This review page is supported in part by the sponsor whose ad is displayed above
Reclocking
This feature can be manually disengaged in three ways. One's available on the X-03, two are on the G25U. On the X-03, you can press the blue-encircled Clock Mode front panel button. No lit blue circle, no clock mode. With the X-03 in clock mode, you can alternately cycle through G.'s Clock Generator settings, for the A, B or C output you've connected the X-03's clock input to via one of the included BNC-to-BNC cables. The defeat option hides between the 176.4/192kHz and 44.1/48kHz options. It confirms by showing no blue LED in that bank; by turning off the X-03's clock mode button; and by showing a brief No Word in its display. Lastly, you can set the Master switch of the G25U for the 48kHz option. The X-03's clock mode button will blink in response, indicating that as a CD/SACD machine, it cannot lock onto the 48kHz frequency though it keeps trying for as long as its clock mode remains selected (and it'll continue to output sound regardless, simply not reclocked).

Clock mode works with or without upsampling. For CD playback, the latter naturally corresponds to the 44.1kHz option. If that's selected, the corresponding blue LED lights up on the G25U while the display of the X-03 doesn't react. Move up to 88.2, however, and the X-03's display reacts briefly with Word 88.2 before it defaults to its regular track listing. Ditto for 176.4. The next setting on the G25U's Clock Generator bank then is word defeat again. This elicits a brief No Word on the X-03 and turns off the clock mode button. Of note about Esoteric's scheme is its synchronous nature for both CD and DVD. The upsampled target frequency is an integer multiple of the native frequency. Hence RedBook upsamples either two times to 88.2kHz or four times to 176.4kHz. DVD's 48kHz base frequency reaches 96kHz or 192kHz on the same whole-number multipliers. Clearly Esoteric doesn't champion asynchronous upsampling. That's rampant elsewhere and ups RedBook's native 44.1kHz to 192kHz via more complex interpolation. It rebuilds every single sample, i.e. doesn't use a single original data point from the disc you insert. A brief explanation of these upsampling differences follows. It was penned by Jeffrey Kalt for his Resolution Audio website. He likewise favors synchronous upsampling.

Asynchronous vs. synchronous upsampling
"I have yet to see an official distinction between upsampling and oversampling. From recent usage, I can only conclude that upsampling is any technique that increases the sample rate, for example from 44.1 to 96kHz. Oversampling, in contrast, is a form of upsampling where the rate is increased by an integer multiple, that is, 4 x, 8 x etc. For audiophile purposes, both processes are used for one reason: to reduce the artifacts of the digital to analog converter. No new information is created. Once the original has been sampled at 44.1kHz, anything above 22kHz is lost forever. The whole idea of upsampling is merely to make the D/A converter behave more ideally and introduce fewer artifacts into the analog audio signal.


"About 7 years ago, a new technique was developed to do asynchronous sample-rate conversion using inexpensive silicon. The original application was for professional use, where mixing digital recordings made from different master clocks required either conversion to analog and back to digital, or had to be done on an expensive computer workstation. The advent of this new silicon allowed studio consoles to provide real-time mixing of these different tracks. Some designers in the audiophile community began using them as jitter-reduction devices, even though they introduced new artifacts as a result. Then some suggested that upsampling was a substitute for higher-resolution formats, implying that "upsampled CD" is comparable, or nearly so, to DSD or 192kHz/24-bit PCM.


"But alas, there is no free lunch. For if this were true, digital audio could be distributed with a sample rate of, say, 100Hz, and then simply upsampled upon playback. There is no question that different up/oversampling methods sound different. Consider that with 16 x oversampling, 15 new data points are interpolated between successive samples. Every 16th point that comes out of the filter is the original data straight off the disc. The intermediate points require additional precision, hence the common use of a 24-bit filter. In contrast, converting from 44.1 to 96kHz represents a 2.17687... ratio. This means that every output sample is the result of a very complex calculation and invariably contains some artifacts not related to the original music signal. Every upsampling product I have heard sounds significantly better at integer multiples - for example, 176.4kHz instead of 192kHz on CD playback."

