According the Brian in this post, the clock in Buffalo II is not only the top-of-the line product from Crystek, but a custom made model with even better specifications.
For those with sharp eyes, the photo of the Buffalo II prototype has a 50 ppm (temperature stability) clock, but the production models will have a 20 ppm part.Temperature stability is related to the quality of the quartz crystal used in the clock. It is therefore safe to assume that this will also improve the jitter figures.
But how good is Crystek's best oscillator?
There is a white paper by Crystek showing "phase noise" curves for the 950 oscillator and compared to a typical oscillator. I copied the charts and overlaid them together. In addition I took a look at the specifications of a ultra low phase noise Wenzel oscillator and a Rubidium oscillator and plotted the data on the same chart. The Rubidium is a "$900 low cost" version. Perhaps they are best at frequency stability over a long period of time and therefor they cost so much. The Wenzel oscillators are the lowest phase noise you can find and are used in High Energy Physics and Radio Astronomy applications. If you work for NASA or CERN, maybe you can borrow one of these clocks and hook it up to your DAC :-).
BTW, "phase noise" is just another way to measure jitter. Jitter, typically measured in psec is the area under the phase noise curve. Thus jitter gives you the total deviation from ideal and phase noise gives you the spectrum of that deviation. However, comparing jitter figures is not easy because manufacturers select different frequency range (say 100Hz to 100KHz or 10Hz to 1MHz) and each gives different numbers. Best to compare phase noise numbers. The graph below shows the phase noise curves. One can see that in order to improve on the clock selected for the Buffalo II DAC, one would have to get a Lab grade oscillator
Update: NEL AE-X3A3 oscillators have better specifications than Crystek 950. The NEL Oven controlled oscillator is even lower noise (not sure how much they cost)
(Clik for larger image)
How is Ultralow Phase Noise Achieved?
According to the author in the Crystek white paper,
"A commodity oscillator is nothing more than an ASIC and a quartz crystal blank. In most cases, it does not even have an internal bypass capacitor. The crystal blank is an AT-cut strip with Q of about 25 K to 45 K. This low Q limits the close-in phase noise. The ASIC with all its transistors limits the floor noise to about -150 dBc/Hz. On the other hand, the true ultralow phase noise oscillator uses a discrete high-performance oscillator topology with a packaged crystal with a Q greater than 70 K for excellent close-in phase noise. The discrete oscillator topology establishes the SNR, and hence the floor is lower than -160 dBc/Hz. Therefore, superior performance is obtained with very high Q crystals and a good discrete topology."How is the Phase-Noise graph obtained? The chart is one half of the power-frequency distribution of the variation from ideal. Instruments are available to generate such charts.
manual has a good tutorial on phase noise
Quarts oscillators is an "old" technology. "Fundamentals of Quartz Oscillators", an application note from HP from the late '70s is a good read.