This document describes the original 1996 first-generation HotBits hardware. The current third-generation HotBits generator uses a commercial Geiger-Müller detector designed for attachment to a computer's serial port, illuminated by a 5 microcurie Cæsium-137 check source. This document is for historical reference only; please see the third-generation HotBits hardware description if you are interested in building your own generator.
The HotBits hardware uses off-the-shelf items with only minor modifications and additional components. By interfacing with the computer's serial port and making all the measurements in software, all the complexity of my original 1986 design is eliminated. To conserve network bandwidth and reduce download time, the colour figures in this document are small images in JPEG form. If your browser cannot display JPEG images, or you want to examine an image in more detail, click on it to display a larger GIF image. After viewing the larger image, your browser's “Back” button will return you to this document.
To detect radioactive decay events, I use a commercial radiation monitor which contains a Geiger-Müller tube detector. Since we're interested in timing individual events, a detector of this kind is essential; many radiation monitors use an ionisation chamber which measures the average level of radiation but does not permit timing individual particles.
The detector I used is called “Monitor 4”, and is manufactured by SEI of Summertown TN 38483, USA. A do-it-yourself kit version of this detector used to be sold by Heathkit, back when such things still existed, but today only assembled versions are available. Here's the monitor as listed in the Edmund Scientific catalogue; you can order one from them (101 E. Gloucester Pike, Barrington NJ 08007-1380, USA; Tel. +1 609 573-6250, Fax: +1 609 573-6295). The price may have changed since the 1996 catalogue I referred to when writing this document, so be sure to call and confirm price and availability.
The first thing any self-respecting engineer does upon receiving a
new gizmo is, of course, tear it apart to see what's inside. Peeking
inside the brown box, we see the following. (This picture is of my modified
monitor—the jack and the two wires aren't there in the factory model.
I'll explain how to add the jack below.)
All the electronic components are mounted on the single printed circuit board. At the top left, protruding beneath the meter is the Geiger tube detector, which has a window at the end to allow alpha and beta particles to enter. When the unit is assembled, the detector looks through a hole in the case, with the fragile window protected by a wire mesh. The round blue thingamajig at the lower left is a beeper that makes the audible “tick” for each count in “Audio” mode. The meter is driven by an integrator circuit which sums the individual count pulses to compute the average level of radiation.
What we're interested in is the individual pulses, so we'll need to
modify the meter to bring out the pulse signals to a jack on the outside.
This turns out to be very easy to do, since there's a test point on
the circuit board where these pulses are available as a 9 volt
CMOS signal. The point we want to connect to happens to be below
the beeper, so remove the screws and set the beeper aside for a
moment. Here's a close-up with the beeper removed (the two red
wires at the bottom left go to the beeper, which is off the bottom of
You can see that the wire which goes to the jack's tip connector goes to the top of resistor R1A which, as a test point, conveniently stands up, making it easy to attach the wire. Here's an additional blow-up that shows the jack tip wire soldered to the top end of R1A. If you're paranoid about blowing out parts in this expensive gadget, put a 1 Megohm resistor between the top of R1A and the jack tip—that will limit the current an accident might dump into the meter and since the interface is CMOS to CMOS, won't impede operation. I didn't bother with this refinement, as you can see from the pictures. I've found the Monitor 4 to be quite rugged at withstanding the odd “oops” event in the breadboarding stage.
The jack's ring terminal goes to the signal ground. The easiest place to attach it is right where the black lead from the battery clip is soldered to the board. Flip the board over and attach the ring wire at the lower left, as shown below. From the bottom, you get a clear view of the Geiger tube. Be careful when manipulating the disassembled monitor not to jar the tube—when the circuit board is removed from the case, it is supported only by its electrical leads at the bottom end.