Pressing them buttons
X-03 owning, button-pushing buckaroo banzais will quickly be disappointed. Simply changing settings in quick slot-machine succession will only render you subjectively deaf. You'll quickly conclude that you can't hear. Any mondo differences. This is a far more subtle game indeed. It requires that one first learn in which domain these settings work. Learn what to listen for. The better a recording you use -- and better here means, recorded with real ambiance -- the quicker you catch up with what's going on. Gurvey's suggestion (which mirrors what many audiophiles already know) is sensible. Live with the reclocker engaged for a while. Get accustomed to its presentation. Then turn it off. This subtractive approach is far more relevant than attempting to discern anything the additive way. Incidentally, emanating from its chassis, my unit produced double-pitched mechanical hum well above the 60Hz power supply band. This was audible in the listening seat at night without traffic noise and signal but completely masked once music played.


Having the new Bel Canto Design PRe3 and S300 pre/power duo for review as well as new cabling by Furutech, I decided to throw all the newcomers into the pot. The outcome was an altogether unfamiliar but highly resolving stew of exceptionally low noise floors. After all, the purpose of this exercise wasn't the final sound per se. It was simply about the difference between the X-03 solo and the X-03/G25U combination. My recording of choice was long-time fave, m.a. recordings' Sera Una Noche [M052A]. It's a masterful nouveau tango meditation on stylistic deconstructionism. Recording quality doesn't get any better than this. Equally important, the playing and compositions are of such a high level that even the repeat listening required to chase the dragon's tail of hairsplitting doesn't wear out its welcome.


The gist of it
What I eventually - um, locked on to were marginally more articulated leading edges. And marginal means just what it says. However, once identified, I could play the Rococo game of "variations on a theme". I could focus on the transients backwards, away from the lead singer or instrumentalist to secondary or tertiary sounds and acoustic interactions farther back in the soundstage or deeper inside the musical fabric. The difference then, between reclocked and normal modes, was simply the extent of how far backwards I could extend this game. Of how small I could make the focus of my attention and still be rewarded with noteworthy substance.


Because these are ultra fine distinctions, they don't really operate within the softer/sharper realm such as significantly more piquant transients would suggest when compared to more rounded-over, less delineated ones. The reclocked sound isn't sharper per se. The upshot in the end is simply just a spatially somewhat fuller, more acute impression of presence. Without a lot of effort, you'll not know why or how it sounds just a bit more live, jumpy or there. It just does. Some might call it more "pop", referring to how sounds stand out from the surrounding silence and lock in more. Others may call it slightly enhanced visibility, referring to how when you focus on percussive sounds zipping through space and hitting a wall in the recording venue, you can trace the rebounds better (or at least imagine that you can).
I'm positive that the GU25's reclocking contributions are audible. I'm far less positive that we can categorically and unambiguously identify why and how. That quite seems a slippery slope greased by imagination and wishful thinking. I'm more comfortable saying that as you live with the reclocker engaged and then take it out of the circuit, a sense of intimacy or directness gets less intense. That much is clear. Far less clear is assigning value to mating the G25U to the bloody excellent all by itself X-03. Consider how it approaches 50% of the X-03's asking price but as a box is far more empty than filled with stuff. Frankly, I can readily imagine far greater advances accomplished with $3,000 applied elsewhere to the endless component upgrade game. The X-03 simply maxes out what's relevant to such an extent that precious little headroom remains to make an appreciable difference (unless you compare it perhaps to a far more expensive component).


