A better Geiger-Müller mood lamp

Hey, you! Yes, you. Ignore this page and check out the newer version of this design instead!
Few weeks ago, John Iovine (of Images SI) published a cool idea for a Geiger-Müller mood lamp - an accent light that changes color in response to background radiation events, and also conveniently warns you the moment the inevitable nuclear apocalypse begins: A doomsday sensor and a fashionable lamp in one! It seemed interesting and gruesome enough for me to want to build it right away. Upon closer inspection, I had some reservations about John's design, though:
  1. It relied on an unidentified transformer and a mystery GM tube that could only be purchased from the author, for about $60, seemingly with no specs included. I tracked down an identically marked transformer at Electronic Goldmine for under $3; the tube remains a mystery - it could be Russian SBM-10 or SBM-21 (link), but it's hard to tell for sure.

  2. In the video, it looked like a pretty poor mood lamp: the appeal of the enclosure itself aside, it just blinks harshly as it cycles through four primary colors, rather than using nice full-palette fades.

  3. The circuit seemed a bit suboptimal: for example, it used a 16 MHz, externally clocked microcontroller and a signal stretcher to implement what could be done with a much cheaper 74HC series counter and a flip-flop, and featured a fairly inefficient inverter circuit clamped with diodes, requiring a MOSFET driver.
So, here's my alternative take on the project - simpler, cooler, more affordable, and hopefully better documented.

Getting a GM tube

This step turned out to be trickier than expected. There are some US-based manufactuers of GM tubes, for example LND Inc - but you must be prepared to shell out around $100. Barring this, most of the reasonably-priced tubes available to hobbyists today are Cold War era Civil Defense surplus, plus surplus imports from former USSR countries. Oddly enough, eBay is the be the best place to go: you can buy excellent tubes for about $15-$20, and many of them are available in continuous stock. Hobbyist stores, such as Sparkfun, Images SI, or Electronic Goldmine seem to be carrying the same items at a 200-400% premium, but with a promise of faster shipping.

Wherever you shop, there are the three characteristics you should be looking at:

There are other differences between tubes, including alpha sensitivity, recovery time, operating temperature, etc - but for this particular use, they do not matter much. I ended up buying SBM-20, one of the most popular tubes, for about $20 + $5 shipping. This tube registers about one event every 3-5 seconds, compared to 30-60 seconds for John's version.

Inverter circuit

WARNING: This portion of the circuit makes use of dangerous voltages. Though the current is very low, you can get shocked - so exercise caution.

Electrically, the tube is a glorified picofarad capacitor: when supplied with a DC voltage on its terminals, it charges up almost immediately, and draws virtualy no current from that point on. When the potential difference is sufficiently high, and the tube is struck by high-energy particle / ray, plasma may form inside, and the tube may spontaneously discharge - resulting in a momentary 80-90% drop of terminal voltage that typically lasts somewhere between 100 and 500 microseconds:

To power the tube, I started with a switched-mode wall wart, capable of delivering somewhere around 500 mA at 5-7V DC (about 100 mA to power the tube, everything else for LEDs). You can almost certainly salvage one from an old cell phone or other obsolete piece of electronics; if you really do not have anything lying around, you can buy one for about $6. To covert this to the high voltage needed by the GM tube, I relied on a simple 555-based oscillator configured to generate an 90 kHz AC signal ($0.50 total). I then fed the signal to a small, high-frequency transformer to form a ghetto inverter.

The 555 chip can source enough current to power some transformers directly, but has another major flaw: its maximum output level is capped at about 2.3V below the supply voltage. Since voltage reduction is exactly opposite to the desired effect, I used a cheap, medium-power NPN transistor (PN2222A, $0.05), as a proxy - this way, voltage drop stays around 0.3V:

The 10 uF capacitor is there to counteract the tendency for 555 to momentarily short-circuit the supply on state transitions (also avoidable by using ICM7555 instead), while the frequency-matched 100 nF cap minimizes harmonics / flyback effects caused by feeding the transformer an unsightly square wave. Lastly, the trim pot allows you to fine-tune the voltage on the other side of the transformer by varying signal amplitude. You can skip all these elements, and clamp the output voltage with a suitable diode instead (e.g., 1N4004 for 400V), but that's just rude and wasteful.

The only missing piece is the transformer. I used Coilcraft FL2015 series ($2.50), but any make of miniature, high-frequency step-up transformers for fluorescent lights will do (e.g., Cooper or JW Miller). These transformers usually come with winding ratios between 1:50 and 1:150, and work well with the 90 kHz signal range. To pick the right variant, divide the recommended GM tube voltage by supply voltage minus about 1 volt; then find a transformer with a winding ratio closest to this number (rounding up). For example, with a 6V supply and a 400V tube, a 1:80 to 1:100 winding ratio should be OK (you do not have to be very precise).

