I beg your indulgence for a moment, as I step into the Way-Back Machine, and emerge on a quiet afternoon in the spring of 1986. I am working at a high-end audio store, and on this particular day I find myself perusing the most recent issue of The Absolute Sound. Therein, I find a rave review of the Audio Research D-115 Mk II power amplifier by veteran contributor Andrew G. Benjamin. Our store is a top-ten Audio Research dealer, so I am naturally pleased that AGB's effusive enthusiasm and insightful delineation of the D-115's merits will fuel further interest in our premier line of tube electronics. However, I feel some consternation at the magnitude of AGB's praise, especially his unequivocal proclamation that, “Perhaps if I heard a better device, I would catch some faults here or there.” I think to myself, “Well, I guess he hasn't heard the Audio Research M-100s.” I know from personal experience that, while the D-115 is indeed a sensational $3000 stereo amplifier, the $5000 per pair Audio Research M-100 mono amplifiers are even better still.
This little anecdote kept popping into my head as I contemplated how to present the results of four years' worth of comparative listening tests, covering eight different sound cards. While I haven't heard every sound card out there, I sure have spent a good deal of “quality time” testing many of the prime contenders.
In the December 2007 issue, I outlined the basic architecture of a computer-based music server, and detailed what I had heard using the analog outputs of the Echo Mia MIDI ($199) sound card and Echo Gina 3G ($439) external PCI audio interface. These preliminary tests instilled a deep appreciation for the uniquely compelling sonic benefits of computer-sourced music playback, most notably “an unprecedented solidity, providing a welcome foundation to the music” and “...a certain ineffable rightness to the music server's sound, which somehow makes the process of listening to digital music easier, less demanding of subconscious mental effort.”
I can happily report that long-term evaluation of another half-dozen sound cards has only confirmed the validity of those first impressions. While the performance of USB, FireWire, and network-streaming devices continues to evolve and improve, the design of internal PCI audio interfaces is already a mature science, based on established engineering standards and robust, proven technologies. The best sound cards set a benchmark standard against which alternative technologies must ultimately be judged.
As most readers of The Absolute Sound have discovered, every new listening experience adds to our accumulated knowledge base. Over the course of this survey, the process of auditioning each successive product added further perspective and context, ultimately culminating in a surprising and exciting conclusion.
After my initial tests of the analog outputs of the Echo Mia MIDI and Gina 3G, I turned to their digital outputs, driving my Bryston BDA-1 digital-to-analog converter. Despite my preconceptions favoring the digital connectivity architecture of the Gina 3G, I was surprised to find that the digital output of the Mia MIDI actually sounded better, with a more incisive, open, and rhythmically agile presentation. While the analog outputs of both Echo cards sounded fine at their respective price points, the Bryston DAC is in an entirely different league, as has been described in these pages in Issues 193 and 194.
Since both Echo cards are limited to a maximum sample rate of 96kHz, I moved on to products that support all the common sample rates up to 192kHz. First up was the M-Audio Audiophile 192 ($200), which features RCA digital S/PDIF jacks on its mounting bracket, and balanced analog outputs on quarter-inch TRS phono jacks at the end of a short DSUB breakout cable. I auditioned the Audiophile 192's analog outputs only briefly, finding trade-offs versus the Echo Gina 3G. The Audiophile 192 exhibited a slightly washed-out character, rendering music with a desaturated tonal palette compared to the Gina 3G's more substantial and colorful presentation.
In contrast, the digital output of the M-Audio Audiophile 192 handily outperformed those of both Echo cards. I won't bother listing all the areas in which sound quality improved, since the bottom line was simply this: You name it, and it got better. The net effect was a wholesale increase in both sonic realism and musical communicativeness. However, as I continued to use the Audiophile 192 for many months, I ran into a few problems. Its driver was more temperamental than that of the Echo cards, occasionally freezing up and requiring a computer reboot. It was also subject to occasional dropouts, which I rarely encountered with the other cards tested. Finally, the Audiophile 192 was not capable of transmitting a 176.4kHz signal via its digital output; rather, it downconverted this sample rate to 88.2kHz.
