Features

The most notable aspect of Benchmark Media Systems' DAC3 HGC ($2195), which I favorably reviewed in the November 2017 Stereophile (footnote 1), is its low noise floor. John Atkinson's measurements corroborated Benchmark's claim that the DAC3 is capable of "at least" 21-bit performance. While significantly less than the theoretical potential of a 24-bit data format, 21 bits is still the state of the art, and corresponds to a dynamic range—the ratio of the highest achievable digital-domain volume to the DAC's internal noise—of 128dB. That's well above the dynamic range that most power amplifiers can achieve. A good-measuring high-end solid-state amplifier is likely to have a dynamic range—the highest attainable ratio of signal to noise—of about 100dB ref. its maximum power.

Benchmark created an amplifier that more than matched the DAC3's dynamic range: the AHB2 ($2995). Bridged to mono, the AHB2 has a claimed A-weighted dynamic range of 135dB (footnote 2). Kalman Rubinson reviewed the AHB2 in the November 2015 issue; it now sits in Class A of Stereophile's "Recommended Components."

With the DAC3 HGC and the AHB2, Benchmark seems to have achieved the lowest-noise, lowest-distortion system of source and amplification on the planet. The signals this combo sends to loudspeakers should be cleaner—truer to the source—than any other audio components can achieve. Wanting to know what that means musically, I asked Benchmark to send me a pair of AHB2 amplifiers to complement the DAC3 HGC already in my possession.

Setup
This article is unusual in that my goal was not to take the measure of a particular component but to determine whether Benchmark's entire low-noise system could resolve musical detail that a more typical audiophile system, with noise and distortion levels that are low but not as low, could not. More broadly, I wanted to determine whether the two systems sounded meaningfully different—especially in ways I could correlate with the presence or absence of distortion and noise.

I set up two systems independent of each other except for the data source (my NUC-based server, which runs Roon ROCK) and loudspeakers (my longtime reference DeVore Fidelity Nines).

I configured the Benchmark system for optimal technical performance, with the DAC3's internal output pads disabled for the greatest possible voltage. In this configuration, the DAC3's maximum output is an extraordinarily high 18.3V—too high for any other amp-and-speaker combination I'm aware of, but appropriate for the AHB2s at their lowest gain setting: a very low 9.2dB. Benchmark does things this way because the combination of high source output voltage and modest amplifier gain is optimal for minimizing noise and distortion. The AHB2's two other gain settings, of 17 and 23dB, can be used in more mainstream systems. I used the Benchmark amps' bridged outputs because, in contrast to most other amplifier designs, the AHB2s get even quieter, by 3dB, when the amplifiers are bridged.

The speaker cables, Benchmark's own, employ SpeakON connectors on the amplifier end and, on the other, banana plugs from River Cable with a nifty screw-down locking mechanism—a great cure for loose-fitting banana plugs. I used the power cords Benchmark provided; they're nicer than most stock cords. Balanced interconnects were high-quality Canare star-quad microphone cables (L-4E6S) with Neutrik connectors. This is a professional-audio system, so why not use pro cables?

In addition to the Roon ROCK and DeVore speakers, the comparison system consisted of PS Audio DirectStream DAC and BHK Signature preamp and, variously, PS Audio BHK 300 and Pass Laboratories XA60.8 monoblocks. All connections among these fully balanced components were made with Clarus's top-of-the-line Crimson interconnects and Clarus speaker cables. The ROCK was hooked up to the two DACs with USB links from AudioQuest. Switching between systems was quick and easy: disconnect one pair of speaker cables, connect the other pair, and switch zones in Roon.

But first, levels had to be matched. Starting with my main system, I adjusted the volume so that the musical peaks of the recordings I planned to use were loud but tolerable. Then I played the pink-noise track from Stereophile's first Test CD (Stereophile STPH002-2), measuring the level with a sound-pressure-level (SPL) meter. I hooked up the Benchmark system and adjusted the DAC3's volume control so that the white-noise SPL reading was within 0.1dB of the other system's level. After setting the levels, I left the volume controls alone. The maximum system SPL was 92dB (footnote 3).

