Hello guys,
Switching the LED on and off does *NOT* save power over a series resistor solution for the same brightness level. It actually uses more power or produces less light because the LED is operated at the upper part of it's voltage/current curve where the efficiency drops off. It's only a true switching regulator that can save power.
This can be proved several ways, but the simplest is the "averaged AC model" approach, where the turn on and turn off of the switch is regarded as a simple resistor of lower value than a regular series resistor would be. Here we would see the lower values resistor itself drops less voltage, that's true, but it also has to deliver more current if the LED brightness is going to be the same as with the series resistor. And because the LED brightness goes down with increased peak current (not average current) because the efficiency goes down with increased current, we actually end up getting LESS light output from the same average current through the LED !
If we hold the LED to the ridged non realistic spec of having constant efficiency over it's current range, then we see the same brightness and the SAME power wasted as heat in the switch as would be wasted in the series resistor. Hard to believe because the 'switch' has low resistance, much lower than the series resistor, but a full analysis of this circuit shows the truth that they are both equal. Another view is that the PWM is attempting to do a power conversion from that of the power supply to that of the LED, but it is not a true power conversion just a power loss. The only way to get a true power conversion (and possibly higher efficiency) is to include an energy storage element like an inductor (and other required parts).
There are still times when a regular PWM (without energy storage) is beneficial, and that's when the LED is to be dimmed using PWM alone. Then the LED can be dimmed without having to use a heavy duty potentiometer or something like that. But in the case of multiple LED's it seems easier just to switch some off like we had been saying.
Still dont believe it? Do two simulations, one using PWM and one with a series resistor. Find the power lost in the switch and the series resistor and compare.