The next contender proved to be quite a delight, with some unique, appealing features. The $139 ESI Maya44's RCA digital S/PDIF output jack is also optimally located on the card's mounting bracket, accompanied by an optical TosLink S/PDIF jack. Two pairs of single-ended analog inputs and outputs are efficiently configured on stereo quarter-inch TRS phono jacks. The output stage of the second analog output pair (channels 3 and 4) features a very low source impedance that can drive headphones directly. In use, the Maya44 operated with rock-solid stability, but unfortunately also downconverted 176.4kHz signals to 88.2kHz via its digital output.
The Maya44's coaxial S/PDIF digital output delivered a light, airy sound via the Bryston DAC. Its character was rather impressionistic, lacking a bit of body, foundation, and tonal richness. Direct comparison with the M-Audio Audiophile 192 clearly favored the Maya44, which delineated instrumental lines with greater ease, and allowed music to flow and breathe with a compellingly engaging naturalness. The Audiophile 192 sounded more forward, with a weightier bass range, but the Maya44 rendered the sounds of actual instruments playing in the lower octaves with superior pitch definition, rhythmic subtlety, and dynamic inflection.
The intrinsic sonic character of the Maya44's analog outputs paralleled that of its digital output, a theme that was consistently repeated across the span of this survey. The Maya44's second analog output pair was the better of the two, with punchier, tighter bass and superior focus. Sonic artifacts such as slight grain, a bit of edginess on transient peaks, and diminished soundfield scale were clearly evident when compared against the far more capable Bryston DAC, but these colorations were so astutely balanced that they never interfered with the Maya44's ability to play music with a lively sense of involvement. Setting the Maya44's analog-domain output-level control to -2dB reduced the sharpness on high-level peaks to an appreciable degree. All told, the Maya44 once again raised the bar another notch closer to the elusive goal of sonic realism.
Moving on to the feature-rich RME Hammerfall HDSP-9632 ($549), I was finally rewarded with a sound card that could transmit all sample rates including 176.4kHz via both coaxial S/PDIF and balanced AES/EBU digital outputs. Alas, I immediately ran into a snag due to the manner in which Foobar 2000 interacts with RME's ASIO Hammerfall DSP driver to assign ASIO channels to the card's physical outputs. This problem threatened to require manually switching between three different ASIO configurations depending on the sample rate of the file being played—clearly an untenable situation.
Fortunately, the freeware utility ASIO4ALL came to the rescue, since it can limit the number of ASIO channels visible to the music playback program. With ASIO4ALL, a single Foobar ASIO channel assignment configuration enabled the HDSP-9632 to play files of any sample rate without incident. Both the RME ASIO driver and the universal ASIO4ALL driver bypass the Windows audio subsystem to deliver a bit-accurate signal to the sound card hardware, and I heard no audible difference between the two options.
I also tested another music playback program, Media Monkey, which assigns ASIO channels in an entirely different manner. Media Monkey was able to use RME's native ASIO Hammerfall DSP driver to play files of all sample rates via either the HDSP-9632's stereo analog or digital outputs without any problem whatsoever.
When I began assessing the RME card's S/PDIF output, my reactions were mixed. The RME sounded clean and full, in a forthright and controlled manner, but also exhibited a closed-in, muted quality that lacked the beguiling ebullience of the Maya44. However, as I revisited familiar source material, I found myself able to hear deeper into the music via the RME, discerning musical strands that the Maya44 left buried in the mix.
Then I happened to play one selection—Ravel's Pavane pour une Infant défunte, performed by the Minnesota Orchestra under Stanislaw Skrowaczewski (extracted from Classic Records 24/96 DAD 1025)—that abruptly ended the need for further comparison, so unexpected and definitive were the differences between the two cards. When I played the Pavane through the RME card, after a few seconds I just laughed out loud, blurting “Oh! Now I get it!” From the opening notes, the music immediately unfolded with a stately, measured, restrained, “walking” rhythm. Searching for words in my head, I first thought “funereal,” but discarded that descriptor as too somber; this was lighter, but still reverential and formal. So I decided that it sounded more like a “procession.”
I did a quick Wikipedia search, and found pavane defined as “a slow processional dance that enjoyed great popularity in the courts of Europe during the sixteenth and seventeenth centuries.” While this piece sounded lovely via the Maya44, the RME card presented it in an entirely different, revelatory (or perhaps Ravelatory) light, clarifying the meaning and structure of the music with obvious, manifestly superior lucidity. Even more significantly, it made clear that Skrowaczewski's interpretation of the work masterfully conveyed the composition's essence, and that both the original recording and the subsequent digital transfer successfully preserved the expressive dynamic shadings and timing cues of that performance.
