At this point, atomic clocks are old news. They’ve been quietly keeping our world on schedule for decades now, and have been through several iterations with each generation gaining more accuracy. They generally all work under the same physical principle though — a radio signal stimulates a gas at a specific frequency, and the response of the gas is used to tune the frequency. This yields high accuracy and high precision — the spacing between each “tick” of an atomic clock doesn’t vary by much, and the ticks cumulatively track the time with very little drift.
All of this had [alnwlsn] thinking about whether he could make an “atomic” clock that measures actual radioactive decay, rather than relying on the hyperfine transition states of atoms. Frustratingly, most of the radioactive materials that are readily available have pretty long half-lives — on the order of decades or centuries. Trying to quantify small changes in the energy output of such a sample over the course of seconds or minutes would be impossible, so he decided to focus on the byproduct of decay — the particles being emitted.
He used a microcontroller to count clicks from a Geiger-Müller tube, and used the count to calculate elapsed time by multiplying by a calibration factor (the expected number of clicks per second). While this is wildly inaccurate in the short term (he’s actually used the same system to generate random numbers), over time it smooths out and can provide a meaningful reading. After one year of continuous operation, the counter was only off by about 26 minutes, or 4.4 seconds per day. That’s better than most mechanical wristwatches (though a traditional Rubidium atomic clock would be less than six milliseconds off, and NIST’s Strontium clock would be within 6.67×10-11 seconds).
The end result is a probabilistic radiometric timepiece that has style (he even built a clock face with hands, rather than just displaying the time on an LCD). Better yet, it’s got a status page where you can check on on how it’s running. We’ve seen quite a few atomic clocks over the years, but this one is unique and a great entry into the 2025 One Hertz Challenge.
Thanks for the writeup! If you decide to check the status page, it’s only got an uptime of a couple days; there was an extended power outage this weekend. Got to start counting all over again. It was only up for a couple weeks before that though, so doesn’t hurt quite as much as losing a months long run (which has happened)
Looking at your “one minute” video, if I saw a clock ticking that way, I wouldn’t even notice the inconsistency. I’m guessing this is because the number of counts per second is high enough that it really doesn’t vary all that much. If you used a smaller sample (or placed it farther from the Geiger tube), you’d get a lower calibration count, and more people saying, “what’s with your clock?”
Yes, that’s right. The actual rate is about 73.8 geiger clicks/second, and I’m sure there’s some statistics I don’t know off the top of my head that will tell you how much it varies from second to second. Main goal with this was to go for the long term, like weeks or months, to see what happens, and I figured my best shot at that was to have a higher click rate.
I wonder how much decay changes the rate over time and how that might impact it. Or if there could be any extra statistical magic that could help squeeze more out of the clock.
The decay rate drops according to the half life of the source, but in this case [alnwlsn] is using a lump of pitchblende which contains an unknown quantity of uranium, of different isotopes, and all their decay products.
Working out the different radio-isotopes involved, and their various proportions, and then looking up their half-lives to get an estimate of the activity would take a while. Or you can do what [alnwlsn] actually did, which is to assume that as the main isotopes have very long half-lives, that the decay rate will be essentially constant over the course of a year or so.
If you wanted this clock to operate for more than a few decades you would have to estimate the decay rate, but for measuring seconds, it’s fine.
It’s deja vu all over again, as a clock lover, that’s fine with me. Weren’t we here yesterday? Maybe a clock that uses hackaday clock posts as the tick?
I thought accuracy referred to how close a measurement is to the true or accepted value, while precision refers to how close multiple measurements are to each other.
So wouldn’t this be more precise than it is accurate?
Well compared to yesterday’s version of an atomic decay clock it was wildly unprecise, but relatively accurate.
Fun to see these two interpretations of an atomic clock btw, thanks!
I guess you could say that I aimed to be accurate over precise, but it ended up being more precise than accurate (as a wall clock, it’s a pretty lousy one, but the seconds seem to tick evenly).
While I was writing up this project, I was shocked to see another entry with the same basic idea. Not that either of us could claim that this is anything new, there’s this one from 2018:
https://hackaday.io/project/156715-gammaclock
and I found a number of historical examples here:
https://www.ahsoc.org/blog/decaying-clocks/