Dice Rolls From The Beginning Of Time

Generating random numbers might seem like a trivial task, that is until the numbers need to be truly random for cryptography or security reasons. When that’s the case, it turns out that these numbers are really “pseudo-random” and follow a predictable pattern. Devices that can produce truly random numbers often do it by sampling random events in the real world rather than relying on a computer to do it directly, like this machine which simulates a dice roll by looking at the cosmic microwave background radiation.

The cosmic microwave background radiation exists in the infrared at the farthest edges of the observable universe as a remnant of the big bang. It’s an excellent source of randomness, but tapping into it poses a bit of a challenge. For this build, [iSax] is using an old Soviet-era Geiger tube to detect the appropriate signal, and a Nixie tube to display the dice roll. After the device detects two particles from the Big Bang, the device measures the amount of time that passed between the detection of both particles and uses this number to calculate the dice roll.

While it takes a little bit longer to roll this dice than a traditional one since it has to wait to detect the right kind of particles, if you really need the randomness it can’t be beat. It certainly works as dice, but we can also see some use for generating truly random numbers for other applications as well. For some other sources of random inspiration be sure to check out our own [Voja Antonic]’s deep dive into truly random number generation.

30 thoughts on “Dice Rolls From The Beginning Of Time

  1. > cosmic microwave background as defined:
    > the peak wavelength is 1.063 mm
    >

    Small nitpick. This machine does not use the CMB. As the energy is not enough to excite a Geiger counter tube. This device measures ionizing radiation. Microwaves are not ionizing. They create rotation in molecules, not changes in charge.

    CMB is electromagnetic radiation in a very low energy spectrum. It has nothing to do with gamma rays, alpha and beta particles or muons.

    /rant over

    Other than that it is a very nice project. Just not using physics as advertised here.

    1. Yeah, I was thinking pretty much the same thing. I clicked the link, checked out the project page and watched the video hoping he would describe why it’s the CMB and not just regular background radiation that he thinks he is sampling. I’ve got nothing.

    2. Also, if it was the CMB he was measuring wouldn’t that make it not random? Isn’t the fact that it is uniform across time and space the whole reason we can know that the CMB is a real thing and not just noise?

      1. Of course it’s noise. It’s thermal noise.

        When we say it’s uniform that’s like saying every radioactive atom is uniform. Every one you measure acts the same. Every photon you receive from the CMB is the same, but it’s still random.

    3. Yeah, a misinterpretation or loss in translation.
      It’s background radiation, and it’s cosmic rays. But it’s not remotely related to microwaves.

      More accurately: depending on your altitude and local geology, about half is from cosmic rays, about half is from local terrestrial sources.

    4. Ok, last response I promise. Just wanted to say it’s a nice build even if we don’t all agree on exactly how it is working.

      Also… it would be pretty cool if one of those watches in the background had radium in it. One might also consider bringing a close by and seeing what changes.

    5. Right, I mean there’s an oxymoron right here:

      “The cosmic microwave background radiation exists in the infrared”

      Microwave isn’t infrared, and neither is detected by a Geiger counter. As others have said, what you’re really picking up is a mix of terrestrial radioactive sources and cosmic rays, neither of which has anything to do with the Big Bang. In fact the particles it detects due to cosmic rays generally are produced in Earth’s atmosphere from an incoming high energy particle (which wasn’t produced in the Big Bang either).

    6. Yup, not CMB. I’ve got the same tube and it’s just picking up general radiation – granite, bananas, vial of pitchblende when I need numbers faster, etc.

      I had the same idea to use nixies, so mine is now in pieces waiting for me to build a new case for it… :(

  2. It’s true to say that radiation (half-life decay emissions such as alpha, beta or gamma particles) are the only truly random events. Anyone using random numbers made solely by computer calculation are not random.

    1. Not at all true. Any Brownian motion or thermal noise source is also random noise. The trick is how to generate random numbers from it without colouring it (biasing the distribution). The unambiguous discrete signals from radioactive decay events make that part (almost) trivial. The OP’s naïve strategy of selecting evens or odds from a fast counter almost makes it.

      1. My immediate thought was that there is no guarantee the timer he’s using doesn’t have some bias on even vs odd measurements.

        What would be nice is if he fed data from it through one of the randomization testing suites.

      1. A gamma particle (or gamma ray photon) is a photon with an energy greater than visible light. Distinguished from an X ray photon that might be of similar energies by virtue of originating in the nucleus of an atom, not from electrons.

        Yes, a photon is an elementary particle. It carries energy and has momentum, and has spin and angular momentum. A photon of sufficient energy can decompose into an electron and a positron when it strikes another particle. This is actually the main mechanism of absorption of high energy gammas.

  3. “While it takes a little bit longer to roll this dice than a traditional one since it has to wait to detect the right kind of particles…”

    The trick for using this method in a production setting is to use the slowly generated random numbers to keep re-seeding a faster pseudo-random number generator.

  4. An easier and faster way to do this is to sample radio background noise and turn it into a bitstream. It might be possible to affect it by “creating noise” however the fact that you don’t know what frequency or at what time I am looking at the noise makes it fairly securely random. Often a psuedorandom generator is used to decide when to sample the noise stream making it another level closer to really random.

    1. So you are saying that your random number generator might not actually be random at all from the perspective of someone who is actually measuring the entropy of your noise stream.

      1. Randomness does not automatically mean exclusiveness, a data stream might be 100% random in that there is no way at all to predict if the next bit will be a one or a zero, but if i replicate your design i could afterwards predict with a certain chance whether you got the same bit value that i got. For cryptography exclusive access of the random data is also essential.

  5. Senior project I did was random number generator using brownian motion random walk. We created a cell which contained a small particles suspended in a liquid and pointed a modified USB microscope at it. Whatever particle happened to be in view, you’d measure the direction of motion in real time and use that as the seed. What limits all of these type of generators is the range of random numbers possible. To be truly random, you need an infinite range, but that’s not physically possible. But they are usually random enough for the intended purpose.

    1. You have an incomplete understanding of the problem of randomness.

      If you truly had a random number generator that is guaranteed to produce a uniform distribution of numbers with infinite range, then it would always report infinity.

      No! That’s impossible! I hear you insist. OK. Divide the number space in half. Half the numbers are between zero and half of infinity. How much is half of infinity? Answer: still infinity. Any random fraction of infinity is still infinity.

      Now, if you specify you want a random int16, or random bit, or a random float between zero and 1, or even a random float with a gaussian distribution of a particular width and mean, then that’s beginning to demonstrate some understanding of the question.

  6. I’m been skeptical of fancy schemes that measure phenomena to extract randomness, ever since the CMU lava lamp discovery. As legend has it (I can’t find a credible source to cite) they had a web cam watching a lava lamp. Stuck the assembly in a closet, and happily used its randomness for months. Eventually someone looked in the closet and found that the sole illumination had burned out. They were using the dark noise in the sensor. Never needed the lava lamp at all. It “works” about as well to leave on the lens cap. Acknowledgment: didn’t watch the “CMB” video.

  7. Congratulations on a wonderful use of technology to create TRNG! I have seen other inventions using Tritium as a source of entropy from a startup based in Poland.

    “While it takes a little bit longer to roll this dice than a traditional one since it has to wait to detect the right kind of particles, if you really need the randomness it can’t be beat.”

    As for the statement that it can be beat, I must challenge you with a rack mount system designed for data centers that creates pure entropy from a physical phenomenon, rolling real dice in a controlled environment. https://youtu.be/XI1NVYvDBlk

Leave a Reply to ljCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.