A rubidium time source is slightly misnamed: fundamentally it is an extremely stable frequency source. Your FE-5650A, for example, produces a very accurate output such as 10 MHz. Electronics can then count those cycles to make seconds, minutes and hours.
The clever part is that it uses rubidium-87 atoms as the reference that keeps an internal quartz oscillator on frequency. Rubidium-87 has a very specific hyperfine transition at about 6.834682611 GHz. This frequency is a property of the atom, rather than something determined by the dimensions of a quartz crystal.
A simplified picture is:
RUBIDIUM ATOMIC STANDARD
Rb-87 atoms
|
| atomic resonance
| ~6.834682611 GHz
v
+----------------+
Light -----------> | Rubidium vapour | ----------> Photodetector
+----------------+
^
|
Microwave signal
^
|
Frequency synthesis
^
|
+----------------+
| Quartz crystal |
| oscillator |
+----------------+
|
+-----------> 10 MHz OUTPUT
|
correction
^
|
Feedback electronics
^
|
Photodetector
The process works like this. Light optically pumps the rubidium atoms into particular energy states. A microwave signal derived from the quartz oscillator is simultaneously applied to the rubidium cell. When that microwave frequency reaches the rubidium atomic resonance, the atoms change state, which causes a small but measurable change in how much light passes through the cell. A photodetector detects that change.
The electronics continually ask, in effect:
“Is my microwave frequency exactly centred on the rubidium resonance?”
If it is slightly high or low, an error signal adjusts the quartz oscillator. It is therefore a feedback loop:
Quartz oscillator
|
v
Frequency multiplier/synthesiser
|
v
~6.834 GHz microwave
|
v
Rubidium atoms
|
v
Optical absorption
|
v
Photodetector
|
v
Error signal
|
+--------------------+
|
v
Correct quartz
This explains something particularly relevant to the behaviour you’ve seen with your FE-5650A: the 10 MHz output actually comes from a quartz oscillator. The rubidium isn’t directly generating a 10 MHz electrical signal. Instead, the atomic resonance continuously disciplines the quartz oscillator. This is why you can see an apparently good, steady 10 MHz waveform on your Rigol very soon after switching the unit on, even though the rubidium system has not yet reached thermal equilibrium or achieved atomic lock. NIST describes commercial rubidium standards in essentially this way.
Once the lamp/cell, heaters and associated electronics have stabilised, the servo finds the atomic resonance and locks the quartz oscillator to it. That’s the point at which your 10 MHz changes from merely being a good quartz-derived signal into an atomic-referenced frequency.
And converting that into an actual clock is straightforward:
10,000,000 Hz from FE-5650A
|
v
Divide by
10,000,000
|
v
1 Hz
|
v
one pulse/sec
|
v
digital counters
|
v
HH : MM : SS
There is one important distinction: rubidium gives you an extraordinarily good rate of time, but by itself doesn’t tell you what time it is. It can produce seconds of extremely consistent length, but doesn’t inherently know that it is, say, 20:24 on 15 September 2026.
That’s where your Leo Bodnar GPSDO becomes useful. GPS can provide the absolute UTC-related time/1-PPS reference, while the FE-5650A provides the excellent local atomic frequency reference.
So your planned clock can combine the two strengths:
FE-5650A RUBIDIUM
extremely stable
10 MHz
|
|
v
CLOCK ELECTRONICS
^
|
|
GPS / Leo Bodnar
UTC / 1 PPS reference
|
v
HH : MM : SS
In other words, GPS tells the clock when a second starts; rubidium gives it an exceptionally stable ruler with which to measure the seconds thereafter. That combination is a very nice basis for the atomic clock you are building.