Technical:<p>When I learned exactly how an HP 5061 atomic clock works, in the process of repairing them.<p>The tube is a marvel of modern engineering, a physics lab in a package small enough to hold in your hand. In the end, it takes quantum mechanics, and turns it into a razor thin filter/rf detector. If you put in 9,192,631,770 Hz, you get a strong DC signal output, a few hundred Hz higher or lower and the output drops off rapidly.<p>HP engineers did some VERY clever work. They start with a high quality crystal oscillator, in an oven to help stabilize it's output. They multiply it's output frequency up to 9.2 GHz using a PLL to get the odd bits, and straight harmonic generators for most of it.<p>A small phase modulation is added to the signal, which in effect spreads out the probe frequency enough to allow finding the peak output. As the equipment starts up, it's precise enough to drift into detection range once it gets up to temperature.<p>The gentle rocking back and forth of the clock frequency to probe the sides of what is effectively a razor thin, quantum stabilized, RF filter/detector with microamp output. was genius. Doubly so when they used the phase of the output variation to tell which side of the peak they were on, and to then lock to the center of the range.<p>If you're on the low side, the signal goes up with frequency<p>If you're centered in the middle, the signal goes up with both sides, and you get double the swing frequency out<p>If you're on the high side, the signal goes down with frequency<p>They use this signal to then add a very small, and very slowly changing steering signal to the 5,000,000 Hz crystal oscillator... which then locks it into frequency with an accuracy of about one cycle/day. In effect they take it's error bars from +/- 10 Hz to about 0.001 Hz.<p>We connected two of them to the X and Y channels of an oscilloscope, and observed a diagonal oval that didn't drift, even after we came back from lunch, nothing had changed.<p>There are even more bits of clever in the device, but that gets into Zeeman splitting, etc.<p>Other atomic clocks, such as the Rubidium Clocks that are available surplus in E-Bay work in a similar manner, but instead of a stream of Cesium atoms, they have a cell of Rubidium gas that becomes 1% more opaque to a specific wavelength of light when excited with RF at 6,834,682,608 Hz