To do the G25U proper justice, I had to back away from the X-03 and hitch it to a transport with presumably looser jitter tolerances than the very latest
VRDS miracle sled. The $1,850 Opera Audio Reference 2.2 Linear to the rescue. I ran that into my Zanden Audio Model 5000 Signature DAC and inserted the G25U in-between. Remember, the latter is a D-to-D converter. It never exits the digital domain. To make sound, you'll need to reenter the analog domain with an outboard DAC (or a one-box CD player that has digital i/o ports to allow looping through the Esoteric). Because the Zanden's 1985 Philips 16-bit/44.1kHz chip can't lock onto any upsampled data streams, the G25U's only possible positive contribution now was its retiming of the data stream. No reclocking the Linear transport, no upsampling, just a lowering of jitter by an unknown quantity (Opera doesn't publish a jitter spec for its player).


Even though not relevant to this particular exercise, the sound of this setup compared to the Esoteric front end was definitely softer and fluffier, obvious enough to be detected a half minute into the first Sera Una Noche track. As already noted in the X-03 feature review, this player -- whether a function of its transport or DAC section I don't know -- is ferociously dynamic to clearly go beyond the Zanden separates in this regard. Comparing retimed vs. original data with the Opera piece as transport involved physical cable swaps. The Consonance lacks dual digital outputs that could have fed the Zanden DAC's multiple inputs, one looped through the Esoteric, the other going direct.


Thankfully, the differences were somewhat more pronounced than before. Same taste, just more of it. More Sam Tellig-style there. But $3,000 worth more thereness? That's the Hamlet question. We're well on the far side of diminishing returns here. In a truly dialed system, bona fide improvements -- rather than differences -- are harder and harder to come by. Chances are, truly dialed means low jitter already. Depending on how low, the G25U could equate to gilding the lily. With video sources of lower intrinsic jitter rejection, I'd imagine far more dramatic results, of the same nature as already described but so much more pronounced as to elicit the desired "Hot damn!" response that eluded me.
Conclusions - more than one!
As a mate to the X-03, the G25U is a last-resort option. I'd first spend $3,000 elsewhere to maximize the quantity of improvements before I'd spend that amount on the subtle though desirable effect the reclocker makes when you already run a source of the X-03's caliber. If you own a one-box machine, the G25U can't be used unless your player offers an uncommon digital input besides a digital output. Whether the expense of the Esoteric plus two digital cables will warrant the results in such a case I cannot predict. If you currently own digital separates but not an inferior transport -- i.e. a discarded CD player with a digital output or a cheap DVD player used solely as spinner -- the G25U may still be the diamond-encrusted fashion watch; expensive but no more accurate than the cheap Casio quartz where it concerns relevance in day-to-day living/listening, i.e. not theoretical superiority.
If you do use a cheap Pioneer DVD player into an affordable DAC like a Bel Canto or Birdland Audio, chances are that your system could benefit from $3,000 spent elsewhere more than on the G25U. While I don't dispute its effects, I simply didn't have a "bad-enough" front end on hand to demonstrate the kind of performance increase I'd call commensurate with the expense. In the end then, I'm not completely sold yet on the viability of this technology for the vast majority of audiophile customers. Alas if, as Gurvey says, video customers make up the majority of Mr. G's actual audience, I was clearly the wrong chap to hit up for this review. Hence keep my assessment in that context.


There is a flip side to this, however. When I compare the $10,500 Esoteric X-03/G25U combo to the $43,000 Zanden Audio 2000p/5000s combo, I think that 7 out of 10 listeners would sonically favor the Esoteric presentation. Once you told them the price difference, all 10 might opt for the Esoteric. That's very significant. How many audiophiles play in these leagues to begin with is a different matter. If you do, the X-03/G25U combo gets my highest recommendation. It's the Zanden's equal and to some ears, might actually surpass it. Just don't think you need the G25U to enjoy the X-03. You'll get 97% of the way without it. While I suspect that my reverse endorsement for this mighty CD/SACD player wasn't exactly what Mark Gurvey had hoped to get out of this particular review, it's the underlying conclusion I took away from it. After all, if the razzle-dazzle technology implemented in the reclocker operates with such subtlety on the X-03, it means that the latter is scarily good on its very own terms.


