Why, Why, Why???

The Arcade Fire chose a digital distribution model for their new album, The Suburbs. Digital-only is $7.99. CD plus download is $12.99. The deluxe edition is a double 12″ vinyl plus download for $24.99.

Downloads are available in 320 kbps MP3, FLAC, or Apple Lossless. There is also an AAC version that has synchronized visuals. Fulfillment was handled by Topspin. Downloads were available the day before the album’s official release.

The vinyl release is obviously important to the band. They mastered the tracks to lacquer disk, and then digitized the disks for the CD, with the goal of preserving the vinyl sound for both the LP and CD releases.

All of this is awesome. Arcade Fire is a dynamic band and a new release is always exciting. I pre-ordered the album (LP deluxe edition) quite some time ago. But when the album arrived, the frustration began.

I downloaded the FLAC and MP3 files. My plan was to use the FLAC files to burn a CD for quality listening and the MP3s for listening in the car.

Unfortunately, the MP3s contained no ID3 tags, which is annoying but not necessarily wrong, and I could fix that Easily enough. But before starting, I decided to open the FLAC files in a audio editor. And then the fun began.

Just about every track on the album has oddly placed track points – in the middle of fades and nowhere near zero crossings.

I’ve seen small mastering errors on CDs for years. It’s not uncommon at all. CDs are effectively linear devices with small index points added for the convenience of finding individual tracks, much like the visible bands on vinyl records.

When everyone listed on vinyl, CD, or even cassette, where the album listening experience was essentially linear, if the track points were a little bit off, it was no big deal. Putting your CD track index 200 samples into a song would hardly be noticeable.

Unfortunately, the album is dead. Individual songs are king. I don’t like this, but it’s true. Even for me – most of my listening is done to MP3 copies of music ripped from the CDs I purchase.

MP3 is a windowed compression format. Its algorithms depend on having chunks of conveniently-sized audio samples to work with. And music is rarely edited to exactly fill these chunks, so what does the MP3 encoder do? Add silence.

If you take a standalone song, which fades up from silence at the beginning and fades down to silence at the end, you will never notice the extra little bit of silence. It might only be 2 ms, 2/1000 of a second. Hardly matters, right?

Well, thanks to the Beatles and Brian Wilson, since the late 1960’s, some artists have viewed the album as a long-form listening experience (which I like), and have effectively eliminated the gaps between songs. One song might cross-fade into another, or they may just plain overlap.

When you have a crossfade, the track index point on a CD has to be at one precise moment in time – one sample, which represents 1/48000 of a second. (20 nanoseconds, 0.0000283 seconds.) A 20 nanosecond long point becomes (say) 2 milliseconds, or 100 times longer in duration. This you can hear. You hear song A fade out, a small pause, and then song B fades in. And that’s best case.

Worst case is the track index comes (say) 10 milliseconds early. Now, you hear Song A fade out, a pause, then Song A continues to fade out while Song B starts to fade in. To me, this is very disruptive to the listening experience.

“But I don’t hear pauses between my tracks!” you exclaim. Are you listening to your music on iTunes? You have an iPod? Yeah, I knew it. iTunes cheats. It’s playing tricks to get rid of those gaps. It cross-fades EVERY song transition. It overlaps everything and these short gaps get played over by other audio.

It’s a clever solution to the problem, actually. But it messes with the album’s intent a little teeny tiny bit. Maybe you don’t notice. Maybe the artist doesn’t notice. Maybe no one does. But that little quarter-second cross-fade plays DJ with the first and last 125 milliseconds of each track. If there is silence at the head and tail of each track, then the gap between tracks is now a bit shorter. If there is a crossfade, now you’re not hearing all the music.

Artist’s intent and all that. Musicians who still work at an album level should be horrified that the MP3 player is changing their music to compensate for the change the MP3 format is making to their music.

Little changes can often have a bigger impact than you realize. When I discovered Bob Dylan (from an 8 track copy of Greatest Hits Volume 1), I was a poor high school student and I couldn’t afford to buy new records. So I visited every library that lent records in town and listened to every Dylan record I could. The copy of Desire that I had contained skips in the Black Diamond Bay track. I listened to that record (dubbed to cassette tape) for years, and though the skips were annoying, I got used to them. One of the skips was very subtle – it perfectly skipped an entire line of one chorus, and I didn’t even realize it was there. When I got the album on CD years later, I was glad to not have the skips, but that one chorus bugged me for years because it just didn’t sound right. It has taken a long, long time for me to get used to Black Diamond Bay the way it was actually mastered as opposed to how I heard it on that borrowed record.

