found this info awhile back...think it came from knna.

PWM is of course a choice to dimming LEDs, but in general, direct dimming, meaning using a lower current level is way more efficient, at least from a energy efficiency standpoint.
Here we need to take a second, and think this question carefully.
First off, PMW refers to a way of dimming electronic devices by switching on/off very fast. The effect on the output of the circuit depends of its topology. It may produce a reduced continous current, or switching on/off LEDs aswell.
On many commercial drivers, dimming is achieved by a PWM signal, but the output to LEDs is continous, not switched.
A true PMW driver that gives the output to LEDs switched aswell, may be carefully designed to avoid transients. The highest the frecuency used, more problematic becomes the design. When using very high switching frecuencies, with individual pulses on the nanoseconds or still shorter durations (hundred of picoseconds), not all electronic components are able to handle it. Rise and fall time responses of components used must me considered.
In general the shorter PWM pulses used are in the range of microseconds. Response time of most LEDs are of a few nanoseconds. Going shorter is impossible with standard equipment.
As reference to those not used to the metric system, those prefix are diminutives:
mili: 1/1000. or 10^-3
micro: 1/1000000 or 10^-6
nano: 1/1000000000 or 10^-9
pico: 1/1000000000000 or 10^-12
Second consideretion, that refers to the nergy efficiency of the system. A LED is always less efficient as the current level goes up. So PWM is always less efficient electrically. Thus, any photosynthetic gain from the pulsed light might should compensate the lower efficiency emitting light.
The highest the peak current used, for the shorter pulses, the lower the efficiency.
Most experiments with pulsed light on plants are performed in the range of hundred of microseconds, up to a few miliseconds.
In this sense, what matter is the averaged voltage and current over a second period.
Say you are using a 50% PMW scheme: half time on, half time off. During that time you use half the current than if you were using continous. For example, 1A on continous mode equals to 2A peak current with a 50% PWM scheme (at least, it happen with relatively long pulses, as they go shorter, raise and fall time of the components may vary the average). From a current standpoint, both ways are the same. But not if you see at voltage, because voltage required to run LEDs at 2A is higher than to run them at 1A. Thus the average power burned is going to be slighty higher over a second period on the PWM scheme (how much? check datasheet for how much higher is Vf at 2A than at 1A).
Not only that, the LED not emits double light at 2A than at 1A, but way less. So the PWM scheme burns more power and emits less light. The highest the difference between the current used (thus, the higher the PWM frecuency used), the highest the efficiency loss.
Still when using relatively long pulses, a loss of 25% of energy efficiency is pretty common. Any photosynthetic gain from pulsing might be over 25% to worth doing it.
For sure. Always take in mind that voltage-current response curve of LEDs is exponential. Little increases on voltage leads to huge increments on current. What destroy LEDs is excess current. A peak current over specs result on the breakage of the very thin gold wires inside the LEDs, that act as a fuse. Often there is a wide margin for it, as many LEDs rated for 1A may support currents over 2A, and those rated to 2A often support 3A and more, but once you are over specs, you never know when its going to fail.
Notice that the study focused on the different ways of giving light to plants, for given amount of light. Its not interested on study how efficiently is that light delivered, that is an important part of the equation on actual applications.
And notice that when used 2ms pulses, photosynthesis for equal amount of light reduced to half. Why messing with pulsing then?
Aditionally, light densities used, 50 uE/m2 are very low in actual conditions. We use densities 10X of that on our grows (and same for tomato grows).
The study used a PWM scheme 150us long (148.5 off/1.5on), a 1% PWM (1%on/99%off). As a second has 1000000 us, 1000000/150=6666 cycles per second= 6666Hz= 6.67KHz was the frecuency used.
micro=μ. But for keiboard confort, u is often used instead of μ. So a microsecond is written us instead of μs, and a microEinstein, uE instead of μE (a Einstein is a mol of photons, 6.02 *10^23 photons).
pico is 1/1000 of micro.
I have linked a thread about it on my sign. But for doing it, lm emission (way better, mW emission) must be know accurately, aswell as the spectrum. With unkown bins, its impossible to know it.
Glad to find people who likes to experiement. Sure your experience is going to be usefull for all us.
:Peace: knna