Mark Levinson No.390S CD Processor
I flashed back to this talking-head experience when I recently read that the typical life of a recorded music medium is that same 20 years. Which begs the question: What am I doing reviewing a CD player now that SACD and DVD-Audio (and MP3) are here?
The fact is that, while SACD and DVD-A both offer clear sonic advantages over CD, neither medium has made much of a dent in the recorded-music market in the four years since their launch. Their time still lies in the future; despite its current commoditization, CD is still the predominant means for audiophiles to get a high-quality musical fix.
Enter the No.390S
At $6700, the Mark Levinson No.390S is expensive for a CD player, but it's housed in the elegant black chassis that evolved from the center section of Levinson's groundbreaking but now sadly discontinued No.30 Reference digital processor, from 1991. (Harman's website refers to the No.30, in its final ".6" incarnation, as a "legacy" product.) The No.390S offers HDCD decoding (such discs illuminate a front-panel LED), and is an internal redesign of the No.39 CD Processor, which Wes Phillips enthused over back in the November 1997 Stereophile. I refer you to his review for a full discussion of the No.390S's functionality and physical attributes. The most important point to note is that while the '390S can be used with a fixed output level driving a preamp, it also has a digitally controlled analog volume control with 0.1dB steps, which allows it to be used directly into the power amplifiers. Take into consideration that it has two digital inputs, and the realization dawns that the 'S can act not only as a source component but also as the control center for a no-compromise two-channel system.
As with all Mark Levinson products, the No.390S's parts and construction quality are to a very high standard. The player's heart is a nicely engineered transport that occupies the entire depth of the case behind the right side of the front panel. The transport's soft bumpers give a nice feel to the elegant, 1/8"-high, belt-driven drawer. Two regular, green-colored, fiberglass printed circuit boards carry the power supplies—separate toroidal transformers are used for the control-transport and audio circuits—and the circuitry for the digital and control circuits. The audio circuitry is carried on a four-layer PCB made of Arlon 25N—an expensive, low-dielectric-constant material—and shielded by an aluminum metal box labeled "No.360S." In fact, the '390S combines Mark Levinson's No.37 CD transport (another "legacy" product, reviewed by Tom Norton back in January 1997) with the D/A and output stages of the No.360S processor (reviewed by Kal Rubinson in October 2003).
Tracing the signal path on this mainly surface-mount board (the two channels are physically separated but identical): Balanced data from the transport section or an external S/PDIF source are upsampled to 352.8kHz and a word length of 24 bits and fed to an Analog Devices AD1853 D/A converter chip (compared with the '360S's Burr-Brown PCM-1704 chip). This 24-bit part is used with hand-selected, bulk metal-foil resistors (calibrated to a tolerance of ±0.0002%), and its two dual-differential channels handle the two balanced signal phases, resulting in four analog output signals per channel. This chip is followed in turn by high-speed AD823 op-amps (these used, I assume, as I/V converters) and the volume control. This operates in the analog domain to preserve DAC resolution and, like that of the No.380S preamp, is implemented with the internal resistor network of an MDAC chip (here a Linear Technology LTC1590). The balanced audio circuitry is based on Analog Devices AD810 video op-amps and Burr-Brown OPA2134 dual SoundPlus op-amp chips.
The No.390S runs warm, so make sure it has adequate ventilation. I performed almost all of my auditioning with sample 1671, and used sample 1563 as a check. The only operational problem I had was with the player's slim CD drawer. Occasionally, after I'd pressed Open, the drawer would close a second or so after opening. Normally this was merely inconvenient, but twice I'd already begun to remove the disc, which then got crunched. I learned to wait before reaching for the CD, but this sort of problem shouldn't happen with a $6700 player.
As KR found with the No.360S processor, the No.390S's high frequencies were smooth and grain-free, its overall presentation open and transparent. Unlike Kal, however, who heard a natural, unexaggerated low-frequency balance in the '360S, I found the '390S to have a rather generous midbass. This was not unpleasant, and lent orchestral recordings a convincing bloom. However, kick drum and Fender bass acquired a bit too much body, which might be a problem in systems that are themselves on the big-bottomed side. Certainly on Ry Cooder and Manuel Galbán's Mambo Sinuendo (Perro Verde/Nonesuch 79691-2) via the Revel Studios, the low end became rather overpowering. But on my recording of small-group jazz by the Jerome Harris Quintet (Rendezvous, STPH013-2), the Levinson's bass usefully brought Harris's acoustic bass guitar forward in the mix.
