Compressed digital audio has become a way of life for Broadcasters. Audio compression works by being destructive, but that doesn’t make it your enemy. Compressing audio saves on disk space and network time, but that doesn’t make it your friend. So, how do you live happily with Compressed Digital Audio?

John Maizels reports.

Early in my professional IT career, I sat in a presentation while a very senior colleague told a gathering of our customers that Product X, the operating system they were about to implement, was just like “The Green Slime”. Our customers laughed, while those of us responsible for the product fidgeted nervously and checked the exits in case a quick getaway was needed. It was.

“Once you bring Product X into your installation”, he said, “it’s like The Green Slime and (he paced this very deliberately) it never quite.. goes... away... One day you’re going to think ‘OK, time to upgrade to another technology’ and that’s when you discover that you’re stuck, locked into a massive data collection and all kinds of processes that can’t be converted. And there you are, with your entire business tied up. So you’d better understand, before you start, what you’re getting into”.

Just like The Green Slime, so it is with Compressed Digital Audio. Seduced by the sales pitch of more audio on less storage, we let this compressed stuff in through the door. Sure, the Green Slime – sorry, Compressed Digital Audio – is cheap, light and effective. Every tech store has a product which lets you load 25 million songs into a 10kb USB dongle and send those songs to someone else in 3.5 seconds on a dial up line. OK, teensy exaggeration. Anyway, in a few years from now (and maybe now) most of the financial and technical constraints which forced us to compress audio for storage will no longer apply, and when that happens… guess what: compressed audio will still be with us in our archives, our libraries, and our transmission systems. Forever, probably, which is sort of a long time.

There is nothing inherently wrong with compressed digital audio, but it comes at a price, and the price is Loss Of Quality. No two ways about it. You can losslessly squeeze about 25% of the space out of audio and it will be exactly the same. But to get meaningful savings, a part of the desired signal also has to be tossed away. Consider that most MP3 compression is at 128kbits/sec… so 90% of the original bits have gone.

Fortunately, with a bit of care and understanding you can get results which sound good, and which won’t leave you with a slimy archive. And that’s what this article is about.

But before we start, a digression into some history. Don’t be put off by the numbers in the next section; you can skip it, but the discussion of the myths makes more sense after a little bit of background.

Well, “uncompressed” would be the obvious, spectacularly unhelpful answer. Uncompressed analogue audio was what we used to have in the golden ages, which is any time before 1990. Back then engineers had heated arguments about whether 20Hz to 20kHz was a good enough frequency response; could 0.5% distortion be tolerated, and would we call that Broadcast Quality? It seems that we did, because pretty much any audio system which achieves those specifications will sound pleasant enough and musical to the ears.

The invention of FM-Stereo in the Thirties forced us all to accept that no frequencies higher than 15kHz would be transmitted, a necessary evil associated with the pilot-tone process. While this undoubtedly makes harpsichords, triangles, and dog whistles sound a bit less tizzie, it’s not a showstopper for enjoyment. The average LP met the standards of Broadcast Quality, and FM audio was only just slightly worse.

Digital audio first arrived with consumers in the form of the CD, a piece of plastic which holds slightly more than an LP’s worth of uncompressed audio as bits, provided the bitrate is kept to 1.5Mbps (mega, or million, bits per second). At that bitrate, we can achieve our traditional audio frequency response (20Hz-20kHz) at 16-bit resolution for 76 minutes and everyone is happy. Or at least, happy enough with 1.5Mbps that it became embodied in a heap of other technologies too. Here’s a water-cooler fact: when we talk about a CD drive burning at 40-times speeds, we mean that it’s 40 times “1.5Mbps”.

Those of you who flick off a Sudoku each morning will have jumped ahead and done the math already. If 1.5Mbps is full speed, then an MP3 at 128kbps (kilo, or thousand, bits per second) is one-twelfth of full speed, or 12:1 compression, or 8% of the original size. We’ve thrown away 92% of the original signal, and it’s gone for ever. That’s gone, and forever, and permanently missing in action; it’s not magically found again when the signal is decompressed.

The ONLY reason that lossy audio compression sounds OK is because the stuff that is tossed away is stuff that the brain wouldn’t notice anyway. It’s a three-card trick of the finest order, and if you’re really careful and clever you can do it more than once and the audience STILL won’t notice. But audio compression is still a trick and, just like bad accounting, if you do it often enough someone eventually notices how you did it and the game is up. The decompressor has to recreate (read: “invent”) all the missing stuff that was deleted in the compression pass. That job can be done well, or badly.

1 2 3 4 5