But I digress. Back to the iTunes “trick”.

If the tracks are loud enough or quiet enough (rock and roll or pop music), you probably won’t notice the crossfade. If the transition is a pure tone, or something close like orchestra strings (classical), or sustained brass horns (jazz), you definitely will notice.

Unfortunately, this cannot be avoided. It’s the nature of the technology. Once the audio is encoded to MP3, the gaps are in there. If my car played FLAC files, which do not suffer this same problem, I would use that format. Instead, for heavily cross-faded albums (The Decemberists’s Hazards Of Love, Wilco’s Yankee Hotel Foxtrot, etc.) I dispense with tracks and encode the album as one big long track. A little inconvenient if you want to listen to one particular part, but I don’t often do that with these albums.

I get bugged by the worst case scenario, when track indexes are not in the optimal locations. It’s most annoying is when there is silence between tracks, and the track index is still in the fade out or fade in of one of the tracks. This drives me nuts. And it is the sign of a mastering engineer who is either sloppy or doesn’t know what they’re doing.

Another sign is track indexes that don’t occur near a zero-crossing. This can cause a “pop” when the track begins, even if it occurs in relative silence, as it can make a very sharp transient, which often contain very high frequencies. The digital-to-analog converter has to try and construct a waveform that fits all of the sample points, and sharp transients often result in large spikes. Starting at or very near the zero crossing greatly reduces the chance of a rough beginning to a song.

I fix these types of mastering errors before making my listening MP3s. I know this is a little crazy. I partly do it because I can – I happen to know what I’m doing and have the tools to do it. I know I’m a cranky middle-aged man with little patience for idiots. This drives the missus nuts watching me fix things she wouldn’t even notice or care about. But I do.

I fixed the mastering errors on the new Arcade Fire album. I even decided to join two tracks (Half Light I and Half Light II) because they really were two parts of the same song. Then I made MP3s and tagged them.

A lot of the songs are cross faded. I thought they mastered every song to its own lacquer disk. Why did they cross-fade anything? Whatever – if that’s the sound they want…

The new album hasn’t yet grown on me. It’s not as immediately memorable as their first two albums are, but I like it and have been listening.

Then I got the e-mail.

There’s a problem with the downloads, they say. Here’s a new download code and a link to a page where you can enter it.

There is no place to enter anything on the page. Whoops – HTML error. The next morning, I’m able to download a new version of the FLAC files. I don’t bother with the MP3s because I suspect the mastering errors are going to be there.

Oh, crap. That means I had to go fix all of them again…

Fortunately, I was able to use the first set of FLAC files as a reference, but this was a pain in the butt for an album I don’t love (yet). And the really annoying thing is I can’t even tell where the “problem” was. Sheesh. Thanks, Arcade Fire.

(One of the tracks was actually a different length, which probably had something to do with it.)


So, I’m doing a lot of complaining here. What do I think should be done? Master everything with silence in between? Abandon the album altogether? That’s a bit extreme.

There are no easy answers. But mastering engineers should be aware of how albums are going to be listened to. Sure, it won’t matter how careful they are when the album is listened to on CD, but what about the iTunes purchase? Or the Amazon MP3 purchase? Or when that CD is ripped to MP3 or AAC?

If your album is a collection of songs, keep them separated.

If you want to make album-length pieces of art, be careful where you put the track indexes, or just dispense with them altogether.

I am increasingly convinced that artists should be making two different mixes of their music – one for people who want to listen to CDs in a fairly quiet environment at a decent volume, and one for people who want to listen to iPods and in their cars.

This may sound crazy, but it’s not. You think the sound mix you listen to of a movie on DVD or on TV is the same as what was used in the theater? No way. They’re different audio environments, so they are best served by different mixes. Why should music be any different?

This is probably a losing battle. I would guess that the vast majority of music is listend to in cars and over crappy earbud headphones, and not in a quiet listening environment. Which is too bad, because the average home theater sound system today is probably hundreds of times better than the high-end “hi-fi” system of the 1960’s.

Between advancements in audio production and reproduction technology, we have the ability to listen to music of higher fidelity today than at any time before – but we end up listening to overly compressed (in both senses) music that sounds like crap.


Post a Comment

Your email is never shared. Required fields are marked *