Every
good engineer likes a challenge, so by 1997 the best
transport/processor combinations had no more jitter than the best
one-box CD players, and offered dynamic range approaching 20 bits and
resolution better than that. But also by 1997, the market for standalone
DACs had shrunk almost to zero. For with the introduction of DVD-Video,
audiophiles had access to a medium that at long last could carry
higher-resolution digital data than the CD. With word lengths of up to
24 bits compared with the CD's 16 and a sampling frequency that could
range up to 96kHz, DVD-Videos could be used for high-resolution
two-channel music signals. And they were, with Classic Records and
Chesky releasing 96k-sampled material—DADs and Super Audio Discs,
respectively—throughout 1998.
By 1997, therefore, there seemed little point in an audiophile investing big bucks in a separate processor that would not operate above a 48kHz sampling frequency. And confounding the matter, it was open to question whether any DVD-Video players would offer a 96kHz data output "in the clear," as the record industry was insisting that, in order to foil would-be pirates, some form of encryption be used for DVD-Audio datalinks (see Sidebar).
By the end of 1998, however, a number of DVD-Video players—from Denon, California Audio Labs, Pioneer, Theta, and Panasonic—offered unencrypted 96kHz data outputs, running a conventional S/PDIF or AES/EBU output at twice the usual speed/bandwidth. The necessary 96kHz-compatible standalone processors are starting to appear. In January, Kal Rubinson auditioned the $350 MSB Link; this month, I am writing about the new $450 X-24K from British company Musical Fidelity; and in future issues we will report on how the 96k-compatible versions of previous state-of-the-art processors from Wadia, Muse, Theta, and Mark Levinson perform. (And Jonathan Scull reports on the sound of the 96kHz- and 192kHz-capable version of the dCS Elgar elsewhere in this issue.)
Technology
The X-24K was launched at the 1998 Heathrow Renaissance Show. An
extruded aluminum cylinder somewhat longer than the rest of Musical
Fidelity's "X" series of components houses the circuitry, with AC power
provided by a 12V wall wart. On the front panel are eight LEDs to
indicate sampling frequency (32kHz, 44.1kHz, 48kHz, 88.2kHz, or 96kHz),
data lock on input one or two, and whether or not the incoming
datastream is pre-emphasized.. On the back panel are the analog outputs,
two coaxial S/PDIF input jacks, and a TosLink optical input, which is
in parallel with input 1. There is also a passthrough data output.
Inside the cylindrical case are four printed circuit boards. A small one behind the front panel carries the indicator LEDs; an even smaller one behind the rear panel carries the TosLink optical input circuitry. The other two, both double-sided, run almost the full length of the chassis. The lower one carries most of the power-supply components and some TTL "glue logic" chips, the upper one the data receiver and DAC chips on its underside and the clock crystal, a 7805 5V regulator chip, and the analog output stages on its top. The receiver is the Crystal CS8414, the DAC/digital filter the Burr-Brown PCM1716, both of these surface-mount chips.
The CS8414 is a new 5V CMOS device intended to work with S/PDIF single-ended and AES/EBU balanced digital datastreams. According to Musical Fidelity's white paper on the X-24K, this chip analyzes the incoming data to determine its sampling frequency. It then locks to it with a second-order phase-locked loop, (attenuating jitter artifacts by up to 30dB), recovers the clock and synchronization signals, and de-multiplexes the audio and digital data.
Introduced in early 1998, the Burr-Brown PCM1716 is one of the Tucson company's affordably priced "audio-grade" two-channel D/A converters. Featuring an "enhanced multi-level, sigma-delta architecture" (according to Burr-Brown's website), it is specified as having 24-bit resolution and offering a dynamic range of 106dB. Operating with sampling frequencies of up to 96kHz, it also incorporates an 8x-oversampling digital low-pass filter with digital domain de-emphasis and 82dB stop-band attenuation. Interestingly, this filter can operate with either a sharp or a slow rolloff, depending on the application.
Because no separate digital filter is required, the X-24K does not offer HDCD decoding, unlike its predecessor, the X-DAC (reviewed by Robert Harley in May 1997, Vol.20 No.5). The analog reconstruction filter is a hybrid Sallen & Key and GIC (General Impedance Converter) type said to offer a "dramatic" reduction in the level of ultrasonic spuriae.
As with such earlier Musical Fidelity digital equipment as the X-DAC, the quality of the passive components used is serviceable rather than top-quality, MF obviously having decided that circuit implementation is more important than absolute parts quality. The analog stages, for example, are based on NE5532 chips. While this 20-year-old dual-op-amp design can give very low distortion and low noise when used optimally, even when driving low-impedance loads, no one would argue that it represents the state of the analog art at the turn of the millennium.
Listening
Immediately prior to dropping the Musical Fidelity into my system, I had been using theWadia 27i
(which was then sent back to the factory for a 96kHz upgrade). Yes, the
cost-no-object Wadia sounded better. But at one twentieth the price,
the X-24K didn't sound too shabby. In fact, it sounded damned good.
The highs were smooth and grain-free. Soundstaging was wide and deep. On Gramophone magazine's 1998 Record of the Year, Frank Martin's Mass for Double Choir (Hyperion CDA67017), for example, the massive acoustic of London's Westminster Cathedral could be easily perceived as surrounding the natural-sounding voices. On Stereophile's new jazz CD featuring the Jerome Harris Quintet (Rendezvous, STPH013-2), the mix of Lexicon-derived reverberation and the natural ambience of what used to be the First Christian Church in Salina, Kansas surrounding Jerome's Taylor bass guitar solo at the start of "Hand by Hand" sounded quite delicious. Even when there was no real soundstage on the recording, as with the Riven soundtrack (Virgin 45425 2), the Musical Fidelity threw an enormous, enveloping soundstage. (A major part of the commercial success of both the Myst and Riven computer games, I feel, was Robyn Miller's genius at composing melodic fragments that could set scenes so forcibly.)
It was mainly in the bass that the Musical Fidelity fell significantly short of what I had been used to from cost-no-object CD playback gear. The new (as of November '98) Joni Mitchell album, Taming the Tiger (Reprise 46451-2), has a rich balance overall that makes perfect musical sense over the Levinson and Wadia processors. But with the X-24K, the double bass on "No Apologies" both loses some of its low-frequency extension and its upper-bass definition, to the detriment of the cut's majesty. (All listening comparisons, by the way, were performed with levels matched to within 0.05dB at 1kHz.)
Similarly, when I was mixing Rendezvous, I wanted a live sound on Billy Drummond's kick drum, with some ring and hall acoustic apparent—no muffled thuds on a Stereophile recording, thank you very much. I set the kick-drum level using the Wadia 27 to decode the data from the Sonic Solutions digital audio workstation. When I auditioned the mix on the X-24K, I wished I had pulled the kick drum fader back a dB or so—some of the tightness had been replaced with boom.
Comparisons
But put these remarks in perspective. This is with the Musical Fidelity's sound compared to a much
more expensive baseline. Compared with the similarly priced Entech
205.2, the Musical Fidelity had a similar treble presentation but
sounded a bit more forward in the midrange. In fact, the Entech made
Joni's cigarette-toned contralto sound a lot more recessed in the mix
than either the Levinson or the Musical Fidelity. The British processor
offered better focus on the voice than did the Entech. Though it had
woollier-sounding lows, the sonic pictures hung together more
coherently. The Entech sounded a little untidy overall.