This review page is supported in part by the sponsor whose ad is displayed above
The digital inquisition
While I still had Mark Gurvey's ear, I asked what most existing G25U customers leash to it. After all, the majority of digital machines lack clock-link inputs. That seems to render the reclocking feature mute. To boot, many dedicated CDPs already sport their own internal 24/192 upsampling. That defeats the other half of brother G.'s feature set. Unless Mark was to tell me that serial upsampling -- putting two upsampling circuits of the same conversion frequencies behind one another -- had unexpected benefits. Not quite.


"It is possible that if an existing playback device with upsampling has a defeat option for that circuitry, the user may benefit from the G25U's upsampling. More importantly, he will benefit due to Esoteric's DSSRL3 circuit. This circuit reduces injections of jitter noise into a digital PCM data stream because it regenerates/retimes that stream thought the upconverter and internal crystal clock prior to output. DSSRL3 drops the noise floor of any form of jitter by up to 1/100th below its original level. However, to use the clock sync function with any other playback device, you need clock-in or word-in. Or you need to route it between a transport and DAC. Those are the only two ways to use that master clock sync feature. All of our video-capable units are rated +/-10ppm so the G25U makes good sense with a dramatic impact. Many (even high-end) players don't publish their jitter spec at all. 10 to 1 is easily heard. 3 to 1 is very subtle but once you take out the G25U after a day or two, you should hear the benefits. Most players are running from 30-50ppm or worse.


"As for upconversion, consumers who have bought the G25U for that alone tell us that they use it with players that have a PCM output but no upconversion. In fact, digital cable and network customers are some of our best users, in addition to conventional CD, DVD player or FM tuner customers with PCM digital outputs. It's a niche no doubt but consumers have bought the G25U, from our research, about 30% of the time for the clock sync and/or DSRLL3 retiming and about 70% of the time for the upconverter.


"DSRLL3 is our 3rd evolution of a digital servo-ratio-locked loop which helps to reduce the generation of jitter in the signal path for digital transmission. This allows higher purity. Input data is stored momentarily in RAM, then output with the time axis controlled by a quartz crystal generator. Since DSRLL3 can lock accurately to any sampling frequency between 28 and 54kHz, input signals from such digital sources as satellite broadcasts (32kHz), CD (44.1kHz) and DAT/DVD (48kHz) can be converted to an output signal with improved time axis accuracy. From any PCM source, DSRLL3 provides a smoother output virtually free of jitter that may have been present in or caused by the signal path itself."

Thus far, I'd labored under a misapprehension. I had believed that without Esoteric's BNC-carried reclocker terminals duplicated on a non-Esoteric machine, you couldn't avail yourself of its internal master clock function. What Mark seemed to be adding now was that simply by using any of the conventional digital i/o ports (single-ended, balanced, Toslink) found on all digital separates, one could access the G25U's reclocking oscillator. One had to merely enter/exit the G25U through its digital i/o ports and presto. That meant inserting brother G between transport and DAC just as one would any other anti-jitter (or anti-digital as it were) box. No special clock-link socketry required. Hey, some of us are slower than others. If that's what the man was saying. Again, not quite.


"In a nutshell, retiming occurs within the data stream when PCM is input and output via our DSRLL3 circuit. But that is retiming the digital data, not relocking the transport. To reclock a transport, you need the word sync or clock sync input (or be able to loop a clock sync signal between a DAC and transport). For example: use the G25U as master clock to the input of a DAC. Now the D/A converter becomes slave to the G25U and the transport becomes slave to the DAC. This makes the G25U the master clock-out device, with the DAC set to clock/word in and connected to its transport set to word/clock in. Now you have one heartbeat from the G25U running all devices at 1ppm. Typical interface connection for word-out to word-in is a simple coax cable with BNC connectors. That's why we supply three cables for the clocking purpose (one for each output). If your transport, player or DAC does not have BNC clock-in/out, then RCA-terminated coax can be substituted."


For the technically inclined, here's further information on reclocking: "As of 1967, the International System of Units (SI) has defined the second as 9,192,631,770 cycles of the radiation which corresponds to the transition between two energy levels of the ground state of the Caesium-133 atom. This definition makes the Caesium oscillator (often called an atomic clock) the primary standard for time and frequency measurements...


"The core of the atomic clock is a microwave cavity. It contains ionized gas, a tunable microwave radio oscillator and a feedback loop to adjust the oscillator to the exact frequency of the absorption characteristic as defined by the behavior of the individual atoms. This adjustment process is where most of the work and complexity of the clock lies. The adjustment attempts to eliminate unwanted side effects such as frequencies from other electron transitions, distortions in quantum fields and temperature effects in the mechanisms. For example, the radio wave's frequency could be deliberately cycled sinusoidally up and down to generate a modulated signal at the photocell. The photocell's signal can then be demodulated to apply feedback to control long-term drift in the radio frequency. In this way, the ultra-precise quantum-mechanical properties of the atomic transition frequency of the Caesium can be used to tune the microwave oscillator to the same frequency (except for a small amount of experimental error). In practice, the feedback and monitoring mechanism is much more complex than described above. When a clock is first turned on, it takes a while for it to settle down before it can be trusted.

