Update on the Leeslamp
I'm afraid I just can't help myself... After discovering that "rolling your own" transformer would set me back at least 5 euros as I would have to purchase every part of the set separately (two "E" cores, a bobbin and two clips) I figured that using a second hand transformer from an ATX power supply should yield similar results. The standby supply uses a separate switchmode power supply, and its transformer seemed ideal for the job. As I have quite a few of those ATX babies I removed three transformers and tested them with my superduper LCR meter. I had figured that instead of the 300 mH of the laptop backlight converter transformer I would need like only 60 mH of inductance in the drain winding. It turned out that I had primaries of 0.97, 2.6 and 20 mH, with secondaries (feedback) windings of 2.4, 2.7 and 30 µH, respectively. The actual secondaries I just ignored. I decided to go with the 20 mH transformer, as this was the closest that I would get to my simulation.
Another thing was that my frequency would be much higher now. As it turned out, it was 70 kHz. Now the question was if the LEDs would need to be protected against reverse voltages and how to do that. In leeslamp2 I had added an extra capacitor across the LED string to compensate the recovery time of the series diode (BYW26E). I needn't have worried: the recovery time of this device is sufficiently fast to avoid any reverse voltage building across the LEDs. So I just left it out.
Now the only thing left to solve was the power storage. As the rectifier produces an pulsing DC voltage (100 Hz, peak voltage 325 V) it needs to be smoothed. The electrolytics I used have a fairly high internal resistance (ESR) at these frequencies, so I used a polyester foil capacitor. This particular unit was used in a EMC decoupling filter, but any foil capacitor with a sufficiently high working voltage works just fine here. My unit is 0.47 µF. It may very well be that an electrolytic will work just fine, though.
Leaves still the issue of primary inductance. I think that the two low values are a bit too low (even though I did not test this) to work confidently. YMMV (your mileage might vary). But how to gauge the value of it? I found that the resistance of the primary is about 2-4 ohms per mH, so an easy resistance measurement with a (digital) multimeter should give a good idea on how to proceed. And if the oscillator doesn't oscillate, just reverse the connections of the feedback winding. It so happened that for my transformer the windings were wound the right way around!
The final results: these 14000 mcd LEDs are soooooo bright! They also get warm to the touch, in use, as does the FET. This transistor doesn't have to work too hard, only the switching times are really slow so there are some losses. Not as much as a dropper resistor though. Which is a lame solution anyway! And I only have to use eleven parts: a bridge rectifier, four resistors, a FET, a zener diode, a fast recovery diode, two capacitors and a transformer. And 37 LEDs.
 Brightness |
 Storage capacitor |
 Storage capacitor again |
 Tosh 2SK1118 |