# buck converter for high current LED

My first question here:

I want to make a simple adjustable current supply for some LEDs I have (approx 3 V and up to 3 amps). I've decided against just using a single FET to control current as it would have to dissipate as much as 9 W of heat. For such a circuit I don't mind a bit of ripple (above say 10 kHz) in the current since its driving an LED which is driving phosphors anyway (the LEDs are white).

I'm not a power engineer so I have the question:

Would the circuit below (which I pulled from my behind) work? I haven't seen anything too similar online. Aside from some ripple equal to whatever the hysteresis (divided by R1) are there any problems? Are there concerns with setting the right oscillation frequency? Should I choose component values to target a specific frequency (I was just going to go with >20 kHz in case something vibrates)?

• Do a search for "hysteretic buck converter" for some ideas. I'm not sure that the circuit you have as configured will work. Apr 15, 2013 at 2:16
• Is there some design objective not mentioned, which requires intensity control by current regulation rather than the more common PWM approach? If you use PWM and drive the MOSFET fully on/off, the dissipation load on the MOSFET becomes much smaller than the estimated 9 Watt mentioned. Apr 15, 2013 at 5:23
• I think you misunderstand how LED's are generally PWMed. They are still either current limited with a resistor or driven from a current source. For the sake of this question and my goals, wasting 9 watts in a transistor or resistor is not okay. And if I already had a current source this circuit would be redundant Apr 15, 2013 at 6:13
• @user1816847 No, I understand the mechanisms well enough - I'm trying to understand whether the crux of the problem is current regulation, or intensity control. Basically, "Please tell us the actual problem, not just your hypothesis on how to solve it". Your comment now provides the clarity that the power loss is a concern even if it isn't on the MOSFET itself, the question left that ambiguous. I'd still suggest clarifying the question somewhat, but that's up to you. Apr 15, 2013 at 8:37

Your idea is OK, but your exact circuit topology is not. Your biggest problem is that the current sense resistor R1 is not in the path of the inductor discharge current. That means it's useless as feedback to decide when to turn the FET back on. It will report no current whenever the FET is off.

One way to fix this is to put R1 on the high side, immediately in series with the top LED. The control voltage is then referenced to the supply instead of ground and the logic needs to be inverted, but the concept should work. Since you are using a pot to produce the control voltage, it can easily be wired backwards to produce the supply-referenced signal. You will need a opamp with common mode range that extends to the positive supply, though.

Another possibility is to keep R1 where it is and use a high side switch instead of the low side switch you show now. That way R1 is always directly in series with the inductor. The anode of D1 would then go to ground.

Tricky issues will be to deal with the opamp slew rate, carefully calculate the inductor charge and discharge profiles, and to make sure the inductor never saturates. There is a chance the circuit will be meta-stable, but that by itself shouldn't really cause any trouble as long as the overall oscillations are high enough frequency.

I can see where you are going with this but the circuit won't oscillate. I did a quick simulation and came up with this diagram: -

Be aware that your LEDs I modelled with a 20R resistor and that this circuit needs refining for it to work correctly with your requirements.

V2 is the demand voltage and it's currently set at 0.05V meaning the demand is very low i.e. the 20R load takes about 20mA. With V2 at 1V there is 200mA in the load. Like I say this will need refining for your circuits but it should give you a good idea how to get it oscillating. The trick is R5 and C1 and taking R4 from the current sense resistor R1. It then should work but it needs refining - did I mention that!!!