June 2001

MSB Technology Link DAC III with 24/192 Input Option

Speed

Speed, The Prequel. In case you forgot, it stars Meanu Reeks and Sandy Bollocks. Aural munitions expert Denise Flopper is the bad girl. Meanu stalls Denise on his elevator shaft with a rare Guaneri bow. "Pop quiz, f-hole. What’s the difference between 24-bit/96kHz and 24-bit/192kHz upsampling?" Denise doesn’t hesitate for a 1/16th of a note. "192 causes more hearing damage, moron!" She pushes the red button. Kaboom! Ears clip, brains melt. Another virtual L.A. is movie history. [sniff]

Seriously, though, I was curious how such a pop quiz about digital processing speeds would play out in the real world. I had covered some of the engineering arguments between upsampling versus oversampling versus interpolation in an earlier "Y-Files" installment. Colleague Wes Phillips had explored the same subject in one of his features . Now I wanted to hear how designer explanations translated into actual sonics. To compare apples to apples, I needed a single player that would allow direct comparisons between both modes. It would preferably even offer one-flick downsampling to standard 16 bits/44kHz. Wishful thinking? Not.

MSB to the rescue

The standard Link DAC III at $399 USD is the most popular affordable DAC on the market. Doug Blackburn reviewed it and awarded it the SoundStage! Network Reviewers’ Choice in 1999. SoundStager! Roger Kanno followed up this review of the DAC’s base version with an evaluation of certain upgrades. A manufacturer’s reply to a later Stereophile review of the 24/96 version then pointed to yet another option. In that reply, MSB held that if the writer thought their DAC fabulous, he really ought to try it with the 24/192 input option. Have them install their proprietary 24/192-upsampler card in a transport of his choosing, then connect both via their high-speed MSB Network. A quick phone call to the factory confirmed that my memory was indeed trustworthy. The upgrade existed and hadn’t been reviewed yet. And yes, they could install their 24-bit/192kHz upsampler card in my Marantz CDR630 semi-professional CD recorder. Being a recorder, the precision of the internal laser pickup was probably quite in excess of an equally priced sub-$1000 regular CD player. This would perhaps not make for impressive name recognition but still a truly decent transport. They’d further modify the Marantz by adding the phone/modem-type RG45 output port for their MSB Network. I could then run their supplied Network cable between this plug and the XLR-based Network input option on the DAC. This would give me the 24-bit/192kHz decoding I was after. At the same time, I could run the S/PDIF coaxial digital output of the Marantz into the standard coax input of the Link DAC. This coaxial connection would produce either 16-bit/44kHz or, via the MSB’s rear-panel toggle switch, 24-bit/96kHz processing. With the Network connected to the DAC, the DAC would automatically default to 192kHz decoding. The Marantz would send output signal simultaneously via both outputs. Simply unplugging the XLR connector would instantly give me the two other modes. Except for physically pulling out one single XLR connection, no further interruptions would be required. I could play musical chairs between 16 bits/44kHz, 24 bits/96kHz and 24 bits/192kHz until I went deaf. In case my ears got confused in the process, the differently colored and positioned status LEDs on the DAC’s front panel would serve as reminder of what chair I’d be sitting on. Idiot proof just as reviewers like it. Perfect!

Theory

Before we play, a bit of theory first. Why would installing the upsampler card in a transport rather than the DAC be advantageous? What was the deal with the "super high-speed" proprietary cable interface? In fact, what makes one cable faster than another? Mind you, I know from experience that my computer USB terminal is faster than my parallel port. I don’t really understand why though. Another phone call to MSB was in order. Senior engineer and Link DAC designer Dustin Symanski would jump-start my upstairs neon bulb. Flicker, flicker, here we go.

The upsampler card is in the transport rather than the DAC because of jitter. Capturing all clock data before they’re sent downstream via the interface cable reduces those timing errors by a very significant margin. Sounds sensible. The MSB Network itself consists of an eight-channel balanced wire pair. One send and one receive chip decode the data on opposite ends in the transport and converter. This chip/cable format supports a low-voltage, differential signal-transmission rate of 295MHz. The differential mode optimizes noise rejection similar to balanced operation. Compared to the 5.6MHz standard of the S/PDIF interface, the MSB Network is 52 times faster. Dustin explained that along standard coaxial digital cables, a data register controls the signal voltage steps that translate as the digital 1s and 0s. How fast the register can switch between low- and high-voltage values becomes the limiting factor of transmission speed. By comparison, the MSB Network register chip is a veritable speed demon. It reconfigures the parallel entry data of the laser pick-up into five series channels for the bit, master, left and right clocks plus the data channel. (MSB will use the three extra channels for volume control and bi-directional component communications in the future). The DAC then reconfigures the series datastream of the signal transmission back into parallel data. With output from the RG45 port, DACs other than MSB’s own couldn’t decipher the signal from this particular upsampler card.

