I am trying to understand cesium atomic clocks better.

I am not getting HOW the cesium oscillation is actually being counted.

So from my understanding of an older atomic clock:

  1. Cesium gets heated.

  2. Heated cesium goes through magnets.

  3. Excited cesium atoms get discarded and non-excited cesium atoms get hit with microwave radiation at about $9,192,631,770 \, \mathrm{Hz} .$

  4. This excites electrons to the hyperfine splitting upper energy level of the $6\text{s}$ orbital.

  5. Electrons that get excited hit a detector, and the ratio between these and the total number entering the microwave chamber is calculated, subsequently the frequency of the microwave generator is adjusted to reach the highest ratio. To this point to me all we did was measure that cesium has a two energy levels separated by some energy,$$ {E}_{\text{photon}} ~=~ h f ~=~ h \times {9,192,631,770} \, \mathrm{Hz} \,,$$where $h$ is Planck's constant and $f$ is a photon's frequency.

So great – but how is this actually being counted? Is there a device somewhere that says – ah ok, maximum cesium atoms are being excited, let's start counting... start counting what though?!? And what device is doing the actual counting, as in: $1,$ $2,$ $3,$ $4,$ ..., $9,192,631,770,$ and then saying, BOOM, that's a second.

  • 1
    $\begingroup$ Not sure about the latest state of the art technology, but the standard way of doing this sort of thing is to use your good old oscillator that isn't all that accurate for the clock and to use the adjusted frequency of the microwave generator as a feedback signal to adjust the frequency of the oscillator. $\endgroup$ Apr 5, 2016 at 22:39

2 Answers 2


Nowadays it's reasonable to make an actual digital counter that operates at 10 GHz. Even so, you might rather use a prescaler to divide this frequency down to 1/4 or 1/8 of the microwave frequency before you actually start counting cycles.

In olden times you might have generated your 10 GHz signal by multiplying up a lower frequency (maybe in the 100's of MHz?) and you'd be tuning that low frequency to adjust the microwave frequency. Then you would count the cycles of the lower frequency to measure off a desired time interval. (Note I don't know how early atomic clocks were designed, this is just one practical way of doing it)


The July, 1968 Hewlett-Packard Journal (on-line here) gives this block diagram for how a rubidium frequency standard was built at that time:

enter image description here

So my guess wasn't too far off.

As far as I know, commercial atomic clocks typically give a 10 MHz clock as output, and they never expose their 9.19 GHz clock (or 6.83 GHz in the case of rubidium) to the user.


The counting is done by the signal generator.

The clock contains a signal generator that generates the 9,192,631,770 Hz signal going to the microwave transmitter. This generator contains a counter that counts every cycle the generator creates (though as The Photon says it may use a prescaler).

  • $\begingroup$ But this seems circular -- how does the signal generator know what 9,192,631,770 Hz is, if it doesn't know what a second is yet. If it knew then what would be the point of the cesium thing in the first place. $\endgroup$
    – sci-guy
    Apr 6, 2016 at 16:51
  • $\begingroup$ @renegade: the signal generator doesn't know what frequency it is, it just knows that it's frequency gives the maximum absorption by the caesium vapour. It counts its cycles, and our interpretation is that when the signal generator is tuned to maximum absorption 9,192,631,770 of its cycles makes one second. $\endgroup$ Apr 6, 2016 at 17:07
  • $\begingroup$ Final question, so what exactly is cycling in the signal generator - like what is it counting? $\endgroup$
    – sci-guy
    Apr 6, 2016 at 17:29
  • $\begingroup$ @renegade: the signal generator generates an oscillating voltage. The counting is just counting how many times the voltage oscillates through one complete cycle. $\endgroup$ Apr 6, 2016 at 19:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.