Every few years a product comes along that changes the way you work. It has just happened again. If your studio contains any equipment that produces digital audio signals, you need to know about a new IC chip, the AD1890 from Analog Devices
Most engineers sooner or later experience trouble copying signals of different formats and incompatible sample rates. Larger studios face the problem of synchronizing all their digital equipment. Fully digital solutions to these problems are very expensive. This will change soon, since manufacturers are working now to bring you products based on this new circuit.
The AD1890 is a stereo sample rate converter that costs less than $50. It comes in 20-bit professional and 16-bit consumer grade versions, but here we will be looking only at the pro version.
Okay - but how does it sound? Many people believe that great specs mean great sound. This is not necessarily true, though specs can often give you some good hints. There were two separate prototypes available for test, both built around the AD1890. It is reasonable to assume that production boxes will use a very similar circuit, since there are only a few ways to hook up this device.
Most people will use the AD1890 as a sample rate converter, so the test converted audio from 48 kHz to 44.1 kHz, as will as the reverse. The reference system was a high-quality D/A-A/D converter setup that I prefer for analog mastering. For a listening environment, I used our mastering facility, which includes two time-aligned monitor systems, custom amplifiers, tube electrostatic headphones, and similar tools of the trade. The source material was mostly original master tapes along with some CDs. Music ranged from rock to classical organ music.
The decision? Very close, except at the extreme top, where the the AD1890 was more detailed. The analog system lacked a bit of clarity by comparison. When I turned the playback system way up (on a quiet passage), I found the noise floor of the AD1890 to be noticeably lower. This might explain where some of the extra detail came from: less interference. Regardless, my frame of reference shifted, and the analog system now feels like a very subtle special effect to be used for adding top-end smoothness. The AD1890 is a trustworthy reference, and that is what studio engineers need.
It won't be long before you can get your hands on commercial processors based on the AD1890. Check one out on your own source material. In the meantime, rumor has it that Analog Devices can't get people to return their evaluation systems. This sort of backhanded compliment signals a winning product. And me? Of course I'll return the prototypes. Just as soon as I remember where I mislaid them. Check back with me in 1999.
Analog-to-digital converters and their opposites, digital-to-analog converters (DACs), require a perfectly stable reference signal, called a clock, to work correctly. Both consumer SPDIF and professional AES/EBU data formats require that the clock be encoded into the same signal as the audio information. This allows one wire to carry everything needed to produce a stereo image on your speakers. When the data reaches the receiving end of the wire, the clock is separated from the audio. The DAC then uses the clock to reconstruct the audio signal. In a perfect world, we're done.
But what if the wire is too long, or of the wrong type? Or the method of extracting the clock - or even the DAC itself - is sensitive to minor problems with the incoming signal? This is not theory, it happens frequently. Many studios run audio through multiple conversions from analog to digital and back again, often through cabling that is less than perfect. Suddenly the clock is not a stable reference, it is "jittered." The DAC can no longer do it's job properly. Welcome to the Twilight Zone, where output does not equal input. Clarity is lost as your mix passes through what you assumed was a perfect audio transmission system.
Here is where the AD1890 shines. First, it is extremely forgiving of variations in the incoming data. Second, the signal it sends out is controlled by a clock that is completely isolated from the incoming clock signal. It "re-clocks" the data stream, and any jitter that was embedded in the original just plain disappears. The output is an improvement over the input.