I earlier did simple go/no-go testing for a bunch of the LED boards that I designed for lighting fixtures (see Summer project), but I thought it would be interesting to characterize at least one of the boards more thoroughly. So last night I wired up a little circuit to control the current through board to plot an I-vs-V curve:
I measured the voltage across the board and the current through it with multimeters (using the better multimeter for the current measurement). I was expecting a constant current when the voltage was high enough, then a linear decrease in current as the voltage was lowered down to a cutoff threshold. That is pretty much what I got:
The constant-current region was not as crisply defined as I had expected, and the current was lower than I had intended. The 10.6Ω impedance in the linear region initially came as a bit of surprise, but when I added together the resistance of my current sense resistor and the dynamic resistance of the LED in the relevant region it seems about right.
The measurements were hard to make, because the current did not remain constant, but tended to drop as I was measuring, particularly in the high-current regime. I believe that this droop is due to thermal effects—the current drops as the board warms up, and I did not wait for the board to reach equilibrium temperature. The lower-than-expected constant current is probably also due to thermal effects, since it was based on scaling up tests done at low currents, where there would have been no significant heating.
To test this hypothesis, I set up a different experiment this morning, connecting the board in series with a 20Ω resistor and connecting both to a 12V power supply, monitoring the voltage across the resistor (and hence the current through the board) using PteroDAQ on the FRDM KL25Z board.
So my “constant-current” circuit isn’t really constant current—it is very temperature dependent. The change in brightness is about the same as I would get from a 13° change in the position of the control potentiometer for my dimmer. I can live with that in the design, but it is a much bigger temperature dependence than I had expected.
According to my infrared thermometer, the heatsink got up to about 60°C at the end of the run, with 9.75V across the board and 0.116A through it, for a power dissipation of about 1.13W. If the room was at about 20°C, that means a temperature gain of about 35.4°C/W.
The LEDs get derated to about 93% of their room-temperature efficiency at 60°C, so when combined with the current drop to 116mA, I expect about 75 lumens for each board when it is fully on. Maximum efficiency would be at the knee of the I-vs-V plot, where the voltage is about 6.67V, getting 75 lumens for 0.77W, or 97 lumen/W. (The temperature may not get as high at that voltage, since the power dissipation is less—the terminal temperature would probably be only about 48°C, which means the current would drop less and efficiency would be slightly higher.) At my design voltage of 9V, the efficiency is only about 72 lumens/W. The LED boards seem to be able to run at 12V, where the power dissipation would be 1.4W/board and the efficiency only 54 lumens/W.
One surprise for me in the testing was how low a current would still produce light. I observed dim light down to about 4µA, giving the LED boards a dynamic range of about 32000 in brightness (4.5 decades), while my PWM circuit only has a dynamic range of about 43 (1.6 decades). The range on the dimmer is adequate, since even a night light produces about 15 lumens, and my dimmer goes down to about 2 lumens. But it is clear that one could design for a much wider dynamic range.
Redoing the I-vs-V plot to have a logarithmic scale for the current, we can see the whole dynamic range.
The LED boards could be operated at as low as 5V, but the brightness is very low at that voltage (about 0.15 lumens)—not suitable for a room light or even a night light. A 7.4V Li-ion battery pack would be a good match to the LED boards. A $14, 2200mAh battery should be able to power the LED board at full brightness for about 10 hours, and at reduced brightness for another 8–10 hours. I’m not currently planning any battery-operated lights, but it is nice to know that they are doable.