Boulder 1012 D/A preamplifier:
Page 2 The 1012's inputs are fully programmable; my unit displayed "Krell" for analog input 1. A subliminal message? No, van Allen sighed, just the remainder of a demo he'd done earlier with Krell gear. But the user interfaces (naming inputs, adjusting gain and phase for each input, etc.) on many newer components are really well-thought-out, making settings more accessible—fun for the whole family! (The best of the user interfaces is still BAT's, I think, closely followed by Mark Levinson's.)

Like many components with audiophile pretensions, the Boulder 1012 features fully regulated and separate power supplies for the analog and digital circuits, which I consider de rigueur these days for high-priced audiophile gear. In another bow to modernity (I was informed by van Allen's e-mail), the 1012 features "optical decoupling of the digital and the analog boards." Now that seems to be a good thing to me, knowing as I do just how ugly ground contamination can sound.

Reading further, I was encouraged to see that the digital circuits are actually shut down while analog sources are active—as they should be. Interestingly, Boulder says that all volume and balance settings are accomplished in the analog domain. There is no digital volume control in this paean to high technology. As in Boulder's 2010 preamplifier, the 1012's volume control uses discrete resistors, is controlled by microcircuits, and enjoys 200 "super accurate" 0.5dB steps to cover its 100dB range.

You have to get the hang of dialing in the right volume on the remote. It doesn't change slowly, as with the Mark Levinson No.32; or more quickly when the button is held down, as with the BAT VK-50SE, which also features a unique and useful Fade function (it fades back to where you started when the Fade button is pressed again). No, the 1012 is a veritable speed demon via its remote; when adjusting the volume, I often overshot the mark, which necessitated backtracking to a more civilized listening level. (Usually, after K-10 has cranked it up!) I'm not complaining, for heaven's sake, just noting it for you Type-A audiophiles out there—like me!

Up and over, it's in—and the Knicks do it again!
So what's the Boulder's claim to fame? It can't be just another 16-bit/44.1kHz processor, can it? No fear. Just to gleefully muddy the waters a bit more, Boulder calls their processing system "Upandoversampling, Boulder Style."

I feel a migraine coming on. Boulder's take on the high-speed racket is that they want to achieve the "maximum bit rate that the DAC can accept without introducing the ringing associated with analog filter theory."

As Boulder explains it, 24 bits at 96kHz equals a bit rate of 2,304,00 bits per second, 192kHz equals 4,608,000bps, and the blazing Boulder does 705.6kHz, which works out to 16,934,000bps! Kinda like a Formula One car coming off its jacks after an eight-second pit stop, wheels already spinning, launched back onto the track in a screech of fishtailed tire smoke.

And please, let us remember, it ain't over till the fat lady sings. And she doesn't let fly till you're feeding the 1012 a 48kHz datastream, which results in an astonishing 768.0kHz out of the processor. That's the maximum bit rate the Burr-Brown 1704k DAC can take, so for 88.2kHz and 96kHz sample rates, the 1012 halves the oversampling ratio to 8x.

According to JA, Boulder's "Upandoversampling" doesn't appear to be different to the 16x oversampling used by Krell and Theta in their processors from a few years back. However, it is realized using 21st-Century circuitry. According to the e-mail I received from van Allen, I learned that a "fundamental goal" Boulder had in mind—and reached, he said—was avoiding rounding off the math. Boulder's 64-bit math ensures, van Allen asserted, the highest practical resolution of 24 bits theoretically available today.

This kind of computing power, according to Boulder, requires a very powerful DSP engine. The 1012 uses the Texas Instruments 320C6701 chip, capable, van Allen asserts, of 1.2 gigaflops, or 1.2 billion calculations per second. Ninety percent of that computing power is used by Boulder to do the impressive math for the processor—the calculations to upsample to 24 bits at a minimum speed of 705.6kHz in one step. Boulder claims this one-step processing eliminates "introduced jitter through addedsignal path and connections." Van Allen continued his technical description by stating that Boulder's algorithm is an "Eigen Type," applicable only in the digital domain. "It has no analog corollary, and is the closest thing to a perfect digital filter." He's not exactly shy.

Boulder has always been a fan of modular amplification, since they first used the late Deane Jensen's 990 discrete op-amp module in the mid-1980s. The 983 gain stage used in the 1012 is a "less complicated but still very effective" version of their "cost no object" 993 gain stage, as found in their 2050 amplifier (reviewed in September 1998, Vol.21 No.9). Analog filtering of the 1012's DAC section is accomplished via a three-stage Bessel filter, which is said to minimize group delay.