Correct way to drive a pulsed LED

That duty cycle means that the LED is on only (500*10^-6)/(15*10^-3)=0.033 or 3.3% of the time, so it appears to be about 3.3% as bright if it were powered at the same level, continuously.

Find a way to increase the duty cycle.

All diodes have some resistive losses, in addition to the junction drop. When you pulse the current, the instantaneous resistive losses go up proportional to the square of the peak current. That heat must then diffuse out of the places where the resistance occurs to the whole thermal mass of the device and then to the air while the peak temperature at the resistive spots stays below the threshold of damage. LEDs that are designed for low duty cycle operation have had the series resistance lowered as much as is practical (double wire bonds, thicker metalization on the die, etc.), and thus, have higher peak current capability. These changes often increase the price.

That said, the 90 mA peak current spec is time limited according to the repetition rate. On page 5 of the data sheet:

there is a graph of how long a pulse can last at the peak value for various repetition rates. There is a line for 100Hz (10 ms period), so your 66 Hz line would be just to the right of that. This line tells you that your pulses can last about 2000 to 3000 us at the full peak current rating if they are repeated every 15 ms. The LEDs designed for low duty cycle operation will show a curve that sweeps up much higher for short pulses, instead of leveling off at a fixed value (of 3 times the DC current rating), like these curves do.

So extending the on time to 2000us out of every 15ms would increase the average light output by about 4 times (about 2 visible steps in apparent brightness) what you are getting with 500 us pulses, at the same current.

All right, I will do some more tests tonight.

I did do experimenting with duty cycle last night.

1/33th vs 1/22th vs 1/11th. All with 100 ohn resistors I could barely tell the difference between 1/22 and 1/11. There was a noticeable difference between the 1/22 / 1/11 and the 1/33. With a 25ohm resistor on the 1/33 led it looked as bright as the 1/22 and 1/11. The 1/22 and 1/11 looked VERY close.

So no matter what duty cycle, I shouldn't drive the LED with more than

90ma? The maximum pulsed value on the datasheet?

My display is organized in characters of 8w x 11h pixels, 22w x 9h characters.

I'm planning on desgining a refresh controller that handles one or two lines. So the maximum amount of LEDs ON at one given time for a 9 line display would be 8*22*5=880, 880*.09ma=79.2A. Now if I had all 9 rows refreshing at the same time 8*22*9=1584, 1584*.09ma=142.56A!!!!

Parts count would be lower refreshing 2 rows at once, and current would be less. So 90ma wouldn't dammage these LEDs? A duty cycle of 1/22 at

70hz

Thanks, Grant

Thanks for the help! I will make sure to post pictures Friday night when we start constructing it. :)

Thanks, Grant

8 pixels wide for each character, 22 characters per line, 9 lines total. 880 LEDs if all are lit.

However it seems like with the resistor being equal between the 1/22 and 1/11 the light output is "the same". I could be wasting current on the 1/11 point though...

My idea was to use my canon camera with manual exposure settings and no flash. I should then be able to see a difference in Photo Shop when I look at the pixel values.

One thing I DID notice was that I could see a difference when looking at the LED from the side. Fron the side there is a little bright green point and it was a little brighter on the 1/11 one.

I'll see what I come up with tonight.

Thanks, Grant

That's what the data sheet says to me.

You will have to explain to me what some of those constants represent.

Total current is proportional to total light output if none of the LED's are in series with another.

That's what the data sheet tells me. That would correspond to

1/(70*22)=649 us every 14.3 ms.

logjam wrote: (snip)

Eyes are notorious at being poor judges of light intensity, especially point sources and pure colors and trying to remember one intensity when you see the other. Measure the light with a photo transistor. Or do the experiment at night with the LED alternating between the two duty cycles on alternate seconds and look at it from 100 yards away. You should be able to see the two-to-one intensity change, clearly.

I have a Canon digital camera and found it to have a "gamma" of about .7. Pixel values would then be proportional to light successfully received raised to the .7 power. A pixel value of 100 represents half the brightness of one whose value is 162. I tried this with my monitor displaying alternatively a small solid white area and a same-size/shape area with alternating lines same white and black - with the camera defocused.

Makes me think that the LEDs could be gallium phosphide red - a chemistry sometimes known as "low current red". These have a nonlinearity favoring low steady currents over high pulsed currents with short duty cycle. If you have such an LED being fed low duty cycle high peak current pulses, then expect an increase in brightness by adding a capacitor in parallel with the LED. Just don't apply a capacitor to the LED while the capacitor is charged well past the LED voltage - this could blow the LED.

Gallium phosphide "low current red" LEDs do have significant color and spectral changes at higher currents when red tinted filtering is lacking or thin. At 90 milliamps they can be more yellow than red, with a "green spot" if viewed through eyeglasses that have a prismatic effect.

Most red LED digital displays have (or had in their heydays) gallium arsenide phosphide, which have a more-opposite nonlinearity that usually benefits from higher instantaneous currents even at same average current. Those even caused a myth to arise that human vision has some sort of "peak-detection" mechanism to explain why these LEDs appeared brighter with pulsed current than with steady current where the average current was unchanged.

Gallium phosphide green lacks the low-current-favoring of gallium phosphide red but is closer to being like GaAsP red, BTW.

- Don ( snipped-for-privacy@misty.com)