LED perceived brightness and efficiency, linear vs. PWM control?

[LightForce]

Hi,

Could someone briefly explain, which setting gives me better results in perceived brightness/runtime ratio?

Let it be the two buck converters. The two has simple high/low/candle settings and identical efficiency of 80%. This efficiency will remain flat at all modes in both the converters. Let the linear converter has 1000 mA high, 100 mA low, 10 mA candle mode, and the digital converter 100/10/1% duty cycle, with ON LED current equal to 1000 mA.

Let it be the LED. It will have 220 lm and Vf=3.4V at 1000 mA, 33 lm; 2.9V at 100 mA and 4 lm; 2.7V at 10 mA

My questions:

What will be the energy usage on low and candle mode in both drivers?
What will be the light output in lumens?
What will be the perceived brightness? (I heard that "PWMed" lumens are different from constant current lumens... Is it true?)

I got lost in these all numbers...

Cheers,
Damian

[evan9162]

This is easy, you've provided all the needed information, just calculate it.

Constant current:
1000mA: 220 lm, 3.4W, 65 lm/W
100mA: 33 lm, 0.29W, 114 lm/W
10mA: 4 lm, 0.027W, 148 lm/W

Now for the PWM:
100%: 220 lm, 3.4W, 65 lm/W

10%: 22 lm, 0.34W, 65 lm/W
Calculations:
output = 220 lm * 10% = 22 lm
power usage = 3.4V * 1A * 10% = 0.34W
efficiency = 22lm / 0.34W = 65 lm/W

1%: 2.2 lm, 0.034W, 65 lm/W
Calculations:
output = 220 lm * 1% = 2.2 lm
power usage = 3.4V * 1A * 1% = 0.034W
efficiency = 2.2 lm / 0.034W = 65 lm/W (seeing a pattern here?)

Let's bring it all together:
At 1000mA/100%, they're both the same (as expected)

At 100mA/10%, constant current outputs 34 lm vs 22 lm (50% more), and uses 0.29W vs 0.34W. Constant current efficiency is almost double that of PWM.

At 10mA, 1%, constant current outputs 4 lm vs 2.2 (almost double), and uses 0.027W vs 0.034W. Constant current efficiency is more than double that of PWM.

Once you exceed a certian frequency with PWM, your eyes integrate the individual pulses into a continuous light source, the apparent brightness which is equal to the average brightness over time. Thus, the calculated output due to the PWM duty cycle matches exactly with what your eyes will percieve the brightness to be.

Notice that with PWM, things don't get any more efficient with lower current, which is a defining characteristic of LEDs. PWM is less efficient than constant current for dimming LEDs. People do not use PWM to increase efficiency with LEDs, they use it to reduce the color shifting that happens when you dim an LED. However, the color shifting is not something that's noticable or that even matters for general lighting applications, so when trying to efficiently dim an LED, PWM is definitely not the way to go.

[LightForce]

Thanks alot evan9162

Yes, this is obvious what you said. But I've heard somewhere, I think even here on CPF that human retina detects the maximum luminosity, rather than an average, so being right with this theory perceived brightness will follows nonlinearly with linear change of duty ratio (higher at low duties). I wonder if it's true, especially that you said it's not.

D.

[evan9162]

It's not true. There's so much misinformation about PWM dimming here, it's sad. People just make assumptions rather than sit down and do the math. We have to debunk the PWM myth at least every 6 months here. It's been happening that way for at least 5 years, if not more.

If the human retina detected maximum luminosity, how would PWM dimming ever dim anything? If that was true, your eyes would see maximum brightness all the time, even with a 1% duty cycle. Suddenly, you're breaking laws of thermodynamics. just think about it rationally for a few minutes, and you'll come to realize that such an assertion is not supportable at all.

[chimo]

Here's an excellent thread from the past that supports evan9162's post.

http://www.candlepowerforums.com/vb/...ad.php?t=70073

[LightForce]

Oh, yes I remember this thread - deffinitely should be a sticky!

[zehnmm]

What a great thread! Thanks to evan9162! Definitely sticky material.

[MrAl]

Hi again,

Yes this has been discussed many times here on CPF, starting around
the year 2001 or something like that. We talked about basically the
same thing, dimming with a change in current or with PWM. The change
in current allows the LED to work more efficiently because it's efficiency
curve goes up for lower currents, while for PWM it's always being banged
by the max current so there is no efficiency increase because there is
no change in dc operating point.
There is reason to believe that PWM lowers the efficiency even more because
of another issue that comes up with the human eye...since the light pulse
contains very high harmonics it is reasonable to assume that the eye can
not respond at all to some of these harmonics and so some energy might be
lost that way too, although it is difficult to say just how much without doing
some tests.