Do not buy bulky, low-frequency transformers meant for 50/60 Hz signals (e.g., Tamura SB2812 series), or extremely low power, current-sensing transformers (e.g., Murata 5300 series). Neither of these options will perform as expected in this circuit.

Note: if you bought a 1000V tube, you might find it difficult to find a transformer with a suitable winding ratio. In such a case, you can go for half the required value - and add a simple diode-based voltage doubler later on, instead.

The high-voltage side

On the other side of the transformer, I used two ceramic capacitors and two 1000V 1N4007 diodes ($0.20 total) to restore something resembling a DC signal:

The tube itself is connected to this DC supply through a 4.7M resistor. This is to prevent undesirably high currents from flowing through the plasma and overheating or otherwise damaging the tube; the exact value will be recommended by the manufacturer, but somewhere around 5 megs works in most cases. Furthermore, the cathode of the tube is grounded through a small resistor (scale it proportionally with the other). Normally, the potential across this element will be zero volts - but when the tube starts conducting, a difference of about 4-10V will be momentarily created, just enough to feed to a CMOS chip:

(Red trace is the potential across GM tube terminals while a discharge occurs, 400V in its peak; blue trace is the potential across the 1k resistor, 5V in its peak.)

Before connecting the tube, be sure to dial the trim pot in the inverter circuit to maximum resistance. Attach an oscilloscope probe across the 200 pF capacitor, and slowly turn the potentiometer back until the manufacturer-recommended output voltage is seen. A voltmeter might not do - some of them have an insufficient internal resistance to accurately measure this voltage. If you don't have a scope, do not worry: you can adjust the potentiometer once the entire device is completed, turning it until the tube starts registering events with the desired frequency - and then few degrees more. GM tubes have a wide operating plateau, usually spanning 50-100V or so, so it's hard to get it wrong.

Color picker and LED drivers

The lamp should change colors whenever a radiation event is detected by the GM tube: the intervals between color changes will be unpredictable, and the pace of changes correspond to the levels of radioactivity. So far, so good - but the sequence of colors should be random too, partly for aesthetic reasons, and partly - as suggested by John - to put other people's amazing precognition skills to an ultimate test.

So, random colors it is! I grabbed the 90 kHz signal already available from the 555 chip, and fed it to 74HC393, a simple 4-bit counter chip ($0.30). This setup causes the counter to cycle through all the 16 possible output states over 5000 times every second. The chip is then connected to a flip-flop register, such as 74HC175 ($0.30). This register, in turn, is timed by a much slower, asynchronous signal: edge of the pulses received from the GM tube itself. It will load and hold the current state of the counter at random intervals - just what I need.

Now, 74HC175 can't, by itself, drive significant loads - so I added a medium-power Darlington driver, ULN2003A ($0.30), in order to control four banks of LEDs and make this circuit a proper lamp; a couple of capacitors and resistors in between the two chips ensure smooth and responsive color fades when a bank is turned on or off:

I used sixteen cheap 6000 mcd LEDs organized into four banks (4 * blue, 4 * green, 4 * red, and 4 * white - about $5 total). LED current needs to be limited to a manufacturer-recommended value; it's usually 20-40 mA, so a 47-68 ohm 1/4 W resistor per bank in intermittent duty should do. For high-power LEDs, power resistors or a proper current regulator might be necessary.

Hint: if your tube is too sensitive for your taste, and color changes occur a bit too frequently - simply reroute the GM tube signal from 74HC175 to the second, currently unused counter on the 74HC393 chip (pin 13); wire the counter's pin 12 (reset) to Vcc; and connect pin 10, 9, or 8 back to 74HC175. This will divide the click frequency by a factor of 2, 4, or 8, respectively.

Making the enclosure

The assembled circuit looked something like this:

External dimensions are around 6.5 x 9 cm. By itself, it looks ugly. I experimented with several fancy enclosure designs, but eventually settled on a stylish but very simple project box. I machined a quick two-part positive mold master on my CNC mill in RenShape 5440 prototyping board (this took about 2 hours), then poured ShinEtsu KE 1310ST silicone over this master to create flexible, negative molds:

In these silicone molds, I cast final, rigid polyurethane parts using Innovative Polymers IE-3075 (project box, dyed black) and OC-7086 (oversized window, with a shimmering pigment to diffuse light neatly):

The box is painted with an automotive clear coat mixed with a gunmetal, pearlescent pigment (the photo does not do it justice - it's mostly black). Four small, orange-colored rubber legs were also cast and affixed to the bottom.

UPDATE (Nov 2010): a newer, more lamp-like case is shown here:

Questions? Comments?

You can reach me at lcamtuf@coredump.cx.

Your lucky number: [an error occurred while processing this directive]