Now all that's left is adding the jack to the case and hooking up the wires. Choose the smallest jack you can find, and be careful to position it on the top of the case so neither it nor the matching plug, when inserted, will touch any of the components inside. Finally, re-install the beeper on its mounting posts, replace the circuit board in the case (be careful to seat the Geiger tube in its support at the top of the case, and make sure the protective screen hasn't fallen out of the window), and button everything up. Then make sure the monitor still works. Here's my modified Monitor 4 with the count pulse jack visible on the right side of the case.
Now that the detector is done, all we need is an interface so it can be hooked up to the computer. If you have a digital I/O board that accepts 9 volt input signals and has reed-relay output, you can just hook up the modified Monitor 4 directly to it, but most people don't have such fancy hardware. The HotBits computer interface attaches, instead, to an IBM PC-compatible serial (COMn) port. I could have used a parallel (printer port) just as well, but most computers only have one built-in printer port and it is often already connected to a printer. The schematic for the interface is as follows.
Fourmilab has RJ45 Category 5 wiring throughout, so I decided to make the cable from the interface to the computer RJ45 with an adaptor at the computer end for the DB-9 serial port connector. The connections to the serial port at the left of the schematic use the standard RJ45 wire colour codes. The entire interface fits snugly inside a standard RJ45 wall mount jack box, making for a neat and tidy package.
One advantage of using RJ45 cable is that the generator doesn't have
to be anywhere near the computer. In fact, it's located three floors
down in a converted 70,000 litre subterranean water cistern with metre-thick
concrete walls and patched to a jack next to the computer at the main
network patch panel. No need to worry about stray radiation zipping
around the computer room! The adaptor on the computer end is a
commercial RJ45 to DB9 female converter, wired like this:
Turning back to the schematic of the interface, note first that only the serial port's modem status and control lines are used—the actual serial data input and output are not used. To power up the interface, the computer raises the Request to Send (RTS) line to positive voltage. This charges the capacitors which power the the integrated circuit U1. The diode is necessary because RS-232 levels swing between +12 and −12 volts, so we have to prevent a negative voltage from being applied to the tantalum capacitor and CMOS hex inverter, which could destroy them.
The serial port's signal ground is connected to the signal ground of the Monitor 4 detector by the ring of the jack. The tip of the jack connects the pulses indicating individual radioactive decay events to the input of the inverter. The 1.2 Megohm resistor limits the current flowing through the inverter's protection diodes if the pulse from the Monitor 4 exceeds the supply rail voltage in the interface (as the two are powered independently, this is entirely possible). The output of the inverter drives the Clear to Send (CTS) modem status line, which appears as a bit in the serial port's modem status register. The function of the inverter is solely to buffer the signal from the Monitor 4 and provide sufficient current to drive the low-impedance RS-232 port—the circuit into which we tapped cannot drive the port directly. (Why did I use an inverter instead of a non-inverting buffer? Because it's easy enough to compensate for the pulse inversion in the driver software, and an inverter was all I could find in the junk box!)
Geiger tube detectors exhibit a phenomenon known as saturation which reduces their sensitivity when radiation exceeds a given level. Essentially, the gas in the tube becomes ionised in bulk, begins to conduct continuously, and is thus less sensitive to individual particles. To avoid this, we need to be able to power down the detector periodically to “give it a rest”. To accomplish this, the Data Terminal Ready (DTR) modem control line is connected (through a diode, so it doesn't conduct when driven negative) to a reed relay with a 500 Ohm coil impedance. The reed relay is inserted in the positive power lead feeding the detector, which permits the software to power the detector on and off at will.
Radiation is ubiquitous in our fair universe so, strictly speaking, there's no need for a radiation source—we could just rely on background radiation, primarily due to cosmic rays and radioactive decay of natural thorium and uranium in the Earth's crust. Background radiation increases with altitude since the higher you go, the less the atmosphere shields you from cosmic rays. At sea level, cosmic rays account for only about 30 millirems per year, but at our altitude of 800 metres in the Jura mountains of Switzerland, the background radiation is about 165 millirems per year. By comparison, an X-ray is about 50 millirems.