The RME card's AES/EBU digital output didn't fare as well, sounding brighter and more mechanical, with an electronic haze that diminished the magical rhythmic dexterity so evident via coaxial S/PDIF. My initial impression of a mellow, midrange-centric, overly restrained character was confirmed and amplified when I auditioned the RME card's analog outputs, which reminded me of a vintage Linn Sondek turntable in both virtues and liabilities: musically expressive, yet colored by excess warmth, curiously curtailed reverberant decays, and an intimate, close-up perspective that flattered recordings of small ensembles, but never opened up enough to convey the power and impact of a full orchestra.
After my positive experience with ESI's Maya44, I approached the oddly named ESI Juli@ card ($199) with eager anticipation. The Juli@ card features a clever analog input/output configuration, with a reversible plug-in board that is fitted with RCA jacks on one side, and balanced TRS jacks on the other. Alas, I was disappointed to find the coaxial S/PDIF RCA jacks relegated to thin breakout cables tenuously connected to the card by a loose-fitting multi-pin DIN plug.
Despite my concern about this less-than-ideal digital output configuration, the Juli@ immediately distinguished itself as something quite special driving the Bryston BDA-1 DAC. Rather than the Maya44's euphonic, water-color presentation, the Juli@ illuminated the music with unprecedented clarity, focus, and resolution. Rather than the RME's restrained, closed-down character, the Juli@ breathed with life, air, and energy. The Juli@ card rendered complex timbres effortlessly, with an unsurpassed ability to convey note shape—the dynamic and harmonic growth and decay of each note over time.
Spatial cues were presented with captivating verisimilitude. Finally, here was a card that preserved the full scale of well-recorded orchestral performances, with three-dimensional instrumental body and natural reverberant decay into the ambient air of the recording venue. Of particular note was the Juli@'s imperturbable dynamic stability. Even during energetic crescendos, the Juli@ card maintained a consistent perspective; instruments in the rear of the soundfield stayed in the rear of the soundfield. In contrast, every other sound card in this survey exhibited sonic artifacts during high-energy peaks, either coarsening textures, adding glare, or projecting the sound forward.
This sense of unflustered, well-behaved equanimity categorically differentiated the ESI Juli@ from every other sound card tested. Whether delicate chamber music, free-wheeling improvisational jazz, or raucous rock ’n’ roll, the Juli@ not only reproduced the recorded sound in a manifestly realistic manner, it played the music with expressive subtlety and enthralling vitality.
After hearing the standard-setting performance of the Juli@'s digital output, I crossed my fingers as I connected its RCA analog outputs to my preamp. My hopes were rewarded, as the sound of the analog outputs mirrored the considerable merits of the card's digital output. No, the Juli@ card's analog outputs did not deliver the soundstage width and depth, unfettered dynamic peaks, rich tonal complexity, and rhythmic precision of the Bryston DAC, but the essence of the presentation was uncannily preserved, only on a smaller, less resolved scale. The Juli@'s cardinal virtues remained obvious via the RCA analog outputs, diminished in degree, but not in character. Even after periods of extended listening, I found myself shaking my head in disbelief, admitting to myself that “I could live with this,” so well-balanced were the sonic compromises of the Juli@'s RCA analog outputs. However, I was utterly chagrined to find that the Juli@'s balanced TRS analog outputs sounded comparatively shrill, unrefined, and fatiguing. I sighed with relief after reversing the Juli@'s analog output board again, returning to the sweet, even-tempered music flowing forth from its single-ended RCA analog outputs.
The final two sound cards arrived courtesy of well-respected Lynx Studio Technology. The L-22 ($749) is a stereo card capable of internal decoding of all sample rates up to 192kHz for playback via its analog outputs, but limited to a maximum of 96kHz via its AES/EBU digital output. The AES-16 ($695) is a versatile multichannel digital-only card with comprehensive onboard mixing capabilities. All inputs and outputs on both cards are provided on DSUB connectors.