We also talked about dimming via PWM and how it might be better because
after all the LED is not 'on' for the whole time, as when dimming with a
variable resistor where the current is lower but the LED is 'on' for the whole
time. The PWM loses again, because when the LED is pumped with PWM
its peak has to be much higher than with constant current to get the same
brightness, and the energy lost in the series resistance (internal or external
to the LED) is the same as the energy lost in the variable resistance, but
again with PWM there is no efficiency boost because of a decrease in the
dc operating point. At one time i actually proved this with a set of equations,
i think back in 2001 if i remember right and posted these on CPF.

The conclusion reached was that the only luminous advantage PWM had
over pure variable current was that the color could be controlled better.
There is another advantage however, in that the series pass element
(transistor) using PWM does not have to dissipate the high power that
a linear operating device has to handle, so you can use a smaller and
*much* more compact device (MOSFET).

[VidPro]

my eyes percieve the average. except when we were using it for Strobe light at lower rates.

there is no doubt that PWM is not as efficent as running the Led at Lower total currents , and therfore at higher efficency levels.

but i would rather have PWM then no level adjustment. no mater how inneficient PWM is, if you only light up the led for 1% of the time, you Still get the longer runtimes.
also it SEEMS that temperature is a major factor, while being higher efficency with current control is more light & less heat, at LEAST with PWM the heat is reduced, not at the actual silicon gate, but it can passivly move away from that gate.

so while PWM is garbage compared to full control, it still has uses, because you can get more runtime and less heat , NOT that current control wouldnt do that better, its just that PWM can get that done.

See i might have said it right this time, instead of arguing with Newbie about it being "BETTER" or "OF NO VALUE AT ALL".

Then if you OVERDRIVE the led in pulses, then PWM is still overdriving the led, there does not HAVE TO BE overdrive to have PWM.
i know of no book on PWM that says you cant do PWM at well below spec within the (more) higher efficency levels of the led.

you try and explain that, without getting into a 50 page diatribe about how current control is better than PWM, it IS nobody doubts that, but PWM can functionally run the led, heat, and the visable light lower, and increase runtime.

the METHOD of current control, then becomes important, if your going to just TOSS out the extra power, using a resistive current control, you would have to balace that out with your efficency of the PWM and the power required to run the pwm controller itself.
and the better efficency of the current control with losses that occur due to the current control curcuit.

what it probably comes down to is the total package that is created and implemented not what one sees on a bench power supply. 20% loss here 50% loss there , 30% losses at the led, percentage of losses due to overheating, lifetime of the curcuit, and the teeney package they make it.
PWM based devices can be made to act in similarity to current control devices by just averaging the pulses out into one more hunk of lossy electronics, before it goes to the led itself.

lest we leave out one factor too in PHOPSHOR leds, is phospher persistance, the phosphers used would effect the persistance of the phospors. people would say that 60htz CRT monitors would drive them buggy , but there was huge variations in phosphor persistance used in monitors. what one person saw the other one did not, the claim was the refresh rate differance 60-200 say fed to the monitor, but the lines were still being drawn on the monitor in little pieces, it was only the phosphors that kept the light to percived levels. Phosphor persistance was why some percieved 60 to not be enough, and some 100 not enough. then turn around they get LCD monitors which have a light valve persistance , and they think they can run the light valve at at 200 refresh rate.

led manufactures really like thier on time and off time specs , .5ms turn on time, why dont they put in higher persistance phosphors ever in led item? what would be the ramifications? ever see a led with more or less phosphor persistance?

which brings up "gate triggering" at some current level some of the "organic" gates will not triger at all, i call it organic because the etching and doping process is a very rough process done with organic acids and baths and such. (at the molecular level)
Question: are all gates in a led die triggering when there is not enough current for a electon jump of the gate? ahh see you didnt think about that :-) if at low currents only close gates are triggering, what is the life of the low gates, vrses the high jumping gate. (something like that) only way to tell is very long term testing.
Dont we basically know that? lets see, the wavelength of the emitted light has something to do with the distance right? and PWM maintains the color of the light the same, but current control doesnt, hmmmmm . could it be that with current control shorter length waves are produced because shorter jumping locations are still operating, at the lower voltages :-)

ya see at the molecular level as seen in other silicon devices, there is this rough molecular surface, not some magic perfection, that means that some jumps are longer and some are shorter to get across to the other side. the more "precise" they make the led gate, the more efficent and the more singular wavelenght would be produced from it.

the reason i bring that up, is because (sloppy) multiseries parellel leds wont have all the gates (or even whole emitters) triggered on low current control, short gates will invariably age faster? PWM has a jump occuring on every gate in the die., which is probably why it eventually kills every gate in the die, and every phospher it emitted to.

so if any of that ^ was true. Why does the voltage required to the led change over time, to LESS for the same current, why does the color change over time? something died inside, something changed in resistance. hmmm, wrack brains, less voltage same current, the resistance went down?? right?