Synchronicity

It turns out that the new 24-bit/192kHz upsampler card differs from earlier MSB offerings. It uses synchronous rather than asynchronous upsampling. Before I could protest by pointing out that there’s no way José to go from 44.1kHz to 192kHz in a synchronous fashion, Symanski admitted that this card actually supports two upsampling rates. With standard CD sources, it upsamples four times to 176.4kHz, just like the Resolution CD55 player. For 24-bit/96kHz sources, it upsamples two times to 192kHz. DVD-A’s native 192kHz format is accepted directly. To be clear, then, with my 16-bit/44.1kHz CDs, I would now be comparing road conditions in first (16 bits/44.1kHz), third (24 bits/96kHz) and fourth gear (24 bits/176.4kHz). My vehicle apparently wasn’t equipped with overdrive 24 bits/192kHz as I first believed.

Symanski claims better sonics though. The simpler math of synchronous upsampling (the new rate is an even multiple of the old) retains all data points of the original stream. It merely interpolates additional ones between them. Asynchronous upsampling involves total signal reconstruction, however. In other words, all original data points are discarded and replaced with new ones. That’s exactly what Jeffrey Kalt of Resolution Audio had pointed out in my earlier article as well. John Stronczer of Bel Canto Design, of course, had a retort. The computing power of modern chips, he argued, renders the entire subject of exactly what conversion frequency is chosen a mute point. I read up on the installation specs of Analog Devices’ AD1896 chip (supports all frequencies up to 192kHz). This confirmed that the manufacturers of the actual chip sets feel the same way as Stronczer. Our remote e-mail debate left me then and now with only my own two ears to decide what really gives. Just for the record, I do have an assumption. Superior results are a function of skillful implementation rather than subscription to any particular approach.

Subtleties, not subtitles

No French or Japanese translations of "kaboom" in our movie here. Easy does it. This lazy man’s advice to Nirvana proved useful. To get a firm grip on what upshifting from 24-bit/96kHz to 24-bit/176kHz processing accomplished -- besides just going faster for speed’s sake -- I connected the Marantz/MSB front-end to the super-high-resolution duo of Bel Canto Design’s PRe1 and bridged eVo 200.4. (The original EVo nomenclature has recently been changed to eVo.) I reasoned that the more resolution the amplification end of things could bring to the party, the more pronounced the differences between the digital processing speeds would be. If I was to grope for subtleties, I now had ultimate magnification power on my side. Bring it on, digital challengers.

I first resampled the beginning few minutes of "Bésame Mucho" on Spain [Verve 314 561 545-2] over and over. This is a truly phenomenal collaboration between Michel Camilo (Jazz piano) and Cameron de la Isla’s favorite Gipsy flamenco guitarist, Tomatito. It afforded me undivided concentration on just two acoustic instruments. No distractions from percussion, double bass or vocals, no unnecessary compositional complexities. This kept my aural rubber securely on the road. Visibility of the terrain proved excellent. 24-bit/96kHz processing sounded drier and a bit more etched, with less ambient recovery. 24-bit/176kHz processing added audible space. With the guitar, this manifested as a fuller, warmer sound. It sounded less metallic and more wooden. With the piano, the decay of undamped strings lengthened significantly. This infused air and a sense of aura. The piano was no longer isolated in space, sharply defined in its outline against black emptiness. It now spilled over and into the surrounding space and thus made this emptiness visible.

I flashed on Kirlian photography. It registers energetic emanations that radiate from living objects. On specialized film, these invisible auras appear as rays of light that shoot outwards just as clairvoyants perceive things. This type of photography gives further credence to esoteric teachings that talk about how human beings are literally made of "frozen" light. Upsampling to 24 bits/176.4kHz simply "thawed" the light of Camilo’s and Tomatito’s instruments. It released them into space to create presence. If you think about it, presence is really defined by how much space it takes up. An insecure wallflower shrinks its very existence to make a statement by its absence, there but not really there. A highly charismatic person fills and overpowers a room even though the physical form might be diminutive. Sonically, this truism made things sound bigger with 24-bit/176kHz processing. A tape measure applied to the aural images, of course, wouldn’t have shown any actual increases. Still, the enhanced ambient radiance seemed to dimensionally expand the soundstage.

