high power led supply design

[caga]

l4960

Hi,

Anybody know good design for feeding high power leds like 1w star?

The first question rises what is the best current / voltage design to feed lets say 6 or 12 of these leds, if you make 6 series in two rows for example 22.2 volts / 0.7A, or with 2 leds 6 rows, 2.1A, 7.4v ???

To make a battery powered design with high efficiency, must use PWM or switch mode converter, for sure, but which of them?
I already tried LM2576, which has a eff of %65-70, vwry low! Also tried L4960 which has %80 eff also seems to be low.

Have UC3843 in mind now but could not find a good solution or example as a start, because UC3843 design is not simple as the others, some mid-advanced feedback is required I think. I found few ideas but all depend on voltage regulating rather than current. No isolation is needed but if the input range is wide lie 8-36v with high efficiency it will super!

Have any idea?

 

[platonas]

Hello caga,
I recoment using one row of LEDs since only one resistor is used in series to control the current passing through, hence lower losses in resistor.
In order to increase the efficiency of any switching regulator , you have to make a good PCB design, use very low ESR electrolytic capacitors , use a good Inductor and make sure the core is not saturated and finally use the appropriate switching diode.
More comments on the above:
In case the inductor is small, it saturates and cannot store more energy, the eficiency drops and the switching device gets hotter. An inductor can be wound with more than one wires in parallel in order to minimise skin effect due to high frequencies of operation. For a 1mm wire you can use two wires of about 0.5mm and wound them together or even more thinner wires if you have.
The diode is also important, choose one that suits the frequency of operation of your circuit, it has to be a fast diode.

The ICs you use are quite good and i cannot see why you have low efficiencies.
Regards
P.

 

[caga]

Thanks.
The LM2576 and L4590 has such eff in their datasheets so I think by design their eff is not suited for battery operations. I do not have the chance of testing uc3843 yet, but know that it is used a lot.

Winding coils with many wires seems to be good idea, I will try.