Mark Levinson No.30.6 Reference D/A processor Page 2

It's easy for reviewers to exaggerate the degree of sonic differences they hear, so don't infer from this description that the No.30.6 "blew away" the '30.5. It didn't. In absolute terms, the differences were small—but they were still important. I recommend '30.5 owners upgrade their processors as soon as budget and the factory's schedule allow. And this is without considering the fact that the new version decodes high-sample-rate material.

As I mentioned in this issue's "As We See It," I was working on the next Stereophile CD project during the preparation of this Follow-Up: 88.2kHz-sampled recordings of the Brahms and Mozart Clarinet Quintets. As Madrigal's Jon Herron told me, Madrigal only had access to 96kHz material during the '30.6's development, so they didn't have direct experience of 88.2k sample rates (footnote 1). Nevertheless, when I fed the double-rate AES/EBU datastream from the dCS 972 to the No.30.6, "88.2kHz" appeared on its display, and an open, free, superbly detailed sound issued forth from the speakers, with significantly better-defined lows.

My microphones roll off above 25kHz, and my hearing doesn't extend at all above 16kHz, so it can't be the extended bandwidth that makes higher sample rates sound so good (footnote 2). But sound good they certainly do, whether it was my own 24/88k2 recordings or the Classic and Chesky DVD-Video 24/96 music discs.

Against the Wadia 27ix: I reviewed the latest version of Wadia's processor in the April '99 Stereophile. On CD playback, I felt the $7950 Wadia to have "impressive transparency" compared with the No.30.5, though with a very slightly more grainy mid-treble with less well-defined lows. But when the 27ix was used to decode 24/96 data, I concluded that "The sheer vividness of the Wadia's presentation, coupled with laid-back high frequencies that resembled nothing so much as the real thing, made this the best I have ever heard from any digital audio technology."

While the Wadia lacks HDCD, it does incorporate a digital volume control, obviating the need for a preamplifier, which makes it an excellent value. The Wadia's unique filter gives better time-domain performance at the expense of premature HF rolloff, typically -3dB at 20kHz. On CD replay, this was just audible as a lack of HF air. But at the high sample rates, this was not an issue, both the Wadia and the Levinson being flat within the audioband.

Whether it was with the Wadia driven from the Levinson's AES/EBU digital output or from the transport's ST-optical output, it took a lot of careful level-matched listening to reveal audible differences between these two processors. But with 24/96 program, the No.30.6 sounded a tad more forward in the low treble, had better defined lows, and room ambience was just a tiny bit more discernible. By contrast, the Wadia's view into the soundstage seemed just a smidgeon more murky, but individual instruments, particularly those with a tenor voice like the cello, seemed slightly better fleshed-out within that stage. But on HDCD CDs, like the superb Elvis Costello/Burt Bacharach collaboration Painted from Memory, Mercury 314 538 002-2), it was no contest, the Levinson edging ahead.

Against the Muse Model 296: Shannon Dickson rated the $3000 Muse, driven by its I2S data input from the Muse 8 DVD/CD transport, as the best digital audio playback he had experienced. I agree that this combination sounds superb, but for these comparisons, I used the same CAL and Denon DVD transports to drive the processors via an AES/EBU link.

The sonic differences were initially subtle. On my 88k2 classical master tapes, I was hard put to identify any significant difference. But the 96k reissue of Muddy Waters' Folksinger (Classic DAD 1020) proved to be the diagnostic tool. Yes, the Muse sounded smooth, with a complete lack of treble grain. This is definitely a Class A bargain. But even with the No.32 compensating for the Muse's 6dB-lower output level, the 296 sounded quieter, less dynamic than the '30.6. Muddy's sudden shouts raised goosebumps with the more expensive processor. In addition, the Levinson's rendition of the double-bass was rounder and deeper.

Summing up: In its '30.6 incarnation, Madrigal's Reference Mark Levinson processor builds on its predecessor's reputation for state-of-the-art CD sound quality and adds high-sample-rate capability. Other than a few DVD players with 96k data outputs, some computer soundcards, and pro recorders like the Nagra-D, such data sources are currently few and far between. But with its "Reference" status assuring its owners that Madrigal intends to support the unit no matter what data sources become ubiquitous, I can confidently recommend the No.30.6 as the digital-music platform for the future. Next month, I look at its measured performance.

Footnote 1: I recorded this project at 88.2kHz in order to keep the downconversion to the CD's 44.1kHz sample rate as transparent as possible. See Stereophile, March 1997, p.3, and February 1999, p.101.

Footnote 2: British engineer Julian Dunn, once with Prismsound, explored this paradox in a paper presented to the 1998 Amsterdam AES Convention (downloadable from His conclusion was that the relaxed transition-band requirements for the digital filter with the higher sample rates reduced the level of audible time-domain spuriae (pre- and post-echoes).

Mark Levinson
2081 South Main Street
P.O. Box 781
Middletown, CT 06457
(860) 346-0896