Lumens for Nothing

This is how I've heard this particular issue described (very similar to Spambot):

You have two flashlights, one with 50% of constant current, and one with a linear 50% pulse. You can essentially double the strength of each "pulse", and if it's fast enough the human eye can't detect the difference and so the light seems brighter even if it's on for only half the time. There's obviously a limit to how hard you can push this sort of thing, and some people are very sensitive to the pulses.

^ correctomundo

some people are very sensitive to the pulses,,,agree!

This is similar in concept to the way flourescent light fixtures work if I understand correctly. They pulse at a frequency that is so fast (60 times a second?) the human eye thinks the light is on continously. I can tell that my Ultrafire LED light is using this pulse method to dim the light in low mode. If I wave it around I can see the pulsing effect more clearly.

120, the input AC is bridged. ;)

Makes sense, they do this with laser diodes. That said I am one of those people that is sensitive to this, so no free lumens for me.

Don't we have this already (I mean it looks like PWM)?

It is driving me crazy! from car backlights to trafic lights, it seems like a stroboscope effect to me.

As a child I was already complaining about the TL- lights in the classroom (nowadays I think TL lights have improved)

I hate PWM'ed LED car back lights and PWM'ed LED signs. Creates streaks of dots in your field of vision when you move your eyes, urgh!

Modern fluorescent tube armatures with electronic drivers and CFLs drive the tubes at a very high frequency which is much harder if not impossible to detect by just looking at them, they are also a bit more efficient and promotes a longer tube life. The trade off is that it causes harmonic distortions in the ac net.

The observed flickering is a problem with lower PWM frequencies. Some drivers have PWM frequencies in the sub 100hz region, and these are visible to the human eye. All my torches I bought with reviews demonstrating a pwm frequency in the khz range, and there is absolutely no flicker visible.

Now back to the lumens output, Im not sure how they have achieved more lumens with a pulsed power source. Maybe the LED responds differently, or is in a different part of its power output curve? Anyway.. ill put out a generic PWM vs Constant current example.

My understanding of driving LEDs with PWM, vs constant current (true constant current) is that the PWM provides less lumens than the constant current due to the decreasing output efficiency with increasing current flow as you tend toward maximum drive current.

Therefore - If you drive an LED at 50% current, It will produce more lumens per watt than an LED driven at 100% drive current. Do we agree with this? Here is a chart from a Cree XP-G datasheet.[IMG]http://i51.tinypic.com/55rymf.jpg[/IMG]

Now If we drive an LED at 350ma current, according to this graph, gives us a reference 100% lumen output.

Now lets drive this LED at a higher drive current, but pulse it to average the same average current. Lets say we drive the LED at 1050mA (3x the drive current). That will give us 260% of the light output, but if we only have it on for 1/3 the time, that will give us an Equivalent drive current of 350mA, but only 87% of the light output. (a 13% loss of output)

Actually thats not the whole story, the Vf of the LED will be higher at a higher drive current too, so there will actually be more than 3x the power(watt) input to the LED at 1050mA. Heres another graph. [IMG]http://i53.tinypic.com/4zw8pi.jpg[/IMG] At 350mA, Vf is 3.0v. At 1050mA, its closer to 3.35v. So if we update the comparison to compare Power input vs Output we get the following.

CC - 350mA, 3.0v = 1.05 Watts => 100% output (gives a ratio of 1.05Watts for 100% output)

PWM - 1050mA, 3.35v (0.333% duty cycle) = 1.1725 Watts => 87% output (_equivalent_ to a ratio of 1.34 Watts for 100% output)

So to put it simply, with PWM, you get a 21% loss of output in this example. (assuming my maths and fundamental understanding is correct)

So with that example, Simple calculations assuming constant Vf, we already see a 21% loss in output for the same power input

So why would we use PWM? PWM is a simpler (read: cheaper) way to control led power. Current control circuits are also more difficult to implement over wide outputs, as in a high output CC driver is fine at 1A drive current, but will be harder to keep operating stable at moonlight levels, the problem being a visible oscillation (or a different type of flicker). The other reason is tint, LEDs are rated for tint at a certain operating current, but many will green shift when driven at lower currents. Driving the LED at full current, but pulsing it will allow the same tint, but permit brightness control.

An example is the Sunwayman V10R I have, that is supposedly a true current controlled circuit, when the brightness is dimmed, the tint visibly becomes more green. My D10 ramping torches also are able to show tint shift when the battery runs low. It is a PWM driver normally, however when the battery runs low the battery may only able to provide say 0.5watts of power, but the PWM part still can change duty cycle. When I ramp the brightness up from minimum, I can see the brightness increase as the power to the LED increases. At this stage the power is pulsed at full current. As I am ramping up the brightness, the PWM duty cycle is increasing toward 100%. However I reach the limit of what the battery can provide about half way up the ramping, and the power to the LED doesnt increase. So as the duty cycle continues to move toward 100%, the battery can no longer deliver the power, and the led drive current must fall proportionally. So what I see is an equivalent of a constant current drive at a lower current level, and therefore I get a visible green tint shift from my torch.

Sorry If that was too confusing, especially the last part.

Ignoring all that I said before... Its interesting that in the table of information, it lists the luminous efficiency, and at 13.5A continuous you get 650 lumens, at 20lm/W. At 22.5A 50% Pulsed, you get 1150 lm at 18lm/W. THe lumen values dont add up, but the efficacy is still lower than at the continuous current level??

