Linear LED driver

[dareon]

I am working on making a linear constant current LED driver. Attached is a circuit that I have used and works but has some issues(this is a simplified circuit)

The circuit works well and is pretty simple accept. I need to be able to Dim the LEDs by PWMing the input power. When doing this when the square wave goes away. The input voltage drops to low to maintain current regulation so the opamp saturates the FET. When the wave comes back enough voltage stays in the caps to keep the opamp on and the FET staturated. This causes a short(few microsecond) overshoot in the current going through the LEDs before the opamp can correct.

Any ideas on how to fix this. I am open to totally different circuits but I cannot tolerate much dropout voltage.

[5th element]

Have you tried implementing PWM dimming at the FET rather then at the input supply? In other words you leave everything powered on all the time and simply turn the FET on and off with the PWM signal.

[dareon]

Something like that would probably work. This is for a product and that is the way the customer is currently doing it. I'm not sure I can get them to change it.

[5th element]

So in other words the control system is already in place as a PWM controlled power source.

How about this.

The op-amp has it's own capacitor feed on the power rail such that the power to it could be considered un-interruptable.

The LED string has its own separate capacitor that sits across the LEDs.

When the PWM signal turns things off, the LED string itself is disconnected but the cap across the LEDs is left in place, this helps by keeping some regulated signal present to the inverting opamp input, so the control loop is somewhat maintained, even in the off cycle. Would something like that help?

[richie00boy]

Choking the supply off will never work, remember the op-amp is trying to maintain constant current.

The best way to do it is to chop the voltage reference on pin 3. Keep pin 3 happy with a 1meg resistor to ground.

The second best way to do it is to alter the voltage reference to get the dimming, and not use PWM at all.

The third way, which may result in some overshoot, is to chop the voltage to the FET only and maintain the supply to the reference and op-amp.

If you use PWM, make sure it is not a whole multiple of your supply mains frequency or you will get a pulsating effect.

[dareon]

I don't believe so. I'm not sure how I could disconnect the LED but I think I would still end up with some overshoot.

What I have done now is I compare the input voltage with the a value and when it drops below a value I pull the negative feedback high with a FET. This seems to be working but might not with higher pwm frequencies and duty cycles.

The problem with this is that I have to measure the voltage after then input diode.
When the LEDs stop conducting this voltage never gets drained all the way down and picking a voltage to pull up the feedback is difficult. If the voltage at C4 rolled off quickly and consistently it wouldn't be a problem. But it doesn't this varies with supply voltage.

[dareon]

richi00boy

The problem I might see with the best way is what about input offset voltages? I believe they can even been inverting. I worry in all cases this might not be enough.

[5th element]

The idea of the cap was basically copied from some LED drivers that maintain the state of the output capacitor so that the system can respond quickly when it is turned on after an off period from PWM dimming. I thought that it might help to speed up or reduce the effect of any overshoot when the system is turned back on.

[richie00boy]

As long as you choose an op-amp than can accept rail to rail or ground input it will be fine. CA3140 is one common op-amp that does.