I like to find electronic oddities on eBay and then try to make something interesting out of them. Before the internet I used a company called Greenweld Electronics for surplus and unusual electronic components. In fact – much to my surprise – they are still in business.

A few months back I came across these items on eBay:

They are ‘Rubidium 87 oscillator modules’ or atomic clocks. Used in GPS satellites, mobile phone cells, telescope positioning – in fact anywhere that you need a precise count.

It’s incredibly compact – just 3 1/2 inches square and only 3/4 inch deep!

So, I decided to build a small atomic clock from one of these surplus modules. Luckily their data sheets are readily available online and their interfaces are *reasonably* easy to work with.

The only downside is that they are at/near the end of their service life. Most still have a few years of life left so it’s not all bad news.

**Rubidium And Accuracy Comparisons**

So how accurate is a rubidium clock? To answer that you first have to ask yourself ‘what is time?’.

We all agree as to what one second is, we have to. You’d have to throw every text book on physics in the bin if we didn’t. It’s a made-up unit like the centimeter or inch. It’s just a way to keep everything *consistent*.

Consistency is the key word to understanding a clock’s accuracy. Time is a *relative* and not an *absolute *measurement.

Let’s take two pendulums called A and B. To each of these pendulums is attached a counter. After each swing the counters increase by one *unit. *We set both pendulums swinging and when one (or both) of the counters reach 100 we stop them.

Suppose that pendulum A counts 100 first and B is still at 99. Which one kept the right ‘time’?

The answer is neither! Maybe pendulum A is running fast or pendulum B running slow. All that we can derive from this is that after 100 swings both pendulums have an error of one swing or unit. We can say that our clocks are consistent to 1 part in 100 swings.

To measure the accuracy of a clock we use that delightful thing: *parts-per. *Those of you that have read electronic data sheets know what I mean!

**A Quartz Watch**

A wrist watch that contains a quartz crystal might have an accuracy of 20 *parts-per-million*. So what does that mean for its accuracy?

First we pick a ‘time frame’ over which to measure its accuracy – let’s say one day or 86400 seconds.

Now it’s a simple calculation to determine the error over that period:

(20ppm/10^6)*86400 = 1.728 seconds.

Which is 20 divided by 1 million and then multiplied by 86400.

So the maximum error is 1.73 seconds per day. That can be plus 1.73 seconds or minus 1.73 seconds and everything in between.

**Rubidium and Cesium Clocks**

The X72 in the above photo is rated at one part in a billion after about five minutes of warm up. It can approach one part in a trillion and guaranteed (when new) to attain one part in 100 billion. Old units (like mine) should easily attain one part in 10 billion – not bad!

So applying our above formula we get:

(1/10^11)*86400 = 0.000000864 seconds when new, or 864 nanoseconds per day!

(a well calibrated rubidium oscillator can achieve 8.64 nanoseconds per day)

A cesium beam clock achieves 1/10^13 or one part-per-ten-trillion and a cesium fountain clock 1/10^15 or one part-per-quadrillion.

The NIST-F1 (https://en.wikipedia.org/wiki/NIST-F1) is accurate to 3.1/10^16.

The Universe is about fourteen billion years old. If the NIST-F1 were to have started ‘ticking’ at the moment of the Big-Bang it would now neither have gained nor lost more than 2 minutes and 17 seconds.

A quantum clock has achieved an accuracy of 8.6/10^18.

In part two we’ll have a look inside the X72 to see how it works.

© Dr Mike Finnley 2018

Audio file