Stay in touch…


Read the latest Bitstream

RSS Feed


Look for us at LinkedIn


Follow us on Twitter

Mix Magazine

This installment of The Bitstream column appeared in the July 2000 issue of Mix Magazine.

The Bitstream

This column discusses discusses digital audio ephemera…

An Ode To Engineering

Some things just aren’t right. Take, for instance, canned orange juice. Someone should really put a stop to that. (Tang’s another thing altogether...) You know what’s another evil that simply will not go away? Lying with numbers. OK, stretching “the truth” with numbers. Now we all know that a sufficiently motivated person can make any statistic look good to the uninformed. Even I, dear reader, has been known to inflate concepts a bit. All for the greater good, I assure you.

But, here’s what I’m talking about: All the new electronic gear, audio and otherwise, that you find in the marketplace is digital, with analog mostly relegated to the vintage or boutique buyer. From the Lightpipe–equipped low end for the DIY musician to the stuff sporting a MADI spigot and hefty price tag, its digital. And never has there been so much hype made from so little.

Let’s suppose that you’re in the market for some new equipment. Who isn’t, after all, always lusting after the shiny and new. Or, let’s say that you want to become more informed about your chosen vocation. Err, not playing scorching leads on your new Tele, Bubba. Being an “engineer.” Of course, the audio industry has it’s own reality distortion field whereby most so called engineers have never designed a single tool/engine/device in their entire life. I guess that means I’m an Automotive Engineer, ‘cause I can operate one!? Not quite. As you probably know, the engineer title we share with other, more technical brethren derives from a time in the distant past when poodles roamed the earth unfettered and we actually had to build and maintain our equipment. These days, all you have to do is head over to Sammy’s Sound Shack, fork over the plastic and walk away with your dream box of choice. All without a solitary shred of knowledge about how, internally, said dream box performs it’s miracles. Ah, the good ol’ days when men were men and…never mind.

Anyway, lots of products at Sammy’s are sold as “x bits” or such and such a sample rate but that really doesn’t tell you much. The truth is, as always, not black and white but numerous shades of gray. In general, all digital gear has the same basic advantages; good multigenerational noise and distortion performance with high amplitude signal, theoretically perfect archival longevity and an ever increasing price/performance ratio. For now, let’s look at that first, distinctive attribute, the one that’s fundamentally different from analog. That “perfect sound forever” thing.

A device that has a 24 bit I/O can be advertised as a 24 bit product. But wait, isn't the AES/EBU standard defined as having a 24 bit payload or essence? Yup, for as long as my pea brain can remember. Needless to say, the I/O word length has nothing to do with what happens to the data once its inside the box. With even the fanciest, you could have accidentally set the output word length to truncate to 10 bits with no dither and the AES/EBU output would still be a 24 bit output. Only the first 10 most significant bits would carry any useful information though. And, my, how bad it would sound.

Sample rate specs are also suspect, though less so. All things being equal, a product that operates at 96 kHz should have wider unity gain bandwidth than that same product operating at 48 kHz. Well, so what? I certainly can’t hear a sine wave above 20 kHz! But is that all there is to our sense of hearing? I can’t say with authority that the increased bandwidth alone “sounds better,” but I can say that my ears tell me a well designed piece of gear sounds better operating at 176. 4 kHz than it does at 44.1 k. There’s obvious theory that shows it should exhibit less group delay or phase distortion when operating at higher sample rates but no one knows the “true” reason or even if the “sounds better” effect is statistically significant.

OK, how about word length? That parameter is, in some ways, more obvious and yet more obscure. Our hearing has, on a good day, about 120 dB or 20 bits worth dynamic range. In the past, we’ve gotten along just fine with analog tape that had, if it was lucky, half that dynamic range. Of course, double ended noise reduction addresses that shortcoming but why do we need 144 dB of dynamic range in the AES essence if we can only hear 120?

It’s those dog gone numbers. A minimalist recording whereby a low noise analog source feeds a high quality converter and that performance is released, unedited, to the consumer wouldn’t need more than 16 bits to convey the information reasonably. But, most music isn’t produced that simply. Even classical releases, those pristine records of acoustic events, are usually edited so aggressively that one, two or three thousand edits in an hour–long performance is not unusual. So? In the digital world, almost any process performed on the data, whether it be editing, EQ, gain changes or mixing, results in longer word length data that what you stared with. 6 x 6 = 36. A one digit number time a one digit number equals a two digit number. Same with AES data. You make your 16 bit recording 1 dB louder and boom, you’ve got some extra low order bits filled with “the remainder,” as my grade school math teacher called it. You need extra “headroom,” as it used to be called in the analog days, to allow for the vagarities of day–to–day production. A 24 input digital desk needs to immediately pad any AES input by 18 dB just to maintain a reasonable gain structure for downstream processing and summing.

And that’s not all. Noise is usually spec’d as an absolute noise voltage or indirectly as a signal–to–noise ratio. Very rarely are an specs released that talk about the nature of the noise: it’s spectral makeup, it’s amplitude vs. frequency or whether or not it’s correlated with the signal. That last one can be particularly annoying and comes into sharp focus when lossy codecs are in da house.

Phase response, harmonic and IM distortion, these are also important gauges of how the device will perform but do not directly predict how it will subjectively “sound” unless the performance borders on pathological. On the other hand, that blanket statement isn’t true for jitter. Though most manufacturers fail to state anything about their product’s jitter performance, again it probably wouldn’t say much about the sound since, like noise, there are so many aspects of jitter behavior. But, in general, jitter has a lot to do with the subjective quality of multichannel audio as it heavily impacts the frequency vs. phase response, which in turn affects imaging and soundstage. Alas, few engineers care about such trivialities as soundstage and are content with their pan pot synthesized, two dimensional world. And, so is the consumer.

But, I wander from the subject at hand…And that subject is simply this: Don’t believe sales and marketing hype, good or bad. Start learning about the underlying technology on which you rely every day. The laws of physics have not been recently repealed just yet. Of course, if you’re the sort to buy a new car solely on it’s looks alone, then put the magazine down now and move away from the toilet. For most of us, however, purchasing new gear often means careful deliberation rather than impulse buying. You’ll have to live with your purchase for quite a while. If you don’t possess the expertise or aural acuity require to evaluate some box properly, then ask or buy someone who does. Your biz card sez “engineer,” right? Start acting like it!


Oliver Masciarotte is a consultant on new media content creation issues. He’s just suffered from a double shot of desert sun, with the NAB and SNIA conferences running back–to–back.