Performance of Low-Cost Rubidium Standards

I tested five surplus rubidium frequency standards of the type that were used in cellular base stations to see what their frequency stability and phase noise looked like. The units I tested were: Important note: There's known to be a lot of variability among the thousands of each of these units out there, so my sample size of one or two isn't a guarantee that yours will perform similarly.

The results are summarized below; the raw data and plots are available here.

Frequency Stability

After allowing each unit to warm up for at least 90 minutes, I measured it for 15 hours against an HP 5065A laboratory rubidium standard that is known to perform much better than any of these devices.

This plot shows the Allan Deviation for averaging intervals (tau) from 0.01 to 20,000 seconds. With this measurement device, I tend to discount the longest tau plotted, so consider this reliable out to 10,000 seconds.

This is the same data, but showing only down to 1 second tau; it allows us to look more closely at a more useful part of the spectrum (very short tau are better suited to phase noise measurements).

From these plots, we can see a few interesting things:

Phase Noise

I measured each unit's phase noise for 15 minutes, comparing it to a Wenzel 5 MHz ultra-low-noise oscillator that is far better than any of these units.

Learnings from this plot:

Other Interesting Data

When these oscillators power on, they sweep the frequency of a crystal oscillator across a small range until the unit locks to the atomic resonance. I caught this behaviour in one of the FE-5680 units:

I didn't notice this in the other units, but on power-up the FRS units require several minutes after lock to stabilize: