I've been continuing on with my coil winder project, and I had planned to put a work light on it using some of the new-fangled white LEDs for lighting. Since I had planned for quite some time to also experiment with using them as strobe lights, I decided to combine the two and see how useful a strobe would be on the winder - if the strobe was no good, I'd still have the light anyway.
I knew the main issue would be the phosphor response - both brightness and speed - whether it could reasonably render a still or slow-motion rotary action image. Since the winder will only go to a few hundred RPM, I figured it would be able to handle it.
I have an assortment of LED light assemblies from various bulbs and fixtures I have collected over the past couple of years. None have any usable part ID to search for specs, so it took some experimenting with driving them. I ultimately settled on one from a quad rail spotlight - I bought several from Costco a while back, figuring they'd make good general purpose room and shop/work lights. These fixtures are rated over 800 lumens, so 200 per light, and the internal driver supply is a CC type 350 mA up to 42V. The lights are simply paralleled, so if equally shared, each takes about 90 mA. I charted the I-V characteristics on the curve tracer, then figured I'd run the one light on about 50-75 mA at 35V. The light modules each have six SMT white LEDs on a metal substrate, compressed against an aluminum heatsink disc by a plastic ring that also mounts a plastic lens to form the spotlight pattern.
I took one apart and trimmed the disc down a bit, and rigged new hardware to replace the socket inside the light hood of an old 12V bulb type gooseneck reading lamp, which is then mounted on the winder control box perched above the work zone. To get operating voltage, I found a DC-DC converter to stack on the main power 13.8V supply to get to the upper-40V range. Ballasting of the LED module is done with incandescent lamps, which run red-orange (for long life) in lighting mode, and cold in strobe mode. The higher no-load voltage of the supply, combined with the much lower cold resistance of the lamps, allows for high LED strobe current pulses with duty factors around one percent or less.
The bottom of the LED unit is driven by a small (2N7000) MOSFET to common via a one ohm monitoring resistor. The MOSFET gate is temporarily rigged to be driven by a pulse generator (HP8111A), so the repetition rate and strobe pulse width can be controlled. The light switch sets the modes - off, strobe, and on. The MOSFET switches on and off only - the LED current is determined by the ballast circuit characteristics, so there is no need for any active current regulation loop. Ultimately, the MOSFET will be driven digitally from either the winder shaft or the traverse position encoder information, and fully automatic and synchronous at all speeds.
I couldn't find much info on the LED or phoshor characteristics, other than that the LEDs can easily take about three times their rated current at less than 10% duty factor, and there are concerns about perceptible flicker in the intensity from line ripple and PWM activity. I also found that most common ones likely use GaN or InGaN blue LEDs along with Ce-based yellow phosphors, so they will likely behave similarly, whether good or bad for this application.
I was disappointed at the current - for this application of course I would prefer a hundred times as much, but that's not likely since the higher current density will at some point blow a bond wire or shatter a die, even though the average power is quite low. If I can get it to survive at ten times, that would be decent. Right now the ballast is set up to deliver about three to four times the rated current. I haven't hooked up a scope yet to see it on the monitoring resistor - I'm just measuring the average current with a DMM. So, in lighting mode, it's about 75 mA continuous (and about 175 lumens out = nice), and in strobe mode about 250-350 mA peak.
This is getting kind of long, so I will stop here, and continue on another post with the experiment results. I also have a bunch of pictures to upload, so I have to figure that out.
David Hess
Gary Robert Bosworth
John Sykes