MrAl: certain Triacs have a holding current of up to 40mA.....and many led bulbs, when dimmed, don't draw that kind of current anyway.
RonSimpson, i see your point, but concurring with the loss factors that MrAl points out, the primary side regulation method seems a bit inaccurate to me. I do however appreciate its cheapness, and lower component count.
When you go on powerint.com site, you can use their design software to do a smps design, and you can pick either primary side or secondary side regulation...when you pick primary side, the tolerance on vout is considerably wider. I don't really favour this, as say, if you're 10% over in output power then that's wasted energy, specially in lighting where there are so many lights
The control of peak current via the Control voltage is rather unusual in LT3799....I played with the simulation model with a linearly decreasing control voltage, and the vin_sense voltage fixed at various levels.....for each vin_sense level, the higher Control voltage values did not affect the peak current level (-say between 1 and 2v for control voltage)...then when control voltage went below 1v, say, then the peak current level changed.
I was expecting the behaviour of the "control" voltage to be a little like the "COMP" pin voltage of the uc3842, but it wasn't.