I bought a few of these high brightness RGB LED’s from my favorite sketchy site, DealExtreme:
They’ll probably arrive in a few weeks or so. But in the meantime, how do you drive these 350mA LED’s? Sparkfun sells one:
However, they’re almost $50! That’s pretty expensive, considering the LEDs themselves only cost at most $20. DealExtreme sells inexpensive LED drivers but it’s awkward having to use three boards to dim each channel individually. So, I’ve decided to learn about DC/DC converters and make my own driver board. I decided to make it as flexible as possible: be able to drive LED’s in series up to 20V with a wide input range and deliver 700mA. Also, if I could make them low cost then I could potentially sell them too. I considered several driver IC chips and iterated through a design for each of them:
- National LM3402. This seemed like the best chip to use. Very powerful, and pin compatible with the LM3404 for higher currents. I spent a while designing for it but then realized that it operates in a strange mode where the output current isn’t set by one sense resistor, but by the output voltage, inductor, etc. Scrapped.
- ON Semi NCP3065. The next choice. Very cheap, but it was limited to a switching frequency of 125kHz. It was difficult getting it to work for a wide range of input/outputs and still be small. Scrapped
- Supertex HV9918. This chip was a little more obscure than the other ones (ie. not on Digikey), but it seemed perfect for the job. It could do high switching frequencies and appeared simple to use.
Since I basically have no experience with DC/DC converters, I decided to spend a weekend trying to simulate my design as much as possible. But there aren’t any SPICE models for any of these chips, and the design guide was especially sparse for the chip I decided to use in the end, the HV9918. So, I attempted to model the device myself, based on their block diagram and chip specs. I think I was mostly successful:
Blue shows the inductor current, and green shows the LED current. The 0.47uF output capacitor across the LED’s reduces the ripple by quite a lot, from 300mA to 30mA.
The HV9918 offers dimming from both a digital input and also an analog 0-2V input, which is great for relieving the microcontroller from doing the dimming. However, the dimming is highly non-linear below 0.1% duty cycle, as the inductor has to charge up, and the output capacitor has to discharge, so it is not possible to dim smoothly at very low light levels. Also, since this chip uses a high-side current sense, it is difficult to add an op-amp to alter the sense voltage as I had originally planned with the other chips.
When the parts finally come, I’ll see how well my model matches up with the chip itself. I’ll also be testing them at 700mA driving some Luxeon Rebel LED’s.
I’ll be designing my board for the Sparkfun driver mechanical layout so that it is compatible with their LED board/heatsink solution. If there’s demand for them, I’ll try to sell them. My intended price points are $20 for the three-driver version and $10 for the single driver version.