After giving up previously, I finally found some details on ifi's GMT clock besides the vague blurb included in the list of features on ifi's Zen DAC V2 page. The relevant page was apparently part of an internal ifi presentation on various technologies used in ifi products, by their R&D division, for a business-oriented division.
The GMT clock was apparently developed for AMR's high-end gear, and has become ifi's universal clock, the best grades of which apparently have jitter on the order of 0.3 pS, and a very clean spectrum from 9 kHz to 15 kHz (which is all that's shown in the relevant plot, above, so perhaps it's the critical range). For information on how such low jitter can be obtained, see Analog-to-Digital Converter Clock Optimization: A Test Engineering Perspective.
Whatever the Zen DAC V2's jitter-rating is, its high end is as clean as any analog gear I've heard, which includes moving-coil cartridges and 45 RPM direct-to-disc LPs. This is partly due to the fact that it uses the TI/Burr-Brown DSD1793 Advanced Segment DAC, which has low jitter-sensitivity, low noise in general, and 24-bit resolution (see my Zen DAC V2 review for details). Naturally, the quality of the high end depends on the source material, and the cleanest CD I've heard is United We Swing by the Wynton Marsalis Septet.
Reading the aforementioned presentation-page was the first time I'd run across the term "femto clocks," although I've done quite a bit of searching on the subject of jitter. It's another indication of how poorly the audio media has kept us informed of significant advancements in digital audio playback. But then considering that it's clearly downplaying the significance of audio polarity, it's apparently deliberately keeping us in the dark so that we'll flail about trying to obtain musical satisfaction from digital audio, and waste a lot of money on playback gear and recordings in the process. There are probably other, even darker motives, but what matters is that playing good CDs by means of well-designed digital playback gear which uses TI/Burr-Brown Advanced Segment DAC-chips, with a decent system otherwise, and with the correct audio polarity (although some CDs aren't very sensitive to polarity), is the least expensive approach to being able to listen to a particular piece of music when you're in the mood for it, and to obtain musical satisfaction from it.
Without further ado, here's the text from the web-page I found about GMT:
Femto Clocks – Picky about Phase Noise
Background
All Femto clocks are good, very good in fact. They exhibit jitter levels lower than most clock crystals which leads to better sonics. An oft quoted benchmark for accuracy is Femtoseconds (Fs) / parts per million [Fs for jitter, ppm for frequency error].
Back in 2008, before the word “Femto Clock” became all the rage, AMR developed a special type of clock in DP-777 [a $5K digital processor] as part of the “Global Master Timing” (GMT) and “Jitter less” technologies (why special? See below, because not all Femto clocks are the same).
We called it the GMT Clock platform (which is comprised of specialised hardware+software) as it is not just buying a “clock in a can” and job done.
Having worked with all sorts of clocks, including discrete, Rubidium, Superclocks and not the least Femto clocks over the years, we know them quite well.
All Femto Clocks exhibit excellent low phase-noise (measured jitter within the clock). However, as their origins lay in being part of SONET, the popular SONET targeted “Femto-clock” is less desirable as its best phase-noise performance is concentrated in the > 12KHz region (read: at the very top and way above the audible band, so benefits audio less).
As an example, this link highlights the use of Femtoclock technology in the telecommunications sector where they are spec'd for.
This [referring to the relevant jitter-spectrum plot at top of page] is an Optical Comm system (aka SONET - which is a Subset). We added the blue line to highlight the -70dBr region so that when referenced to the AP2 chart in the next section of the micro iDSD, it is more of an “apple to apple” comparison.
The spike at 50KHz is the "signal" As you can see, for quite a few KHz around this region, phase noise is low, this is what matters in this application.
However, the area around the green arrow is the most crucial human audible range of 20Hz > 20kHz where phase noise performance is less impressive in the region of -100dB to -70dB.
Explanation
Therefore, the key for AMR was to design a new system, the “GMT” clock platform which not only exhibits the lowest phase-noise in the crucial audible band, but offers precision (< 0.004ppm tolerance) adjustability with literally millions of possible frequencies (as per the DP-777 "GMT" Technical Paper).
The GMT Clock system designed into the micro iDSD measures <280 Fs, comparable to many Femto-Clocks (because it was designed to give low jitter).
From the [GMT jitter-spectrum plot above], you can see that jitter in 9kHz > 15 kHz is very good, the micro iDSD noise floor goes all the way down to -150dB which is virtually across the board with no spikes.
How this benefits the user
Consistent, across the board negligible jitter means timing is supreme, with just the right amount of attack/decay and of course, tonal accuracy. We are really pleased with the very low jitter performance of the micro iDSD in the most crucial audible range – in fact, we would not mind if customers pit it against significantly more expensive DACs.
We hope you found this interesting as it sheds some light on the particular attention we have paid to parts performance and custom design in the micro iDSD (actually, we took it from the DP-777!).
Addendum: What about Rubidium Clocks?
[Above] is a chart of several types of rubidium clocks. What they all exhibit is many sharp spikes in phase noise. Even though they measure well, some down to -150dB, when they spike, noise levels jump up to -70dB to -90dB.
This is far from ideal which is why we have not used such clocks, neither in iFi nor AMR products. It all boils down to paying close attention to the specific clock/s used and its performance in the audible range.
