Sonic Frontiers SFD-2 D/A processor JA Listens to the SFD-2

Sidebar 4: JA Listens to the SFD-2

Given RH's enthusiastic report on the sound of the Sonic Frontiers, I felt that I should also give it a listen. My system consisted of B&W Silver Signature speakers driven by a pair of Mark Levinson No.20.6 monoblock power amplifiers; the preamplifier was first the Sonic Frontiers SFL-2, then the remote-controlled Mark Levinson No.38S. The digital source was a Mark Levinson No.31 transport connected to the processor via either Madrigal AES/EBU cable or ASM Labs' Mamba ST-optical cable. For reference purposes, I too had the original SFD-2 and the Mark Levinson No.30.5. All analog connections were balanced, using AudioQuest's AudioTruth Lapis x3 interconnect.

My first comparison was with the SFD-2. As good as I found the original Sonic Frontiers processor in its balanced mode, it was really no contest. The Mk.II processor kept the good aspects of the Mk.I's sound quality—its well-delineated soundstaging and punchy low frequencies—but reduced the forward upper-midrange quality to the original unit's tonal balance that, for me at least, made system-matching quite tricky. As RH says above, the Mk.II's balance was more laid-back. This didn't mean that low-level detail was smeared or missing in action. Rather, instead of being artificially thrust forward at me, it was there when I listened for it—just as in real life. This was coupled with a more effortlessly natural reproduction of midrange textures. If you own the original SFD-2, the upgrade to the Mk.II would be well worthwhile, in my opinion.

Comparing the new SFD-2 with the Mark Levinson No.30.5 was more of a contest. With levels carefully matched using the No.38S's programmable volume control (footnote 1), for a while I found it difficult to home in on the exact sonic differences between these two leading contestants. After several discs, however, four consistent trends became apparent.

First, while the Mk.II SFD-2 sounded quite a bit more laid-back than the Mk.I, it still featured a somewhat forward balance in the upper mids. This was most apparent on piano—Stereophile's Robert Silverman Concert CD (STPH005-2), for example—where the instrument acquired a more robust, louder-sounding presentation. By contrast, the Levinson presented the piano slightly behind the plane of the loudspeakers—which, in itself, might lead the naïve listener to assume that the recording was not playing quite so loud. This was also apparent on male voice. The lead vocal on "New York Minute," from the Eagles reunion album, Hell Freezes Over (Geffen GEFD-24725), sounded both more robust and more palpable over the Sonic Frontiers.

Second, the Levinson had fractionally more high-frequency air to its sound. While not something that would be identifiable in anything other than an A/B comparison, it made the recorded ambience sound more tangible. The Albuquerque church where I recorded Concert, for example, was just that little bit more identifiable over the No.30.5 than over the Sonic Frontiers.

Third, while the SFD-2's sound had more midbass body, the No.30.5's had more low-frequency extension. Kickdrum on the Eagles track was presented quite differently over the two processors: the Sonic Frontiers emphasized the midbass weight of the instrument; the Levinson emphasized its transient attack—the sheer kick-you-in-the-solar-plexus impact—and had more ultimate low-frequency extension. I initially thought this might be due to the Sonic Frontiers' highish output impedance at low frequencies (see later), but as the No.38S has an input impedance of 100k ohms, this shouldn't be a factor.

Finally—and, again, you must remember that I'm talking about small distinctions here—the SFD-2 had just a little more mid-treble grain than did the No.30.5. "History Repeats Itself," the Sara K. track featured on Stereophile's Test CD 3 (from Play on Words, Chesky JD105), allowed me to pick this up on the female vocal. There was a trace more sibilant edge to the sound of Sara K.'s mellifluous utterings via the SFD-2. Whether this is due to the Sonic Frontiers' 6DJ8 tubes—which, in my experience, can introduce a bit of mid-treble glare—or the fact that the SFD-2 uses a steel chassis, which will always have some deleterious effect on ultimate sound quality, I have no way of knowing. I can only wonder, however, what the already superb-sounding SFD-2 would sound like with a nonmagnetic, aluminum case.

Ultimately, for playback of non-HDCD recordings I preferred the three-times-the-price Mark Levinson No.30.5, which in my system sounded more natural/neutral. But in other systems, the nod might well go to the Sonic Frontiers SFD-2 Mk.II—as it did in Bob Harley's. However, when it came to HDCD playback, the SFD-2 outstripped the Levinson.

I had one operational difficulty while auditioning the SFD-2 Mk.II: once the ST-optical input had been selected, I couldn't get the decoder to lock to other inputs, such as the AES/EBU, without physically disconnecting the ST cable. This made it more time-consuming to perform all my HDCD comparisons (see next section). Whether this is a sample fault or generally typical of the SFD-2's input circuitry, I don't know.

Summing up, the SFD-2 is right up there with the best digital processors. Given my druthers (and unlimited disposable income), I'd go with the $15,950 Mark Levinson No.30.5 for non-HDCD-encoded software, where it still has the edge over the Canadian processor. And bearing in mind that Madrigal now supplies an HDCD upgrade kit—the PMD100 chip mounted on a small daughterboard and a replacement EPROM—for the No.30.5, that should be one superb-sounding unit (footnote 2). But for those living in the real world, the Sonic Frontiers SFD-2 Mk.II gets you almost all the way there for a lot less money.—John Atkinson

Footnote 1: Those who make a big deal about the difficulty of matching levels should note that all you need is a reasonable AC voltmeter—I use a Fluke 87—and the -20dB, 1kHz tone on the first Stereophile Test CD. Though the No.38S's volume control operates in 0.1dB steps, the fact that the No.30.5's output at 1kHz was almost exactly 2.4dB greater than the SFD-2's meant that any residual level error in my comparisons was below 0.02dB.—John Atkinson

Footnote 2: As soon as RH saw the socket holding the NPC filter chip in the No.30.5 while preparing last October's Follow-Up, it struck him that Madrigal is preparing for HDCD. God willing and the creek don't rise, our Follow-Up report will appear shortly.—John Atkinson