I think the problem is the missing life calculation. Here is a data sheet that specs pulses down to 5%. About 4X the current, but for 1/20th the time.
http://www.farnell.com/datasheets/1498098.pdf
I don't see any 20 ma LEDs with current at .4 amps at 5% duty cycle.

 

I think the problem is missing information from the OP.

 

But I think with the LED I posted if you pulsed it at 1.4 amps for 2.5 ms 0ut or 50 ms it would appear twice as bright as if you were running it at 70 ma. Just not near as bright as running it at 700 ma.

 

My understanding of this is that, at the frequencies used for PWM dimming, that pulsing effect is not relevant. To get the brightening effect, you need to pulse in the range of 10Hz, i.e. strobe the LED, and you can easily detect the fact that it is pulsing, not really "dimming".

If the effect were relevant in normal PWM dimming, you would have the odd effect of an LED appearing brighter as the duty cycle is reduced from 100%, passing thru some maximum brightness at some duty cycle less than 100%, and then falling off the maximum as the duty cycle continues to drop.

I'm not aware of that being observed with PWM dimmers.

It's interesting that you said that. One of the problems with PWM dimming (from my own project experiences) is that the low end of the PWM range is too bright, which I got around last time by making a lookup table to make the LED brightness more what the eye expects.

Now you said that, i'm wondering if the lookup table was cancelling out some perceived extra brightness at low PWM levels. I think i was PWMing about 100-150 Hz region.

 

My limited understanding is that 100Hz would be too fast to get the Broca-Sulzer effect, where a strobing light is perceived as brighter than it should be (based on the time-averaged current).

But it's complicated. There's a pretty good article on it here. It's an old field and there are many other references.

 

Thanks for the reference.

It also could be complicated further by white LEDs which use a phosphor to convert blueish light to white light. The phosphor may need a certain amount of luminance to start glowing properly. In that case a PWM'd white LED might be brighter than a very low DC current which might not get the phosphor working well.