Benchmark DAC3 HGC D/A preamplifier-headphone amplifier
Much has been written about the divide in high-end audio between subjectivists, who trust their ears, and objectivists, who believe that anything not scientifically proven is fake news. I respect both sides and am skeptical of both extremes, and I like to think that’s how most audiophiles feel. High-end audio is about experiencing musicthat’s the whole pointbut scientific and technological rigor lie behind every real advance, past and future. I regret the cynical snake-oil salesmanship, bad thinking, and clumsy engineering that pervade certain parts of our hobby.
A couple of days ago, a friend invited me over to help him solve a problem with his system. He’d installed a new preamp and was getting a lot of AC hum. We traced the problem to the (balanced!) interconnects connecting the new preamp to his DAC; their stray capacitance and inadequate shielding made them an effective antenna for random, stray electrostatic fieldsnot what my friend was wanting to listen to just then.
It gets worse. While both channels of the interconnect pair caused hum, one caused louder hum than the other (we’d switched channels to make sure). This demonstrated unacceptable quality control. Punch line: Those cables retail for $2500/pair (footnote 1).
When my friend called, I’d been deep into writing this review of Benchmark Media Systems’ DAC3 HGC digital-to-analog converter, which lists for $2195$305 less than a pair of those interconnects. The contrast between Benchmark’s relentless focus on engineering (about which I will have more to say forthwith) and this cable manufacturer’s negligence was striking.
Benchmark is so focused on engineering that the company routinely releases new products with changes they don’t expect you to hear. They tell you that upfront. When they released the DAC3’s predecessor, the DAC2, they published an Application Note titled “Benchmark DAC2 vs. DAC1Is There an Audible Difference?” Their answer: “The noise and distortion produced by [the DAC1 and the new DAC2] is well below audibility in normal operating conditions. Nevertheless, there are some situations where the DAC2 can provide an audible improvement.”
What other hi-fi company would introduce an important new product, with big improvements in specifications and several important new features, by questioning whether those improvements are likely to be audible?
Some audiophiles might read that Application Note, shrug, and move on to another company’s more exciting pitch: If the designer doesn’t think it sounds better, why should I be interested? But others, including me, are reassured: I’ve spent most of my life around scientists, and I’ve grown to love science’s traditions of precision, self-deprecation, and understatement (footnote 2). Besides, there’s already enough nonsense in audio. It’s comforting not to have to worry if a company is trying to win you over with false claims and fake technology. High-end audio would do well to emulate science’s humility and meticulous integrity.
All that aside, Benchmark’s apparent faint praise for their own new product is understandable in another way: They were comparing it to its predecessor, the Benchmark DAC1.
Fourteen years ago, with characteristic eloquence, John Marks introduced the DAC1 to the audiophile world in these pages and established it as the rear guard (price-wise) of audio’s high end: “There are less-expensive CD-playing solutions, but I think that the Benchmark (and anything else with similar performance) is at the watershed pointunquestionably among the hills that lead to the highest peaks.”
The DAC1 was added to Class A of Stereophile‘s list of “Recommended Components.” I bought one, as did JA. A lot of other people did, too. It remained in Class A until last year, when Benchmark discontinued the last DAC1 variant. And although John Siau, Benchmark’s Vice President and Director of Engineering, downplayed the DAC2’s sonic superiority to the DAC1, audio writers found it pretty noticeable. In the February 2014 issue, musician and longtime DAC1 owner Erick Lichte very favorably reviewed the DAC2, which then joined the DAC1 in “Recommended Components”but in Class A+.
The DAC1 was a tough act to follow. So, apparently, is the DAC2. Where would the DAC3 land?
The DAC2 was a more significant upgrade than the DAC3, and the changes it contained are still very relevant. The DAC2 added these to the DAC1: an on/off switch; a polarity switch; native DSD conversion (single rate); support for 192kHz PCM; a sophisticated digital/analog hybrid volume control with lower noise; asynchronous upsampling to a higher frequency (211kHz), to make room for gentler anti-alias filters for high-resolution data; a redundant digital architecture for a lower noise floor; and Benchmark’s solution to the problem of “intersample overs.”
I’d never heard of intersample overs. I learned that digital engineers first became aware of the problem in the late 1990s; credit for raising awareness of it goes to Søren Nielsen and Thomas Lund, then both with Danish pro audio company TC Electronic. In the years since, Nielsen and Lund have published several papers on the subject.
When recording and mastering engineers push digital recordings up toward maximum volumeright up to the red line, but not past itthat ought to be okay. But because of the way a DAC’s reconstruction filter works, most modern DACs sometimes give recorded levels a little shove past the maximum. Usually this happens when the filter interpolates between two high-level samples, but other mechanisms are possible.
“When the DSP overloads, a low-level burst of noise splatters across the audio spectrum with the bulk of the energy concentrated in the higher frequencies,” writes John Siau in an Application Note. “These bursts tend to create a false percussive brightness in the high end.” It’s a problem only for “1x” audio dataie, data sampled at 44.1 or 48kHzand for MP3s (footnote 3). With hi-rez recordings, the frequencies likely to be affected are above the range of human hearing.
As long as they’re rare, intersample overs are inaudible. But when a bunch of them occur togetherand they often appear in clusters, especially with loud percussionthose clusters can be audible. And while they’re surely numerous in really hot recordings (think Loudness Wars), even in some highly regarded recordings they’re abundant enough to cause problems. (Siau points to tracks from Steely Dan’s Two Against Nature.)
