We need extreme freq stability for our master freq reference for a particle accelerator that is 770 meters long.? The original 46 year old oscillator was first a HP107BR special that was at 6.289xxxx MHz, then multiplied up with a TCI multiplier assembly. Later it was HP105B, again special crystal in double oven setup. Backup was a Vectron oscillator. Four years ago, a digital engineer replaced the crystal sources with a Meridian End Run GPS-disciplined oscillator and a TCI phase locked assembly with multipliers. This is used to generate a 100.625 Mhz, 201.25 Mhz and 805 MHz reference. With phase stability giving trouble, I took to comparing the GPS ultra-stable reference to the HP105B, 10 MHz vs 6.289xxxx. I used a HP 5372A dual channel frequency analyzer, and recorded histograms and stats of each oscillator. The GPS-disciplined oscillators are clearly inferior by a factor of 10 for medium term stability over a few hours. Two of them were tested. Then I found an old Racal Rubidium oscillator in VXI format. It works great. The moral of this story is, specs on paper aren't always enough. The GPS oscillators claimed 10^-11 and such, over 24 hours. But over shorter term even, they jumped around and covered a few millihertz spread. Stanford Research has a nice crystal oscillator that is also available as a Rubidium oscillator option. These are not very expensive.
