secret, inferior version of Pearl Jam's "Vs."

Despite no documentation, there are at least two separate masterings of Pearl Jam's LP "Vs." floating around. One of them is clearly inferior to the other. Simply put, one of the masters was highpassed at around 40hz, with the resulting digital overs clipped, to generate the new master. Such a master suffers from both reduced bass and greater amounts of clipping.

I haven't seen any documentation, either online or on the printed packaging of "Vs." that I just saw at the store, to suggest that this remaster ever took place. Therefore I'll dub it a "submarine remaster". Just like in the bad old days of vinyl, when different pressings of the same record might have used vastly different mastering practices with their only documentation being their dead wax info, a submarine CD remaster may be considerably different from the original, with a different pressing indicated by a different ISRC.

Here's a plot of the eq differences between the two masterings, measured from the first 60 seconds of "Animal". The bottom axis is frequency in hz; the side axis is attenuation in db. A 1-second window was used with a 50% overlap. At 20hz the masterings differ by 12db, so this ought to be quite audible with a suitably good system.

Here's a zoom-in comparison of the waveforms, illustrating the two stages of clipping in one of the masterings, where one of the clipping stages matches the clipping found in the other mastering. This may not be audible, but it certainly could have been avoided and is unfortunate.

Unfortunately, I don't have ISRCs for the two pressings, but I am 95% certain that the bassier/less clipped master is from the original 1993 CD, and the newer/more distorted master is from the 2004 Sony reissue. (Update: After speaking with greynol further and comparing notes, we now have no idea exactly which master corresponds to which release. Both of our CDs predate 2004, but his has the two clipping slopes identified in the red plot above, while my CD has the single clipping slope of the black plot. Comments here affected by this have been struck.)

The pressings both ReplayGain to the same value - no gain or added dynamic range compression was applied. I don't know which version is used for iTunes/Amazon music downloads. Looking at a copy of the Vs. CD at a store, there is no visual indication on the outside of the case that it is either the 1993 or the 2004 pressing.

So what's my point? Besides being a very odd example of a "remaster", and one that's worth posting about, the circumstances surrounding this release suggest that no notice may be given if a newer pressing of an album is a remaster. Those looking for the "authentic" version of Vs., and certainly the one with the most bass, should almost certainly look towards the original 1993 CD.

More generally, if you're looking for an "authoritative" master of a back catalog CD, it may still be important to buy the original pressing, even if a newer pressing does not advertise any remastering. Or, if the newer pressing surreptitiously corrected a flaw in the original, vice versa. But if the CRCs match between two pressings, then the newer pressing matches the old one exactly. And If they don't, it's still possible the differences are trivial, like a difference in track offsets. A thorough analysis must be done to concretely identify the differences between each master. Got that?

Thanks to greynol for the heads up.

Update: See above for new notes on pre-2004 pressings.

Think classical music mastering does not employ clipping? Think again

This is a waveform plot from the opening two seconds of the third movement of Bela Bartók's second piano concerto, conducted by Pierre Boulez, recorded in this decade, and released by Deutsche Grammophon in 2005. There is an opening drum hit, shown here, that is fairly unequivocally clipped near 0dbFS for 121 samples (2.7 ms). This is a large length of clipping by most standards.

In many music listener's minds this would be considered a gross mastering error. But the sound quality notwithstanding is spectacular, and the long-term loudness plot shows no signs of hypercompression (or indeed, perhaps any compression): the peaks of the music are merely being sawed off. However, I have no independent proof showing that the recording has not been compressed in other ways.




A very similar style of mastering can also be found on two other recent Boulez discs: one of Varèse, and another with his own ..explosante-fixe... and Notations (the latter played by Aimard). These recordings consistently brickwall-clip at or near 0dbFS, for between 4-100 samples, but otherwise are mastered clean as a whistle. But another recent DG CD of mine (Aimard's performance of Bach's Art of the Fugue) has no clipping whatsoever.

It's important to note that, as I've mentioned in the past, drum transients are perhaps the kind of signal most forgiving of clipping/limiting. As they are already very spread out in the frequency domain, the added smearing from this kind of clipping could be rather effectively masked. Listening to this music, there is no "snoking gun" kind of artifact that lets me say with certainty that this distortion is audible.

Nevertheless, that DG would choose to make this sort of compromise in their modern releases is troubling. I used to consider Boulez's recording of Amériques one of my reference tracks for critical listening; I'm considering searching for alternatives now.

Original thread on HydrogenAudio.

Update 19 May 2009: Steve Hoffman notes that the recent DG CD of Boulez's Maher 8 is also clipped.

Metallica's "Death Magnetic": Clips on both CD and Vinyl

Much hay has been raised about the sound quality of this album, although interestingly enough, none of it has been from mainstream music journalism (or even the blogosphere) yet. The mastering engineer has even disowned the sound quality of the record, passing the buck onto the mixing engineer, and ultimately, the band itself.

The MP3 prerelease leaks have been so commonplace (and so bad sounding) that a lot of people are buying the 2LP and 5LP versions thinking that they are getting superior sound quality. However, judging by the large amount of clipping still extant on the vinyl, they probably won't:

Image credits: hdsemaj on stevehoffman.tv. Original post.

Subjectively, some people are reporting that the warmer sound of the record is dampening the clipping somewhat, but really, that's damning with faint praise.

My recommendation is to spend as little money as possible on the release until they sell a better master. Metallica.com is streaming the album for free right now. Beyond that, buy the MP3s. Personally, I smell a rat.

Some thoughts on a new technique for clipping detection

I decided to riff on an idea I had for smarter ways to detect clipping. I'm sure it's not a new approach - a quick search on Google pointed out several papers on the same basic concept - but I'm not aware of it being used for audio, or for mastering evaluation specifically.

Clipping clamps the signal to a constant value. It also tends to occur right in the middle of signal content which is of a high power. If the signal derivative is calculated, the DC component of the clipping is effectively eliminated, bringing the values to 0, while the values before/after the clipping are relatively unmodified. (Specifically, low frequencies are attenuated and high frequencies are boosted.) So naively, one could use this derivative as the basis for a clipping detector - compare y' to y, and if y' is zero or very close to 0 while y is of a high power, you may have clipping. This technique would be immune to attenuated clipping - if it occurs at -10db it should work as well as if it is at 0db.

However, this approach fails when gross frequency response distortions are introduced - like what exists on vinyl. As discussed earlier, vinyl clipping examples exist which are sloped, not flat. The derivative of these a sloped line is a constant nonzero value. The workaround for this is simple: take another derivative, the second derivative, so that this constant nonzero value collapses to 0. In theory this could be extended to an arbitrary number of derivatives, but because high frequencies are amplified, background noise tends to dominate the response after the 2nd derivative, so the 3rd and beyond are pretty useless for vinyl analysis.

What I'm ultimately hoping for is to have the final output be a histogram and running a threshold on that to give an estimate for how many clipped samples exist in the signal. This allows comparisons between signals that are not sample-aligned (as is usually the case with vinyl vs. CD comparisons).

Here's what I have so far. First, some clipped stuff from Leyendecker again, on CD:

Note how the clipped content is neatly crushed to 0 on the second derivative plot - and most importantly, that the histogram plot shows a lot of values on the left, outside the distribution curve with a high signal energy and a low 2nd derivative energy. Those are the clipped samples.

Now for the LP version, different part of the file:

The 2nd derivative crush still occurs, but it's not nearly as prominent as on the CD. And the clip signal is so weak on the 2nd derivative plot (or the noise is so high) that none of it really shows up on the histogram plot; there's no real indication of clipping there.

But it's a start.