Although background radiation can be used, you either have to not need very many random bits or else be very patient, since three counts are required to generate each bit and background radiation counts only occur every few seconds.
This is your detector on background radiation.
To crank up the bit generation rate to something usable for
a server accessible on the Internet, we need a radiation source
more intense than background radiation. Rummaging around in the
well-endowed Fourmilab junk box turned up a 60 microcurie Jordan Nuclear
Krypton-85 (85Kr) source capsule, model BB-0005.
The capsule is about two centimetres long, and has a foil window on
the left side through which the radiation emerges. This capsule,
aimed at the window of the Monitor 4's Geiger tube, was just what
the doctor ordered to start cranking out HotBits at a decent
This is your detector on Krypton-85.
The source capsule generates radiation through the beta decay of Krypton-85:
where the beta particle (an electron, just like those illuminating the computer screen on which you're reading this document, only moving faster) is emitted with an energy of 687 kiloelectron volts (keV) and the gamma ray photon with an energy of 514 keV. Krypton-85 has a half-life of 10.73 years, so the radiation from the capsule is a little more than half as intense as when it was manufactured in 1988. The radiation level impinging on the Geiger tube of the detector is about 50 millirems per hour, as measured by the radiation monitor.
I used the Krypton-85 capsule because I happened to have one lying around. If you're building your own HotBits generator, a better choice would be a 10 microcurie Cesium-137 (137Cs) check source, which usually takes the form of a small plastic disc containing the radioactive material. Cesium-137 has a half-life of 30.3 years, so your investment doesn't waste away quite so quickly, and the lower radiation level (Cesium-137 is also a beta-emitter) is both safer and completely adequate for triggering the detector. In most locations, no license is required to obtain such a check source, which can be ordered through the mail from dealers in such gadgets.
If you can't lay your hands on a purpose-built radiation source, a couple of other alternatives are available. Mantles for gasoline (petrol) and some kerosene (paraffin) lamps are coated with Thorium oxide. There is no stable isotope of Thorium; the only isotope found in nature, Thorium-232, decays by alpha emission with a half-life of 14,000 million years, so there's still plenty left over from the supernova that made the heavy atoms in the Sun, the Earth, and you and me. Obviously, with such a long half-life, you're not going to get intense radiation from a Thorium-coated mantle, but it's still a lot better than background radiation for speeding up bit generation. Be careful when handling the mantle—follow the instructions that came with it, wear gloves, and don't do anything that will make dust you might inhale or swallow. You may be able to simply leave the mantle in its original package, but since alpha radiation is not very penetrating, this might not work.
Another alternative is to visit a shop catering to rock collectors and buy a specimen of a Uranium-bearing ore such as Carnotite or Pitchblende. Natural Uranium has a half-life of 4,500 million years and, like Thorium, decays by alpha particle emission.
If you don't know what you're doing, radioactive material of any kind is better left strictly alone. I am not encouraging you to fool around with hot stuff—especially when there's no need to, since you can get all the random bits you need from the Fourmilab generator.
Before assembling an elegantly-packaged final product, it's nice to know if the darn thing is gonna, you know,… work. Here's a shot of the HotBits prototype, a few seconds after it started working the first time. This was cobbled together from junk box parts and Barney clips in a few minutes. The radiation source used to test the prototype was the check source built into an ionisation chamber radiation meter—that's the white box at the top with the radiation symbol on it. The actual check source is not visible in this picture.
Here's the production model, hooked up to the detector and a battery which provides the switched power to the detector. (In practice, I use a 9 Volt “wall-wart” power supply to avoid worrying about the battery running down.) The RJ45 cable to the computer serial port plugs in at the bottom of the interface box. All the interface components are inside:
The reed relay that controls power to the detector is at the left, and the small circuit board at right contains the rest of the components.