Any practical limit to multiplexing LED minimum duty cycle?

well, yeah, but

At 0.1% duty cycle the mean current would be 2mA, not 20mA

And at 1% the RMS current would be 200mA, enough to cause accelerated aging through resistive heating

Luminous intensity has only a vague correlation with physiological reality when short pulses are in view.

Cheers

Phil "down with photometrists" Hobbs

For what range of values for "short"?

I don't know about the eye, there does not seem to be a strong effect, but from reliability considerations I stay above 10% duty cycle for "normal" brightness. Proabably less is acceptable in dim settings with modern led dice, and more would be better if you are pushing the apparent brightness towards limits.

--sp

If you do not care about background lighting (eye does not seem to care much if indirect), most bright LEDs are visible in the sub-microamp (DC) region. Eye may "see" things differently with PWM drive but same average power/current.

How about dark-adapted, seeing a single ten nanosecond light pulse from a Huggins lamp at minimum voltage (about 30V if i remember correctly)?

Beats the heck out of me what a crude measure of luminous intensity for that 10nSec 32v Huggins lamp pulse is, but it seems bright. 100V drive seems to not increase brightness, tho. The 32V is the ionization potential which makes testing at lower drives not too useful to say the least.

That's what I said.

then I went on to say "*even if* it has a continuous rating of

I didn't say anything about 1% at 2A.

I do not think that is true

If you draw up the efficiency curve normalized to current, then it's most efficient at 13mA. So my choice would be to drive the LEDs at 13mA, and fit the dutycycle to what light output is desired

Cheers

Klaus

We have a current design that I need to be able to match, but if I can do it smarter and better, I would like to do it :-)

Cheers

Klaus

The peak current of the LED is 65mA, so that is probably the point at which lifetime effects come into play

Cheers

Klaus

My experiments show there is none (at least practical). Even 1e-3 is pretty bright in the darkness. The eye has a shocking dynamic range.

Best regards, Piotr

Well, people are making pulses shorter than 1 fs these days, so there are about 13 orders of magnitude to play with before the blinking becomes visible. I don't recall at what time scale the nonlinearity becomes important.

Cheers

Phil Hobbs

I didn't say that we were blind to short pulses, just that the sensitivity goes down. It's easy to see light from femtosecond lasers, for instance.

Cheers

Phil Hobbs

One good way to estimate when you're hitting the LED too hard is to measure the 1/f noise. At high currents you start to get electromigration due to impurity scattering, which gradually causes dislocations to form.

You can watch that happening by measuring the 1/f noise.

Cheers

Phil Hobbs

So do classical emulsion films but ISTR it has to be really very short before there is a perceptible difference in brightness. I suspect driving LED displays with 10mA for 2ms or 20mA for 1ms would not show any detectable difference to the human eye (except perhaps moving in peripheral vision where the length of the dash would be longer).

I find some car LED brake lights particularly annoying in this respect.

Several years ago I did a project using some Lumex SML-LX1206 series red,yellow and green. I was driving them about 15mA if I remember correctly. I did steady on, blink and a fade mode where the led would ramp up from off to full on and back to off over the course of a couple of seconds. I found that any on time less than about 1200us was really off to our eyes in a "server room" type lighting.

I did experiments circa 2002(?) with an 'A/B' switch that would switch an LED between a low-duty pulse, and the equivalent average as a d.c. current.

The answer to Klaus' question is that the practical limit was set by the increasing i^2r losses of running the LED at high peak current. But down to very short pulses, there was no perceptible difference between a pulsed LED's brightness and the same LED running on the equivalent d.c. current.

I think I tested down to ~10uS, IIRC. I was trying to increase energy efficiency of 'white' LEDs, and it quickly became clear that d.c. was best.

(Some older colored LEDs had a "threshold" current. Pulsing those can increase efficiency, depending.)

Cheers, James Arthur