The Art of Noise
If you follow John Atkinson's measurements sidebars, you know that just because a digital audio format is 24-bit doesn't mean you get 24 bits of audiophile goodness in your listening room. The analog side can't match what digital is doing—AC line noise is especially pernicious. JA's measurements of the DAC3 HGC corroborated Benchmark's claim that it's capable of about 128dB of dynamic range, equivalent to about 21 bits. It probably isn't possible to do much better than that.

But what about the rest of the system? I've already mentioned power amps: If noise and distortion limit the system's signal/noise ratio (SNR) to about 90dB—that's pretty good—then even under otherwise optimal conditions, noise and distortion in the amplifier will be higher in level than, and so potentially mask, the quietest 38dB of the DAC3 HGC's output (footnote 4).

This is the problem the AHB2 is intended to solve with its claimed dynamic range of 135dB—that's the ratio of the noise (inputs shorted) relative to the full output voltage. A more relevant figure is 0.0003%, the AHB2's total harmonic distortion plus noise at full power, relative to the maximum output voltage. That corresponds to an SNR of about 110dB. That number should be better bridged, and JA's measurements show that just below full power, the distortion is lower by at least a factor of 10, so it should be possible to achieve significantly better SNR by listening just below maximum volume. Which means that 120dB is in sight.

Assuming, then, that there are no other noisy components in the signal path—no preamp—the AHB2 delivers to your speakers more of the resolution and dynamic range of the DAC3 (or other high-performing DAC) than will most other amplifiers.

What happens after that depends on your speakers and your room. Loudspeakers are distortion factories, especially in the bass. Rooms, too, have a dynamic range. The top of the range is set by the highest volume you can tolerate—or, since music is supposed to be pleasurable, the highest volume you can enjoy. The bottom of the range is set by . . . what, room noise? Yes, but precisely how that works is not entirely clear (no pun intended).

By convention, in-room noise is referenced to an SPL of 20µPa (micro-Pascals), which corresponds to 0dB, the quietest sound the average human ear can hear—though that's sort of average, ignoring how the ear's sensitivity depends on the frequency: in the highs and lows, the volume necessary for audibility is much higher than 0dB, reaching 73dB at 20Hz; in the presence region it's significantly lower, plumbing –7dB at 3kHz.

In my room, on a good day, with windows closed and appliances turned off, I measure an SPL of a few dB above 30dB. That's typical for a "quiet rural area," according to one widely distributed chart, so it's pretty damn good for the middle of Manhattan. On a bad day, with service trucks idling along my stretch of Riverside Drive with pumps or generators running, or with heavy rush-hour traffic on the road itself, the noise level climbs above 40dB, which the same source says is the "lowest limit of urban ambient sound." The noise in my listening room can hit 50dB at times, usually weekends, and late nights in summer when the building's boiler fan isn't running. That's equivalent to a "quiet suburb, conversation at home," or "large electrical transformers at 100 feet."

So there's electronic noise inside components and there's room noise. How they relate depends on your speakers' sensitivity and the volume setting (footnote 5), but in general room noise is louder. To risk stating the obvious, the widest in-room dynamic range is usually achieved at high volumes. But 110dB SPL is the pain threshold for the average human; to go above that is to risk permanently damaging your hearing. An SPL of 110dB is, I think, a reasonable upper limit to the high end of your room's dynamic range. In real rooms, music's quiet passages, and quiet sounds within louder passages, will always lie below the room's noise floor.

---

Footnote 1: Benchmark Media Systems, Inc., 203 E. Hampton Place, Suite 2, Syracuse, NY 13206-1633. Tel: (800) 262-4675, (315) 437-6300. Fax: (315) 437-8119. Web: www.benchmarkmedia.com.

Footnote 2: Bascom H. King, designer of PS Audio's BHK monoblocks, is quoted on Benchmark's website: "I measured less distortion and noise in the Benchmark AHB2 than in any other of the many power amps I've measured over the years." Not only does the Benchmark amplifier have very low distortion; that distortion gets even lower as the power increases, to quite high levels.

Footnote 3: Room noise, then, was higher in level than all but the loudest 60dB of music.

Footnote 4: I'm ignoring the effects of noise-shaped dither, which can improve effective dynamic range beyond 6.02dB per bit.

Footnote 5: This isn't simple, though. When you turn the volume up or down, you're altering the electronics' SNL—the dynamic range in practice.