It’s not just that numbers are open to manipulation. It’s just that all too often, the numbers quoted don’t actually mean much. If you want meaningful measurements of a loudspeaker – the sort of measurements that might actually help you when it comes to room and system matching – then an impedance curve and a low-frequency sweep measured in an anechoic chamber are pretty much your starting points – yet both are comparatively rare these days.
If nominal impedance is next to meaningless, how about implied performance? How many speakers claim ‘phase coherent’ crossover designs? How many of those speaker manufacturers publish square-wave outputs for their products? It seems like there might just be a gap between the theory as described and reality.
“…and not everything that can be counted counts.”
But nowhere is the appeal of numbers more important, more powerful or more shamelessly exploited than digital audio. From day one, digital has been a numbers game, a tendency that’s rearing its head with a new and utterly cynical vengeance. Think back to the early days of CD and the ‘over-sampling wars’. First you had to have a player with 4x oversampling. That was the acid test of high-performance and technology. Then it was 8x oversampling. The Cambridge CD2 broke the double-digit barrier, offering 16x oversampling – and we all just lapped it up. Not only that, we’ve never looked back. Oversampling rates have hit three-figures, bit rates have doubled and sample rates have gone though the roof, while the advent of file-replay and ‘high-res’ streaming has driven the numbers game into a frenzy of inflation and hyperbole. The message is simple, the bigger the number, the better!
‘Hi-Res’ has become a catch all, to the point where some buyers will only invest in ‘Hi-Res’ titles. Yet repeatedly, I find hi-res files that sound worse than standard 44.1 equivalents, while file-replay struggles to match the performance possible from a decent optical disc player. If the whole ‘Hi-Res’ argument holds water, then that simply shouldn’t be the case. Unfortunately, all of the evidence, both in terms of process and results, indicates that a Hi-Res sticker is no panacea for musical success. Yet the whole industry and not a few customers seem to be happily hitching their wagons and wellbeing to the ‘Hi-Res’/streaming train. It’s become an audio article of faith that “higher resolution equates to higher quality”, but that is a mistaken assumption. It fails on two levels: the semantic and the practical/operational.
Let’s deal with the question of terminology first. The words “higher resolution’ imply more information and that can only be good, no? The problem is the distinction between information and data. A higher sampling rate will absolutely generate more data. But for that data to constitute information, it has to be rendered intelligible – and that’s a massive distinction. Consider for a moment, the actual form of music. It consists of notes, each a dominant pitch and harmonics, spaced precisely, one relative to another. You could express it as a row of dots, amplitude on the vertical axis, time on the horizontal. Now add second, third and fourth instruments. They will each generate their own row of dots, but depending on where the musicians are sitting relative to each other, the arrival time of each individual’s notes at the microphone(s) will vary, introducing a third ‘depth’ axis and creating (in any one instant) a three dimensional box, or more properly a stream that continually flows from musicians to microphone.
I can hear what you are saying…
Of course, a single point mic and a small group acoustic recording is the simplest model. Start adding multi-tracks, effects, reverb, delay or overdubs and things get a whole lot more complex. But either way, the thing that identifies this sound as music (as opposed to noise), is the pattern that the notes create: a pattern that the recording needs to capture, the system needs to recreate and your ears/brain need to recognise. And pattern is the key. Information – sound – that is in the right place (in terms of time, amplitude and delay) helps to build the pattern and create an intelligible picture. By definition, anything that isn’t where it should be, when it should be, the size it should be or the pitch it should be simply confuses and obscures the pattern. In other words, data points that don’t fit the pattern are noise, not information.