2 ways leds die, one if you smack it hard it breaks to an open curcuit , resistance up. if you drive it hard, or just over time it does the opposite, resistance increases. in tests this gets so bad, it goes from Diode to Resister completly.
so
little broken gates are passing power through, and probably not putting out any light? then my logical estimation is that any led that had dropped in Vf over time has less light, more fried gates are just passing current through. and they would be the gates that are most likly to go from short jumping to charred remains, not the higher voltage gates?

so if i can wrap my brain around the cause of Vf lowering, and color change, would it be dead, short jumping, low power, high frequency, gates in the emitter are fried? so the color should shift to lower frequency light.

SOOO
if you use current control, you will fry less short gates , preserve the long and short gates for high powered operation, are less likly to have a change in the Vf over time, and are more likly to have more light and less useless resistance in the led.
or
they could just make the leds better :-)

wouldnt that mean :-) that over the 100 hours where the led voltage (for the same current) drops , that in reality in 100 hours the item is basically burning parts of itself up, leetel teeney parts . and any lumens efficency over time, so a Led that did NOT decrease in resistance, or change in Vf , or change in color, would still be in very good shape. In this TIME situation, the low Vf led is NOT more efficent, its lower in Vf for the current because parts of it are going from diode to resister.

if they tighten up the gate, and precision up the gate, then the lower VF there would represent efficency for the single color output. that would explain how they can get High efficency leds in low quantites, perfection that doesnt translate to mass production.

even in-spec operation fries the weakest link in the led, the one being pushed the hardest for the power going in.

[Roger108]

Greetings,

I've been reading a lot of complaints at the Apple forums about LEDs causing headaches and nausea whenever brightness is lowered. PWM is causing a subtle but annoying flicker. Are there any LED monitors out there that use linear rather PWM control?

Thanks

[DIWdiver]

Back in the day when red 7-segment LEDs were king (obviously many years ago), it was found that an LED driven at 50 mA, 10% duty cycle appeared slightly brighter than the same one driven at 5 mA constant current. I recall reading a paper from Texas Instruments about this.

This was important to know then because LED efficiency was almost constant from 5 mA to 50 mA, and switching regulators were never used to drive them (and for the most part still aren't).

Fast forward a few decades, and everything is different; white LEDs are king, they use phosphors to achieve their color, efficiency drops off at higher currents, colors shift with current, and switching regulators are common.

It may or may not be true that a white LED appears brighter if driven with PWM instead of CC. I suspect it is at lower currents and not at higher currents, but it almost doesn't matter. This effect was never huge, and is now swamped by other factors.

As others have ably pointed out, the LED is less efficient running on PWM than on CC. I don't disagree with this conclusion, but I think the numbers you gave us don't apply to all situations.

For example, to get the lumens/watt to drop over 40% as in your example, the LED has to be pushed pretty hard. If designing a light for maximum efficiency, one might use two of these LEDs driven at 500/50/5 mA. The lumens/watt wouldn't drop nearly as much, and you'd get more light at maximum, for the same power. The difference between PWM and CC would consequently be much less. I'm not saying that one of these choices is right and the other wrong (some lights just won't fit 2 LEDs), just that the choices we make have an impact on the outcome.

Also, very few buck regulators have constant efficiency over a wide range of output current. In fact, it's not uncommon for a regulator to have 85% at full output and less than 50%, even as low as 20%, at 1% of max output. This would shift the balance back in favor of PWM. However, PWM at 1% has its own set of problems...

So in general it's safe to say that at higher currents PWM is less efficient than CC. But at less than full output, especially at small fractions of full output, the answer isn't necessarily quite so clear, and probably depends more on the specifics of the design than on whether the LED seems a little brighter when driven with PWM.

D

[MikeAusC]

It's important not to get confused between the effect of Switchmode and PWM drive of an LED - it is possible to PWM the Switchmode drive of an LED.

PWM drive means that the LED current goes fully on and fully off to reduce the average current to the LED. The drive to the LED can be direct drive or via a Switchmode converter.

Switchmode drive is also used to change the average current, but the current is steady and continuous. Switchmode can be used when the battery voltage is significantly different to the battery voltage e.g. driving a single LED at 3.5 volts from 8.4 volts or from 1.5 volts. Switchmode can be used to drive the LED at its maximum current or to dim it. If you see a coil of wire or a block of ferrite, you can be pretty sure it's a Switchmode Converter.

Mostly what the LED sees is essentially the same as DC drive. Alternately it's possible to use a Switchmode driver that delivers pulses to the LED if no capacitor is put after the diode. This reduces the efficiency of the LED (because it's peak current is twice as high) but improves the efficiency of the regulator.

The improved light output from PWM being discussed here will not be seen from a Switchmode Driver - unless it's followed by a PWM dimming stage.