What is with these lossy compression fools and their tweezes to make squashed digital audio even shittier than it already is? Fraunhofer Institute? Whatever happened to that fine German engineering we always hear about? ;)

This "joint stereo" encoding nonsense should be avoided like the plague.

OK, you use the MPEG-1 Layer III format for your little desktop jukebox. Why you don't use MPEG-1 Layer II at high bit rates like you should, I'll never know. But what the fuck, you're a lemming, and it's too late for you and the rest of the schmucks out there in Net Land.

It's bad enough you're chopping the frequency response of your nice spiffy raw uncompressed PCM file. It's bad enough you're sucking the very life out of the beautiful reverb someone worked so arduously to get just right. Now you're going to screw whatever phase coherence they so masterfully preserved, to save a few fucking k.

You deserve the shit you get.

So what exactly is that mysterious "joint stereo" checkbox? The idiot savants at Digidesign have the balls to call this nasty procedure "Intensity Stereo Encoding." Do any of you actually know what this does to your already putrid MP3? It takes even more choices out of your hands.

This evil little algorithm actually decides which frequencies are common to both channels at a particular sample and compresses it as if it was one channel. Oh look, 1.21k is "in phase" here, so we can squash that sample even more and save you disk space.

Let the Old Engineer Who Still Pushes Buttons give you Clue One about phase coherence, children. Don't fuck with it. This procedure is very akin to what the Digital Compact Cassette format did, but with mpeg it's even more lossy compression on top of what you're already doing to the frequency response and complex amplitude curve by simply encoding it in MP3 format to begin with. It's deciding for you what can be squashed even more than it already is, thinking you won't hear the difference since you're already deaf from that fucking car stereo of yours.

Let me tell you a story. Many moons ago I had a teacher who engineered some recordings of a band called Moby Grape back in the late 1960s. One particular hit song was a breakthrough in mixing techniques. It utilized what we now call "M+S" recording, a technique using 2 carefully placed microphones in close proximity. By precisely canceling and reinforcing specific phase relationships between the 2 channels, when listened back in stereo it gave an amazingly "spread out" sound. Background vocals sounded incredible, truly larger than life. There was only one problem. Sum the channels to mono and the background vocals disappeared. Of course they had to find this out when they heard their song on AM radio. Oops! That pesky phase coherence.

For those who may not really know the actual concept of phase, when the "peak" of a waveform coincides with the "valley" of another waveform of the same frequency, the two waveforms combined will cancel each other out and create a null point. Similarly, two peaks can cause a "bulge" where that frequency is overly accentuated at that point in time. You can lose those nice backing vocals. Your reverb can disappear. Your EQ can be shot to shit.

Manipulation of phase relationships has been around forever, but coherence between channels is essential or your mix goes to hell. For years engineers have walked a fine line with this concept, especially broadcast engineers, because their mixes must sound good in mono as well as stereo. A good engineer always listens to his mixes in mono to make sure he's not screwing too much with phasing.

Everybody loves stereo. We happy-go-lucky Recording Engineers Of The 21st Century tend to record almost every element of a multitrack recording in stereo nowadays. It's hard to run out of playback tracks in our nonlinear disk-based recording world. But then even mono tracks can be "stereoized" quite convincingly with simple duplication, panning, delay, and some tweezing of the volume difference between the 2 channels. In any case the resulting stereo assault on the ears must not cancel itself or you are dead. I will kick you right out of the control room, and all the one-eared people in the world will despise you.

Now, back to this joint stereo business. So you have this precisely mixed masterpiece that you've slaved over, making sure every complex waveform you recorded stays coherent when they're all summed to 2 channels, or even one. Then along comes this German scientist who thinks he has every fucking mix there ever was figured out because of his math. Reverb that sounded like it came from God suddenly sounds like you're in a cave with your ears plugged. Precisely spread lead guitar that sounded like it was coming from two places at once now can barely be heard.

This may be an exaggeration of the effects of this cheating algorithm, but you really should do yourselves a favor and spend that precious extra few kilobytes of yours and encode each channel separately. Don't let some bozo decide which frequencies deserve more squashing. He's already decided which frequencies you don't even need to hear. Tell him to leave the ones you still have left alone.

NEXT RANT

HOME