This is followed by a jury-rigged optointerrupter: a
fairly standard IR diode, a
matched phototransistor,
two 5 mm nylon spacers,
top half of a polypropylene beaker,
and copious amounts of hot-melt glue. The diode is connected to +5V through a 220 Ω resistor; the phototransistor
uses a 10 kΩ one, in the usual
topology. That's good enough to detect the light that gets refracted by a passing drop of water:
Next up is a clear plastic box (recycled packaging from an electronic gadget of some sort) and
ATmega48P running at 20 MHz
with an eight-line program that monitors the phototransistor, waits 80 milliseconds (chosen very empirically),
and then turns another pin on for approximately 50 microseconds (ditto):
The MCU is hooked up in a textbook fashion, the only external components being a 20 MHz crystal and a low-ESR capacitor.
The program is just this:
One of the nice things about LEDs is that they have rise and fall times of under a microsecond. That's considerably shorter than
traditional camera flash, which lights up for around 1-2 milliseconds at high power (and goes down to about 50-100 microseconds
only at lowest settings). In principle, LEDs could rival the motion-stopping powers of more exotic
air-gap flash units.
All right, so that's it, let's showcase some photos (with a macro tube attached to Canon 50/1.8). There's enough light to shoot
at f/11. Here's the moment of initial impact - I found this interesting because I don't think I have ever seen such a picture
before:
Here's are two of the more conventional takes, taken within the next 5 milliseconds or so:
And these are from earlier experiments with longer exposures (1/5,000) and fewer LEDs - still OK, but a bit
more blurry:
After about 10 miliseconds, the pretty part is over, and all you can see is a series of waves spreading
away from the site of impact.
Your lucky number: 20111906