Dimming HID lamps

A surprising number of HID lamps can be dimmed when using a suitable driver. The Philips HID-DV Dynavisions are excellent in that regard, they can run pretty much anything provided the lamp voltage is not too low. The power level is adjusted by a 0-10 V input which enables a very simple and easy use of the system. I made a little control box containing a switch and 10-turn precision potentiometer, which can be seen at the bottom, near the multimeter I use to monitor the set voltage level. I opted for BNC connectors out of convenience, also to connect that control box to other instruments and circuits if needed (i.e., such as a parallel variable capacitor if I need to adjust the rate of power change over time).

Here I am running an Iwasaki Eye MT150FD at around 100 W, which is about 67 % of the lamp's nominal wattage. The light color is still excellent and it is possible to go even lower in power, but the CRI will take a hit (although the light color shifts towards a nice clear mercury-like tone, only with a much better color quality). I've tested a wide variety of HID lamps with this setup and only a few were found to be unsuitable in actual usage, such as the color /830 standard ceramic metal halide lamps which turn green at low power due to a too high thallium content and a burner temperature which drops too quickly.

Unsurprisingly, mercury lamps are the easiest to dim and they are also the most stable in light color. I ran the Philips HPL-N 50W shown here on the 70 W ballast, which enables a (useful) power boost as well as dimming (that lamp has a broken resistor and won't start without a high-voltage kick). Now I'm wondering what toll the extra 23 W has on the lamp life... If operated continuously at 73 W, calculations give a 72 % reduction of the mean service life, which is quite severe. On the other hand, the lowest measured power setting with that lamp is 30 W, which corresponds to a 3-times greater (calculated) service life. In theory, there is a way to combine the boost and dim timings so the net total service life is unchanged. A simple calculation gives an interesting solution: it's 73.8 % of the time run at 73 W and 26.2 % at 30 W. The estimated benefit of an operation at low power is such that we could overdrive the lamp by 46 % for nearly three-quarter of the time without impacting the mean service life (16 kh in the present case)... Only a proper field test will tell if that's actually the case - but that's a very interesting outcome nonetheless.