Jbeale designed a high precision decade frequency generator based on a 26Mhz ovenized oscillator with stability of +-1 part-per-billion, or +-0.0000001%. The generator is controlled with a PIC16f1823 microcontroller, which provides interface and frequency division.
Firmware, PCB and schematic are uploaded on his site. He also provided a link to Tom V.B. impressive site that could best be described as:
…one man’s quest for the most accurate clock…
Via the forum.
Actually it’s stable to one-part-per-billion (that’s 1000x than 1 ppm) over a time span of about 1 hour. Over longer periods it can drift more, but it would take a very very long time to drift as much as 1 ppm.
they ebay link you provided from the BOM states
Stability: ±100ppb, I misinterpreted your post, thinking that 1PPB was a typo, sorry
You are correct that the mfr. datasheet at http://pletronics.com/uploads/datasheets/ohm4%205.0v.pdf calls out +/- 100 ppb (per day) for the full temperature range of -40C to 85C. However that is quite conservative for a typical indoor environment. My experience with a few of these parts, in my apartment (which has a delta-T of only a few degrees C) shows typically 1 ppb drift per hour.
I corrected it in the post, thanks for pointing it out.