Overall, particularly when the No.390S was used straight into the power amplifiers with its volume control active, there was an addictive vitality to its sound, a superb retrieval of recorded detail. Even weary old recordings such as The Best of Fairport Convention (A&M Chronicles 069 493 308-2) sounded clean and smooth, Sandy Denny's "Who Knows Where the Time Goes" reminding me why I once thought this English band more than equal to anything that came out of San Francisco in the 1960s. And new CDs, such as the unlikely but astonishingly successful collaboration of singer Ronnie Isley and composer Burt Bacharach (Here I Am, DreamWorks B0001005 02), wanted for nothing (other than a little less midbass on Neil Stubenhaus's five-string bass guitar).
Against my long-term reference, the combo of Mark Levinson No.31.5 transport and No.30.6 DAC, with levels matched to well within 0.1dB at 1kHz with the Levinson preamp's Input Offset function, the more expensive 31.5-30.6 combo had better-defined, better-extended lows. Otherwise, it was a pretty even match, and I might even swear that the No.390S had a slightly more believable presentation of recorded ambience. Certainly its highs were as smooth, and almost as free from grain. This was for CD playback; playing 24-bit WAV files from my PowerBook via a Metric Halo Mobile IO 2882 FireWire audio processor connected via AudioQuest's SVD-4 S/PDIF cable moved the '30.6 way ahead on points, the '390 sounding grainy. And a major irritation for me was that the '390S wouldn't lock on to the MIO 2882 when I played back music sampled at rates higher than 48kHz.
Against the $6000 Musical Fidelity Tri-Vista SACD player—using only CDs, of course—and again with levels matched at 1kHz, the No.390S's soundstage was actually a little deeper than that of the tubed player, but its presence region was a little more forward, its midrange very slightly less rich-sounding. As a result, the chiff that accompanies the start of each note of the flute in the Mozart quartet movement on my Editor's Choice CD (STPH016-2) sounded slightly disconnected on the No.390S, better integrated with the image of the instrument's body on the Tri-Vista. On the other hand, the '390S's character better defined the leading edges of piano notes, resulting in a more delicate presentation of both Robert Silverman's Bösendorfer and Hyperion Knight's Steinway on Editor's Choice. The two players sounded pretty much the same in the bass, however.
I also used the MF player as a transport, feeding the Levinson's digital input using Stereovox's new, inexpensive hdxv S/PDIF link. The No.390S's highs were not quite as grain-free as when I used it as a conventional CD player. However, there was one enormous difference: the Tri-Vista's S/PDIF data output swaps the channels! Sloppy engineering on Musical Fidelity's part; my apologies for not noting this in our review last May.
My third and last comparison was with the Classé CDP-10 ($2000), which had so impressed me last September. While the Canadian player has relatively utilitarian styling, its measured performance is virtually identical to the Levinson's. It, too, offers HDCD decoding, and uses the same Burr-Brown SoundPlus op-amps in its output stage. As with the Musical Fidelity, I compared the players both as players, and with the Classé used as a transport driving the '390S's digital input via the Stereovox datalink.
With each player on its own, using their balanced outputs, I found it impossible to distinguish between the players' high frequencies or soundstaging capabilities. Both were superb in these respects, probably limited more by the ultimate performance of the CD medium than by their own intrinsic abilities. However, when the Classé was used a transport for the Levinson's D/A section, the latter's high frequencies sounded slightly more grainy, less smooth.
I did note a persistent difference in low-frequency behavior. The Classé's presentation lacked the low-frequency bloom of the Levinson's, sounding more lightweight but more delicate. With recordings that were overcooked in the bass, such as Mambo Sinuendo, this was welcome, but it made the Bösendorfer on my recording of the complete Beethoven piano sonatas (OrpheumMasters 75020 14182 7) sound a little more silvery than it should have. By contrast, the Levinson reproduced the piano with more lower-midrange body, more bloom. Similarly with "Danny Boy," from Editor's Choice: The low basses sounded fuller on the Levinson, more supportive of the harmonies, to the benefit of the music.
With its classic appearance, its superb sound from CD, and its transparent-sounding volume control allowing it to be used without a preamplifier, the Mark Levinson No.390S comes close to justifying its asking price of $6700. But as much I want to replace my Nos.31.5-30.6 stack with something smaller and more ergonomically friendly, the measured problems I found with the No.390S's external data input mean I can't abandon my separate processor just yet. Of course, there aren't that many domestic sources around with word lengths greater than 24 bits or sample rates greater than 48kHz—probably a suitable DVD player playing one of the small number of Chesky and Classic DADs is all a typical audiophile will have access to. If you have a DAT recorder or a second CD transport, then this will not be an issue for you, but it was an issue for fussy me. I can recommend the No.390S as a Class A CD player, but I must withhold a full recommendation for it as a CD "processor" until its external input problems are resolved.