>>Ok, i see that there are people in here.
>>
>>I am running a led fixture made out of a fluorescent fixture and a MC PCB: 8 1W leds are soldered to a MCPPCB which is glued to the aluminium shade by means of a glue and thermal paste. See photos for your reference.
>>
>>The 8 leds are mounted into 2 series of 4 leds.
>>
>>
formatting link
>>
>>
>>In my opinion, the heat is partly because those leds are cheap and ineffcient. My question is if i replaced them with better quality chip leds with proven efficiency of about say 130lm/W, and run them at 350mA Would i still need to add over radiators to this homemade fixture ?
>
>Look carefully at the LED specifications. The life time for those "1
>W" LEDs are usually claimed at 150 or 175 mA, dropping rapidly at 300
>or 350 mA.
>
>The 130 lm/W is claimed at 150 mA or similar, but may drop to 100 lm/W
>at 350 mA.
>
>Run the LEDs at 1/2 maximum power and compensate the loss of absolute
>brightness by doubling the number of LEDs. This will also simplify
>thermal design, since the internal junction temperature rise relative
>to heatsink is also halved.
There may be an other advantage. Running LEDs at maximum capacity requires almost constant current supplies, mostly approximate constant voltage with feedback, being careful not to surpass maximum peek voltage. This contrast with the practice in LED displays where duty cycles of 1:10 are common. Apparently low power leds have a much larger peek to continuous ratio, but I never came accross an article drawing attention to that. However if you feed the LED a sawtooth, the average voltage is halved. A sawtooth is easily generated, because it is the natural decay of current supplied by a loaded inductance coil, assuming the voltage over the led is approximately constant, which is of course the LED voltage.
P.S. I've build the circuit at
formatting link
It works, but the intensity and the frequency depends on the quality of the transistors. Non-descript vintage silicon NPN fares much worse than the BC547C (BC547A does fine). The scope reveals that the LED only lights up about 1/3 of the time at a frequency of about 50 Khz.
I would like to try the circuit with germanium transistors (I've still a lot of those) and see how low it would go. What changes to the components would be required? I don't understand the circuit well enough to experiment fruitfully.
Groetjes Albert