Next up was "Keskin Biçak" from Sezen Aksu’s Deliveren [Post 2000-2]. This is another track and album that currently see inordinate play in my system. Aksu is Turkey’s most celebrated ethno-pop diva. Her newest album is a stunner from beginning to end. I found my copy at a Borders listening station. Despite its exoticness, you should be able to find it pretty easily. I recommend it highly! This particular composition employs a huge drum. After the brief intro, it literally swells and then explodes on every first beat, archaic and massive as Sezen’s voice can be. With 24-bit/176kHz processing, the drum was endowed with more reverb. This made for a more complete closure: initial impact, followed by the energization of the vessel’s physical body, it spilling into the recording venue, then the gradual extinction of the sound like an observably slow camera shutter closing in front of your eye. All this occurs, of course, in a very brief moment in each bar. With the added detail, it seemed a bit longer and more complete though. In the four-piece chorus refrain, I was more acutely aware of the individual location of each back-up singer than in 24 bit/96kHz. Frankly, these particular differences seemed a bit more subtle than those during the guitar and piano track. At this point, I started to feel that the 24-bit/176.4kHz mode revealed more minute low-level detail. This was not the sort that affects timbre or bass extension. It was something less obvious. A complex tune like the Sezen Aksu track made it harder to quantify the individual changes. Simply, my ability to focus on minuscule multiplicities proved insufficient.

Back to basics. Out came a recording by Jazz clarinetist Eddie Daniels (Nepenthe [GRP 9607]). Having played the instrument for years, I’d zero in on minutiae I might miss otherwise. I cued up "Sueòos" and concentrated solely on the Blackwood. Yes, the difference was definitely of an ambient nature. Think of your short-lived Pavarotti-like voice when you first move into a new dwelling. You serenade yourself in the empty bathroom and are much impressed. Once the shower curtain, plants, floor mats and towels move in, your voice reverts back to its normal croak. The recording contract is instantly canceled. Using the Network connection gave the clarinet more empty-bathroom-glow. Rather than just hearing the performer in my room, I now also enjoyed the original recording environment. When Eddie Daniels scaled the upper registers and added some air-pressure emphasis, I could clearly hear the reflections this created off the walls in the recording studio. While this may appear of little consequence, let me point something out. Certain audiophiles make much hay of their systems’ laser-etched-against-black-background holography. This has never struck me as realistic. What I hear in any live venue is an interaction between the performers and their acoustic environment. This causes a blending of various sounds, not razor-sharp separation. Granted, many recording sessions isolate their players into different sound booths. They want to prevent this exact blending to maintain maximum control over the individual microphone feeds. It stood to reason, then, that ambient-rich recordings would benefit most from the added retrieval of said ambiance with the higher processing speed.

To test this assumption, I pulled out two tunes that feature stellar recording quality and plenty of ambient cues. "My One and Only Love," part of the soundtrack to the movie Leaving Las Vegas [UNI/ARK 21] as performed by Sting. And the saxophone and organ standard "Sometimes I Feel Like a Motherless Child," of the album Antiphone Blues by Arne Domnerus (a single track on an out-of-print audiophile sampler).

On Leaving Las Vegas, Sting contributes a total of three standards, recorded in his English castle. A lit fireplace in the background adds the occasional crackle. Otherwise, it’s just Sting, plus bass and piano. The performance is absolutely masterful, the recording quality sublime. Now that I knew what to listen for, the difference was immediately obvious. More space, more dimensionality, more realism. The space behind and around Sting’s voice was illuminated by this presence, creating again a backlit or corona effect. This proved true to an even more pronounced extent with the Domnerus track. It was recorded in some kind of chapel. The decay time is easily a few seconds, and the whole venue is positively aglow with reflections. Unplugging the Network connection caused much of this reflective action to flatten out.

Now that I had a handle on the upsampling effects, I wanted to expand my testing with different electronics. Out went the Bel Canto gear, in went my customary Art Audio PX-25 and the under-review Gill Audio Alana tube preamp. I checked off the same tunes of the first go-around. Same difference. The 24-bit/176.4kHz upsampling mode became the space enhancer. I thought of it as the presence audio gear.

Not all is as it seems

Speaking with Symanski again, I asked what he thought was responsible for the performance increase -- the removal of jitter via the MSB Network, or the higher math of upsampling. Dustin was rather candid with his answer. According to him, upsampling has very little to do with it. In fact, had MSB decided to run 24 bits/96kHz via the Network, Dustin feels the results would be virtually the same. What I heard then wasn’t a math-enhanced resolution increase but a qualified absence of jitter. Hmmm. This naturally begs the question why to even bother going off-board into a separate DAC. After all, this very separation of one box into two creates jitter in the first place. I didn’t feel like pointing this out though. First off, MSB is in the outboard DAC business. What were they going to say? Secondly, I thought I had the answer already.