When it comes to squeezing the last drop of performance out of your digital oranges -- imagine I'd said lemons instead -- the reclocker will be considered a luxurious, final but vital addition to your tool chest. After all, it operates in a domain that's otherwise inaccessible - digital retiming. The audible effects are similar to superior resonance control, albeit without the latter's possible tonal balance changes. However, reclocking data that's already very low in jitter by itself simply constitutes an example of how today's technology can nearly eclipse relevance when dealing with the perception mechanisms of ordinary humans. Think Olympic games. Some winners can only be determined with slow-motion replays and highly sophisticated clocks linked to auto-triggered cameras. To the naked human eye, two guys crossed the finish line at exactly the same time.


Is reclocking the new upsampling then? Not by a long shot. If present samples of reclockers are any indication, their implementation isn't a function of "cheap silicon" as Jeffrey Kalt characterized the asynchronous upsampling craze. Outside the $1,500 Apogee Big Ben and a Weiss unit reader Howard Lindo alerted me to, the G25U is the most affordable external reclocker on the market I'm aware of - though it's certainly not cheap. Because the reclocking rather than retiming function does require dedicated socketry which is exceedingly rare at this point -- I believe it might be a near Esoteric exclusive on one-box players, transports and DACs in the consumer sector -- this component segment promises to remain very esoteric for a time to come. It's a highly specialized application niche. It would be interesting to speculate what the trickle-down effect could be, eventually. Esoteric's new Class D integrated/reclocker could be a harbinger of where things might be headed.

PS: A reader forwarded the following link to Ivar Løkken's PhD.-level exploration of digital conversion issues, stating that "the ppm rating for a clock is not its jitter performance but rather its frequency precision. A clock can have high jitter but low ppm." To quote a few pertinent items from the paper, here goes but the technically savvy or interested reader should read it in its entirety (which will also give a different - er, appreciation for various marketing claims in this field): "It is very important to distinguish between different forms of jitter in a digital audio recording or playback chain. The only one that really matters is sampling jitter, deviations in the sampling interval, which will directly translate to distortion in AD or DA conversion.However, we also define clock jitter, which is jitter on the clock generation circuits in the system, as well as interface jitter which is jitter introduced in the transmission of digital signals. CD discs even have physical jitter, often referred to as landscape-jitter and pit-jitter. All these are secondary distortion sources which will again lead to distortion. Good jitter rejection performance is thus a critical part of high-performance data converter design...


"It should also be noted that in real life the jitter function j(t) does not consist of a single frequency but of many distinct frequencies as well as white noise (white jitter). Thus it will produce many distinct sidebands in the output spectrum as well as an elevation of the converter's overall noise floor... The audibility of jitter is also an important factor. When the jitter frequency is low, the sidebands produced fall near the frequency of the signal itself. Then the sidebands will be subject to masking. High jitter frequency on the other hand is much more audible since the sidebands will not be close to the signal. The sensitivity increases rapidly for jitter frequencies from 200Hz to 1kHz. Above 1kHz, it flattens out somewhat, since the signal frequency that can produce unmasked sidebands increases... We can see that the audibility threshold decreases from 500ns at low frequencies to as little as 20ps at 20kHz. Especially when using formats or converters with high sample rates, this will be a major issue. The high-frequency sample-jitter must be rejected vigorously...


"In the case of jitter, there is no substitute for good engineering. Although jitter is a digital problem, it is of an analog nature and is best fought off with good analog design. All parts of the interface chain should have very high signal bandwidth but still low noise. A clock recovery circuitry with good PLLs and high jitter suppression is critical, and the DAC itself should be designed with low sample-jitter sensitivity. However, it has been shown that as we reach and exceed 20-bit effective resolution, the jitter levels have to be extremely small if they are not to degrade performance. Although one can argue from a psychoacoustic point of view that an ENOB of 20 or lower is mostly of academic rather than practical interest, engineers will continue to push the frontiers. And to be able to do that in audio converter design, research in jitter and the challenges related to it is of utmost importance..."
Manufacturer's website