Benchmark’s antidote is to move interpolation off-chip, where more digital headroom can be provided. Solutions exist on the recording side, but Benchmark’s is the only solution I’m aware of that does this during playback (footnote 4).
Compared to the DAC2’s upgrade over the DAC1, the DAC3 upgrade is relatively minormotivated, it would seem, by a new and improved DAC chip from ESS Sabre, the ES9028PRO. The DAC3 adds active harmonic compensation (a feature of the new chip) and lower passband ripple, facilitated by the new chip’s superior filter choices. There are a few other changes I won’t go into.
Harmonic compensation reduces the levels of second- and third-harmonic distortion in the DAC3 from levels that were already very lowalmost certainly inaudiblein the DAC2. Even if this distortion were previously audible, that might not be a bad thing: translating from sciencespeak to musicspeak, the second harmonic is an octave above the fundamental, and the third harmonic is up another fifth; that can make music sound better. Although in most music we don’t make much of a distinction between identical notes an octave apart, a fifth above the octive is pleasing to the ear in much the same way as a fifth above the tonic. Benchmark, though, believes that what comes out of their converter should be the same as what goes in, so they eagerly suppress euphonic distortion.
How much audible improvement do these changes add up to? “I’m quite certain that there should be no audible difference between a DAC2 and a DAC3 given a single pass through the converters,” Siau wrote to me in an e-mail.
The DAC3 comes in three flavors. The HGC ($2195) is the standard for audiophiles; it includes the usual, excellent Benchmark headphone amplifier (with two jacks) and two analog inputs, so that it can be used as a preamp. (There’s no phono preamp, though, as there is in the otherwise similar, and similarly priced, Mytek Brooklyn.) The DAC3 L ($1895) is the same as the HGC but deletes the headphone amp. The DAC3 DX ($2095) is the more professional version: there are no analog inputs, but there’s a balanced (AES/EBU) digital input and a second stereo output bus; you can use one set of analog outputs fixed, and another with the volume control engaged.
Otherwise, the three versions are the same, offering one asynchronous input (USB), two optical inputs (TosLink), and two unbalanced S/PDIF inputs (RCA). Either of the unbalanced digital inputs can be reassigned as a digital pass-through for any other digital inputyou can use the DAC3 to convert a TosLink, USB, or S/PDIF digital signal to S/PDIF. And while the DAC3 won’t play multichannel music, the pass-through works for multichannel inputs.
The DAC3 HGC has another feature that, while hardly sexy, addresses a common and underappreciated problem. Is your volume control usually set above or below the halfway point? In my experience, for most systems the answer is “below”but most systems achieve their best noise performance in the top half of their volume range for both analog and digital volume controls (though for different reasons). Also, analog volume controls exhibit their best channel-matching in that range. The DAC3 offers pads for its balanced outputsattenuatorsthat can be set, via internal jumpers, to 0, 10, or 20dB. Benchmark says that when they’re in use, the DAC3 retains its full signal/noise ratio of 128dB, A-weighted. If you’re paying for 24-bit DACs and hi-rez downloads, you’d best get your noise level down to where you can hear at least some of that extra resolution, and that’s harder than you might think (footnote 5).
Thinking about Listening
Reviewing the DAC3 raises issues of methodology. As I’ve mentioned, the designer of the DAC3 believes it sounds exactly the same as the DAC2, which was very favorably reviewed by Erick Lichte and very favorably measured by John Atkinson. How does one go about reviewing a component whose designer says it sounds exactly like one we’ve already reviewed? Should I even bother to listen to it?
The answer to that question is obvious: Yes. Listening is what Stereophile reviewers do.
But to leave it at that would be facile, as there’s an interesting point here that merits further consideration. What, exactly, am I listening for? Transparency is an important ideait’s the fidelity in high fidelityand it contradicts the ubiquitous high-end notion that every audio component has a sound of its own.
Footnote 1: You can buy 6′ of Mogami professional microphone cable, in any of nine colors, mass-produced to a high standard and assembled by Sescom with Neutrik XLR connectors, from Markertek (www.markertek.com) for $18.99 each. If you prefer Canare cablejust as capableit’s a dollar cheaper.
Footnote 2: In a paper proposing the double helical structure of DNA, James Watson and Francis Crick wrote: “It has not escaped our notice that the specific pairing we have postulated immediately suggests a possible copying mechanism for the genetic material.” Nature 171: 737738 (1953).
Footnote 3: Siau’s comments got me wondering if intersample overs could possibly lie behind MP3’s most obvious artifact, that jingly-keyring test, not to mention the harsh sound of many CDs.
Footnote 4: I’m not counting server DSPheadroom management and the likewhich is common and effective. Also, some DACs, from companies such as EMM Labs and PS Audio, do all their processing off-chip and so presumably don’t have this problem, but I have not verified this.
Footnote 5: As John Atkinson has exhaustively documented, the very best DACsincluding the DAC2, the DAC3 HGC’s predecessoroffer resolution of 21 bits, and many DACS are much worse. Also, to lower noise to the point where you can get real value from hi-rez recordings, you may need a low-noise power amplifier. See Kalman Rubinson’s review of Benchmark’s AHB2 power amplifier in the November 2015 review, and especially JA’s measurements.
NEXT: Page 2 »
Benchmark Media Systems, Inc.
203 E. Hampton Place, Suite 2
Syracuse, NY 13206
Benchmark DAC3 B