The L-22's balanced analog outputs delivered a visceral, full-bodied presentation, rich with tactile immediacy. However, under dynamic conditions, it failed to measure up to the ESI Juli@'s poise and composure. At low to moderate signal levels, the L-22 exhibited a smooth liquidity, but high-level peaks sounded edgy and forward. It offered a vivid, up-front perspective, but with reduced depth, flattened dimensionality, and truncated reverberant decays. The L-22 seemed to emphasize whatever was most prominent in the mix, spotlighting the loudest element of the music with explicit directness and presence, but at the expense of overshadowing quieter instruments playing simultaneously. The L-22's AES/EBU digital outputs sounded remarkably consistent with its analog outputs. Considered on its own, the L-22 certainly sounded nice enough, but repeated comparisons only emphasized its colorations when compared against the more coherent, neutral, better-behaved ESI Juli@.
The Lynx AES-16's AES/EBU digital outputs exhibited an obvious family resemblance to those of the L-22, but with superior performance across the board. Compared to the L-22, the AES-16 sounded cleaner, with improved solidity, tighter focus, reduced temporal smearing, and a more engaging sense of drive. Substituting upgraded AES/EBU cables improved performance even further. Still, direct comparison with the ESI Juli@ laid bare the AES-16's limitations: lackluster rhythmic alacrity, homogenized tonal colors, and a subtle opacity that obscured note decay, hall ambience, and low-level dynamic inflections in a manner analogous to the absorption of sound in humid air. Perhaps I might have had a more enthusiastic reaction to the Lynx cards had I auditioned them earlier, but after hearing the extraordinary ESI Juli@, it became abundantly clear just how much of the music the Lynx cards were missing.
Over the course of this survey, I often found myself frustrated with the trade-offs between the various contenders. Fortunately, the remarkable performance of the inexpensive ESI Juli@ card completely changed the game. Not only did its single-ended RCA analog outputs trump those of the other products reviewed, but its digital output established a new reference standard. Thus, the Juli@ merits an unqualified recommendation, in both relative value and absolute performance. Apparently, I'm not the only one who appreciates the ESI Juli@'s virtues; several high-end manufactures are already using modified Juli@ cards inside their computer music products. It only seems fitting to close this survey by shamelessly lifting the concluding lines from AGB's amplifier review that I mentioned at the outset: For this reviewer, the ESI Juli@ is the first computer music source component “...to succeed in forsaking sonic artifacts. In fact it does more. For a mere electronic device, it luminescently renders the truth of the music.”
Nut & Bolts
The computer used for these tests is quite ordinary, a nine-year-old PC with a 1.47GHz AMD processor, 1 GB of RAM, and no geeky tweaking of arcane operational parameters. The machine is utterly representative of a stock generic PC, other than having been fitted with a quieter power supply and fan-speed regulators to reduce mechanical noise. Thus, the conclusions drawn from these tests can confidently be expected to represent what anyone can achieve with minimal effort.
For consistency over the course of this project, I have continued to use the Windows XP operating system. I encountered no problems installing any of the cards or their software drivers, even with two or three cards installed simultaneously. All of the cards can utilize the professional standard ASIO protocol that bypasses the Windows audio subsystem, thereby insuring bit-perfect signal transfer and playback of music files at their native sample rates. I set ASIO latency to 512 samples in each card's on-screen control panel, which allowed playback of all supported sample rates without drop-outs or audible glitches. I used the program Foobar 2000 for playback and library management. Mac Pro users are advised to choose a music player that circumvents the limitations of Apple's CoreAudio architecture, such as Decibel, Pure Music, or Amarra.
Due to limited physical space, most sound cards use non-standard connectors for analog and digital inputs and outputs, such as multi-pin DSUB or DIN jacks. In such cases, the cards are usually supplied with either short breakout cables fitted with RCA or XLR jacks, or cables of modest-length (typically 6 feet) that can be attached directly to nearby audio components. Unfortunately, these stock cables are often of the “included with a cheap VCR” caliber. Substitution of higher-quality after-market cables offers the potential of further improving performance, as with any critical analog or digital source component. I auditioned each sound card with the cables provided by the manufacturer, and also tested custom upgraded cables wherever possible.
Specs & Pricing
Echo Digital Audio
Prices: Echo Mia MIDI, $200; Echo Gina 3G, $439
ESI Audiotechnik GmbH
Prices: ESI Maya44, $139; ESI Juli@, $199
Lynx Studio Technology, Inc.
Prices: Lynx L-22, $749; Lynx AES-16, $695
Price: M-Audio Audiophile 192, $200
Price: RME Hammerfall HDSP-9632, $549
Foobar 2000 (free)
ASIO Universal ASIO Driver (free)