To verify it, I inserted the Bel Canto Design PRe1 back into the system. I fed the Marantz/MSB combo into input 1 and my resident Cary CD-303 into input 2. Not only are the PRe1’s inputs remote-switchable, the preamp also remembers each input’s last volume setting. The Cary had marginally higher output voltage. For identical levels, it required a minor offset. The memory feature proved priceless and made comparisons a snap. For even further snappadosis, I copied one of my personal CD compilations. After feeding the original and its copy into the two players, I remotely synchronized play. I could now compare differences in mid-track and on the fly. What couch-potato chips reviewers are -- well, at least this one.

Value

I won’t bore you with the details. As far as I could tell, there weren’t any differences. Someone with authentic golden ears instead of my brass lookalikes would have divined performance changes, perhaps -- to me, both front-ends sounded identical. Now we’re coming to the real meat. Let’s do the financial math. The Cary CD-303 retails for $3000. Having heard it and the highly regarded Resolution CD55 side by side, I contend that the Cary is one of the players to beat in its price segment. The MSB bill -- DAC, transport mod and cable -- totals out to $1273 (see breakdown at review’s end). I spent something like $649 on the discounted Marantz. Call the Marantz/MSB duo $2000 after tax then. That’s already a $1000 savings over the Cary player. Now replace the Marantz with a $250 Pioneer DVD machine. While I can’t be sure, of course, I bet the performance differences would be minute. For argument’s sake, let’s say you’d lose 10% of performance. Saving $1400 (or about 50%) against the Cary sure makes a potential minor trade-off pretty appealing. Plus, you’d be able to play DVDs too.

Which brings us straight to the answer why MSB and others continue to develop, produce, market and sell off-board DACs. Despite the jitter problem that needs to be addressed, smart engineering can work around it. Done successfully, it can give you sterling two-piece performance for perhaps half of what equivalent single-box CD playback might cost.

Race cars go far in first gear alone

How much of a difference did I hear between 16-bit/44.1kHz and 24-bit/96kHz processing? A lot less than going the Network route. This gives credence to designer Dustin Symanski’s opinion -- upsampling per se "doesn’t do that much." It could, of course, also mean something else, namely that the MSB Link DAC III is already so accomplished in basic mode that its circuitry doesn’t leave much room for improvement, fancy mathematics be damned. Again, the improvements of the 24-bit/176.4kHz upsampling mode are supposed to primarily be a function of the Network’s jitter reduction, not any upsampled resolution enhancement. Yes, the Network could have been packaged with the 24-bit/96kHz option, for apparently very similar results. But due to DVD-A’s possible future, the 192kHz sampling-rate standard became a natural choice for MSB’s proprietary low-jitter interface. It recalls Audio Alchemy's I-squared-S bus.

As an end user, I don’t really care what the reasons are for the increased realism of the DAC’s 192kHz input option. The improvements are real. Especially in an already rez-optimized system, I find them worthwhile to take those few additional steps towards the suspension of disbelief. As a reviewer, I call it noteworthy that another engineer stepped forward to put yet another dent into the credibility of the number’s war of upsampling. John Stronczer of Bel Canto Design and Kevin Halverson of Muse had already said much the same in my earlier background piece on this subject. The true magic, they said, lies really in the combination of choices a good digital designer makes about the digital filter, the analog reconstruction filter, the power supply and DSP chips. In other words, implementation. This can include upsampling, but doesn’t have to. Upsampling per se isn’t some tooth-fairy panacea that can successfully cover up digital design deficiencies elsewhere (now you know what DDD really stands for).

Kaboom

In the final analysis, how much of L.A. did Denise Flopper really destroy when she whacked that red 24-bit/192kHz upsampler button?

For those who read too much into reviewers’ reports, I’m happy to say that much of L.A. is still standing. Rodeo Drive’s definitely gone (good riddance), and parts of the valley seem smoked. If you care enough about the aural subtleties of space and presence, I can highly recommend investigating the MSB Technology Link DAC III with 24/192 input option. If the upsampling isn’t what caused the improvements as I first assumed, the MSB Network’s jitter suppression surely works as claimed and must be considered the real feature of this option.

Should you love Los Angeles as it is, don’t bother. Even in standard 16-bit/44.1kHz mode, the basic Link DAC III at $399 is a superlative performer. As you climb the corporate audio ladder, the differences become more and more subtle as you leave the already elevated ground-floor office of the MSB. Unless you’re a card-carrying audiophiliac, your ongoing sanity might be better served if you did forego the endless struggle for the top floor. If you disagree, don’t blame me later when your wallet is depleted and you still haven’t arrived.

...Srajan Ebaen
srajan@soundstage.com

• The Earmarked! Current Issue