"A counter counts the waves emitted by the radio transmitter. A computer reads the counter and does the math to convert the number to something that looks like a digital clock or a radio wave to be transmitted. The real clock of course is the original mechanism of cavity, oscillator and feedback loop that maintains the frequency standard on which the clock is based. A number of other atomic clock schemes are in use for other purposes. Rubidium clocks are prized for their low cost, small size (commercial standards are as small as 400 cubic centimeters) and short-term stability. They are used in many commercial, portable and aerospace applications. Hydrogen masers (often manufactured in Russia) have superior short-term stability to other standards but lower long-term accuracy.


"A Rubidium standard is a frequency standard whereby a specified hyperfine transition of electrons in Rubidium-87 atoms is used to control the output frequency. A Rubidium standard consists of a gas cell, which has an inherent long-term instability. This instability relegates the Rubidium standard to its secondary status behind the atomic Caesium clock. Often, one standard is used to fix another. For example, some commercial applications use a Rubidium standard slaved to a GPS receiver. This achieves excellent short-term accuracy, with long-term accuracy equal to (and traceable to) the U.S. national time standards.


"The lifetime of a standard is an important practical issue. Modern Rubidium standard tubes last more than ten years and can cost as little as $50. Caesium reference tubes suitable for national standards currently last about seven years and cost about $35,000. Hydrogen standards have an unlimited lifetime."


As far as where this type of research is headed, "most research focuses on ways to make the clocks smaller, cheaper, more accurate and more reliable. These goals usually conflict. A lot of research currently focuses on various sorts of ion traps. Theoretically, a single ion suspended electromagnetically could be observed for very long periods, increasing the accuracy of the clock while also reducing its size and power consumption. Some researchers have developed clocks with different geometries of ion traps as well. Linear clouds of ions usually have better short-term accuracy than single ions but there are trade-offs. The best developed systems use Mercury ions. Some researchers experiment with other ions. A particular isotope of Ytterbium has a particularly precise resonant frequency in one of its hyperfine transitions. Strontium has a hyperfine transition that is not as precise but can be driven by solid-state lasers. This might permit a very inexpensive, long-lasting compact clock."

Where audio clocks are concerned, the current hierarchy of applications is as follows, again as per Gurvey: "The most accurate would be an atomic clock but that's cost-prohibitive at present. The next accurate is Rubidium, a sub-atomic particle generator outputting an oscillating timing signal accurate to +/- 0.5ppB. The next accurate are various crystal/quartz oscillators, typically accurate from 1 to 50ppm depending on cost, architecture and implementation. The next accurate are quartz or other timing devices which usually range from 50 to 150ppm." So much for clock talk - except that the G25U can be linked to the G-0s via its external clock input to go Rubidium and that each of Mr. G's three word clock outputs can be independently while synchronously set for different upsampling rates.


The other welcome reminder and upshot off all this esoteric tech stuff this was to think of jitter as audible noise. I'd always understood is primarily as timing errors, somewhat analogous to how various crossover topologies in speakers rotate phase to stagger/delay parts of what should arrive at the listener's ear as a unified wave front. But Mark was also referring to lower jitter as noise reduced by a significant margin. Up to 99% in fact. That's a lot. It's a function of Esoteric's DSRLLIII circuit regardless of whether clock-link is engaged or not (incidentally, that circuit is based on the Analog Devices 1893 large-scale integrated chip). Should one think of the G25U more properly as a digital noise scrubber then? After all, audiophiles with their widespread acceptance of severe timing errors introduced by conventional speaker crossovers often believe that timing errors are inaudible. However, those same audiophiles happily spend the long green on extravagant power cords or conditioners to lower their system noise floors. Noise is something our kind relates to a lot better than the somewhat intangible and vaporous time domain. All this by way of demonstrating how my mind was still trying hard to wrap itself around this entire subject of retiming vs. reclocking. And how, exactly, jitter affects the subjective qualities of musical playback at these already ultra-low values. Would the X-03 turn out to be too good to demonstrably and categorically benefit from the G25U?


I should also tell you that the entire intro up to this very point was penned well before ever taking the G25U out of its stout shipping carton. After page 1, I was inclined to think that I'd have an awfully hard time to hear its effects. If at all. With the additional information of page 2 -- specifically as it relates to jitter noise suppression -- I'd flip-flopped believing that perhaps things wouldn't turn out to be quite as hair-splitting after all. It's a fair subject to bring up that reviewers in particular must separate conceptual expectations from actual test sessions. Otherwise you'll hear what you expect to hear. Even price and build quality can influence this. Though the opposite holds true as well. As regards the mighty G-0s for example, learning that commercial Rubidium clocks are available for $50, its ultra-stout $13K sticker could create an inverse reaction without knowing more about the actual implementation. (400 cubic inches as referenced in the above text would certainly not fit inside the big G's chassis and we all know how costly miniaturization can be). But enough now of back story and technology. Could the G25U tango despite its nerdy nomenclature?