alt Hacker NewsFor the averaging, you need to detect some experimental signal over some time window. The stronger that signal, the less time you have to average for to get the same uncertainty. So: is a superradiant atom-only source as described in the paper as 'bright' as atoms coupled to an external resonant cavity? I'm no expert, only curious what the trade-off is. You mentioned the difficulty of getting enough power with an acceptable S/N; my question's along the same lines.
In a normal optical atomic clock, you use a laser whose frequency is determined mainly by the resonant cavity. The laser is made tunable with some additional device, e.g. a piezoelectric actuator, which can make small adjustments to the resonance frequency of the cavity.
In order to tune the resonant cavity, one generates a signal that is proportional with the frequency difference between the current frequency of the laser and the frequency of an absorption line from the spectrum of some reference atoms. There are various methods for the generation of such frequency difference signals, by modulating the laser signal and passing it through some space where the reference atoms or ions are held by various methods, e.g. an optical lattice of neutral atoms, an ion trap or just a cell with metal vapor.
By tuning the cavity, the frequency locked loop ensures that the integral of the frequency difference signal over a long integration time is approximately null, which guarantees that the average frequency of the laser is equal to the frequency of the spectral line of the reference atoms.
If instead of using a laser whose frequency is determined by the resonant cavity, you use one whose frequency is determined by the stimulated emission spectral line of the laser medium, like in the parent article, you no longer need a system of control of the laser frequency. The laser frequency is itself the reference frequency.
Unfortunately, such a laser will have a very low output power. This means that if you detect the output signal of the laser it will have a very low signal-to-noise ratio. Because of this you will still have to average the laser frequency, to filter the noise. Hopefully, for filtering the noise shorter integration times will be sufficient, in comparison with those needed for the existing optical clocks.
While shorter averaging times are a possible advantage, more important is that the frequency should be less sensitive to environmental factors, like vibrations and temperature. This could enable such optical atomic clocks to be used e.g. in vehicles. Nowadays the optical atomic clocks that can be used in vehicles are many orders of magnitude less accurate than those that are restricted to a well protected laboratory environment.