Enquiring minds want to know: - How much did you pay per metre for the strips? - Where did you source them from. - Do you want more ? :-) _____________ Be wary of life of unknown brand LEds when run at (or well above) rated current levels. This CAN be only hundred of hours I intend to do exactly what you are doing in the near sometime. Hopefully, after you have done it so I can copy you :-) I have a range of strips and they are operate with variable degrees of enthusiasm when 12V is applied. Presumably some makers think brighter lights will sell better. They are on Al backing and the brighter ones get excessively hot rather quickly if air cooled alone. Even the ones that are OK at 12V would be in danger of frying if used in a = car. At nominal 12V x 3A your LEDs are dissipating 36 Watt (as you know) so the 6W is only an extra 16% efficiency wise - and the on PCB resistors are probably dropping 1.5V to 3V at 12V in. If you want maximum efficiency and minimum heat you would like to eliminate the on strip series R's BUT you then get LED drive imbalance - so not a good option with what you already have. A simple buck regulator will easily do what you want. This could be 90-95% efficient with due care, but even with little care 85%+ should be easy enough. There are many of these available from ebay and elsewhere and it is very likely not worth building your own. Without looking I'd guess they may cost $US3 or less each. If you get a higher current one you can parallel more strips if desired. The MC34063 that Joe mentioned would do this job well enough with an external MOSFET switch. The ebay units usually use these. The main disadvantage compared with more modern ICs is that the MC340-63 can be pushed to maybe onl;y 200 kHz so inductor sizes are larger than some. However. MOSFET losses get annoying higher up so maybe not so bad. If you use linear PWM and regulate v average or current average then the LEDs will be exposed to the peak currents and may get unhappy. If you add a series inductor to limit the peaks and smooth things out you are back at a buck converter. If you arrange for continuous operation then Vout/Vin is =3D to PWM on/total time ratio (ie high% x Vin) and you can run it open loop. Closed loop control is very easy. You could use a PIC BUT you could also use a 555 :-). If there is ALWAYS some Vin drop then adding a series R such that VLED is OK when Vin is at Vin_min will move dissipation into the resistor. Using PWM via a resistor into a cap will NOT reduce dissipation BUT will move it all into the resistor (apart from minimal Rdson drop) and you can deripple Vin_LED as require by sizing cap appropriately. Rseries_effective =3D Rin /(PWM duty cycle) so you set Rin to be as large as possible at Vin_min and then PWM from there. This is not quite free as the peak currents into the capacitor are now increased by 1/(duty_cycle) and you may have to use a high ripple current part. Russell McMahon On 17 January 2013 17:18, David Duffy (AVD) wrote: > I have the need to regulate the voltage going out to some LED strips. > These strips are simple 12V DC types that have 3 LEDs and a resistor per > section, repeating throughout the strip. > > The voltage available is nominally 12V, but can vary between 12V and > 14V. The ability to handle up to 18V input is also desired by the > client. The current draw of the LED load is approx 3 Amps. > > I was thinking of a simple 12V LDO, but at 2V differential that's 6 > Watts in heat - ouch! Ideally this unit needs to run with minimal heat > dissipation. I also thought about a SEPIC converter, but that would > take too long to design and debug and be overkill for this application. > > My next thought was to use a PIC12 to PWM a logic level MOSEFT and use > the A2D in the PIC to sample the output voltage via an R/C network. In > theory this will allow me to regulate the output and even allow for a > dimming function to be offered as an option. > > There are only 100 units to build initially and maybe some more after > that, but not thousands. A PIC based solution would be low parts count > and fairly easily tweaked to my mind. Does this sound like a reasonable > solution? > David... --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .