The Conversion Problem(s)

The reasons you might be interested in such a device have to do with the nature of digital audio. For a variety of reasons we won't discuss here, we use several different sample rates, and naturally, being human, we've simply got to have different digital languages too. Further, and in accordance to Murphy's Law, the digital source you need to work with probably has a different language (we'll call them formats) and sampling rate than the rate/format you need for your particular job. Hence, you need a little black box that'll do the conversion/translation.

For the case of sample rate conversion, you need a box that will do the necessary arithmetic to change the number of samples (and therefore digital words) per second while maintaining the integrity of the data over time. While this would intuitively seem to be a fairly straightforward process, in fact the math is what my neighbor Tom Bates calls "burdensome," because resolution accuracy is a direct function of the number of numbers crunched and 20 bits of resolution requires a heap o' crunchin'.

In any case, Analog Devices makes a chip called the 1890 which does this very thing to a very high level of performance (20-bit resolution and related clock stability), and, after we opened the Anatek up and peered at the innards, Tom Bates is pretty sure that that chip is the guts of the Anatek.

Meanwhile, AES/EBU is the so-called 'professional" digital format, while S/PDIF is the so-called "consumer" format. Aside from a few issues pertaining to copy protection and the like, the two formats are operationally similar, except they transmit at different voltage amplitudes. The Anatek SR-7 makes the language conversion (the manual describes explicitly what happens to the various word-bits) as well as the voltage conversion. One useful thing it does in the process is to strip away the copy prohibit bit in the S/PDIF format.

Interestingly, if all you want to do is AES/EBU-to-S/PDIF format conversion, Canare makes a little in-line transformer that'll do it just fine for virtually all professional applications.

So Why Don't We Just Run the Device Into Analog and Back?

There's another way to do this stuff, a way that's cheap, simple and to the point. Patch the analog outputs from the digital source to the analog inputs of the digital receiver. This permits any sample rate conversion and completely bypasses the format issue. Worries about analog noise and distortion are unfounded - we're talking straight wire here (noise and distortion at -130 dBV).

The real issue is, rather, the quality of your digital-to-analog and analog-to-digital converters. How much do you like 'em? If you're wild about 'em, go ahead and do it that way. If you're dubious, or not really sure how you feel, than maybe the Anatek way is better. Think of it this way: the chip in the Anatek is equivalent to a state-of-the-art 20-bit D/A and A/D converter set. The errors introduced by the Anatek may be a lot less than the errors introduced by your stock converters, unless they're really primo. If that's a concern for you, and it may be a reasonable one, once you get to a certain level of operations, then the Anatek is a better choice. It isn't the analog that hurts you, it's the quality of the conversion process.