Maybe the 1150 lumens is the measured brightness for when the LED is on??? therefore the output is half that? thats my only guess for now..

okwchin, I agree with you in all your statements (I think).

PWM is not used for gaining free Lumens.

The eye has a certain peaking function on PWM but only slight.

This is confirming some of your statements regarding tint shift.

okwchin said

Ignoring all that I said before... Its interesting that in the table of information, it lists the luminous efficiency, and at 13.5A continuous you get 650 lumens, at 20lm/W. At 22.5A 50% Pulsed, you get 1150 lm at 18lm/W. THe lumen values dont add up, but the efficacy is still lower than at the continuous current level??

Good catch. Going back to page 5 of the original datasheet (http://www.luminus.com/stuff/contentmgr/files/0/50f6c6498fd99d4866dfb0db78e429f7/miscdocs/pds_001310_rev_02_cbt_90_rgb_product_datasheet__illumination.pdf) things make much more sense if the column labeled 22.5 A Pulsed 50% is interpreted as the _average_, not peak, current. In other words the peak current would have been 45 A. The lumens and efficacy numbers would now seem to be consistant.

But the diagram on page 5 is still troubling. For any given current the lumens are definitely a bit higher, so we're not out of the woods yet.

Any thoughts?

Angus

The limit is thermal, not electrical

Which should cover most of the discrepancy unless I'm missing something.

Which is not as much "possible" as "likely"

quote from Don

The limit is thermal, not electrical
Which should cover most of the discrepancy unless I'm missing something.

I'm not sure which (or both) of us is missing something here.

The solid red line indicates about 700 lumens are emitted at 15 A continuous. When the current is pulsed at an average 15 A the output is more like 900 lumens.

The current is the same. The electrical power is the same. Therefore the heating must be the same. Where did the extra 200 lumens come from?

Or to put it another way, I see two different ways of producing 700 lumens
- use a steady 15 A
- use pulsed 11 A

Either way it doesn't make any sense.

Angus

I find that really interesting. The other week, I was doing some regular maintenance work on my car and had it hooked up to my computer to diagnose some relatively minor issue (the alarm going off in the middle of the night for no reason at all... okay, not so minor when your neighbors call you the next day about it; oops). Anyway, after I fixed it (bad inclination sensor which set off the alarm), I was playing around with some of the settings and as I understand it, the daytime running lights (sorry I made you look that one up in the other thread, Pook; I have a PhD in gearheadism... my bad :D ;)) uses standard, regular, run of the mill PWM.

In any case, I could specify any value from 1-100 and the DRLs would dim noticeably when I tested it. The factory value was somewhere in the high 90s, I think, and I didn't mess with it because I happen to think that the German engineers that design these things have forgotten more about it than I could possibly ever know. ;) But the point I was trying to make is that even when I had the DRLs set to 10%, which equals "barely visible" on my car, I never perceived anything strange: no dots, no flicker or anything of that sort. I should probably find someone with a better set of eyes when I get a new light that uses PWM because as it turns out I'm pretty much useless when it comes to detecting that sort of stuff. :D

The daytime running lights have been the standard here in Sweden since long before I was born. I knew that the rest of the world would come around eventually ; )

PWM in only visible if it is done with a low frequency, some cars does this but not all. The only reasons to choose a low frequency is in order to use bad switching transistors or to avoid the need to filter the PWM signal. It is the reason those manafont XM-L drop-ins use such a low PWM frequency, it is in order to avoid using an expensive switching transistor. Higher PWM frequency means a higher number of "switches" per unit of time which means that the switching transistor is going to spend more time in its dynamic range which in turn means that there is going to be more losses in the transistor which in turn means more heat and possibly a burnt out transistor.

PWM above 100Hz is hard to detect if there is no motion, if there is motion then it is very easy to spot. Take a day when it rains or snows, light up the falling droplets or flakes and check the trails left on your retinas by them. If they are continuous, then the PWM uses a high frequency and is not as easily detectable. If the trails shows up as streaks of dots or a trail of short lines, then you have a case of low frequency PWM. I never use my low PWM frequency lights when it snows since I find the "strobe"-phenomenon it creates incredibly annoying.

Exactly as SPAMBOT said, Low frequency PWM is used to keep the costs down. Higher frequency switching needs faster transistors. Faster high power transistors are more expensive!

REgarding the "free lumens" on the graph, I believe it represents the light output for the period that the LED is on. It maybe doesnt count the time when the LED is off.

The IMPORTANT figure there is the lumen output efficiency. the quoted lumens efficiency is still lower at 18lumens per watt vs 20 lm per watt.. (thats actually really really low anyway comapared to an XM-L!! which gets more like 130 lumens per watt)

Interesting point and I agree. Light meters probably work somewhat like sound level meters, and there would be an lowpass RC filter providing a time constant of something like 200 - 500 ms.

Agreed that the efficiencies are important, but we must compare apples to apples.

The 18 lu/W value refers to operation at 22.5 A while the 20 lu/W refers to operation at 13.3 A.

Here's the same diagram with some reference lines added so we can compare the efficiencies under the same conditions.

The power input at 15 A is given by 15 A x 2.4 V which gives 36 W for both cases.

The efficiency with continuous current is 700 lu / 36 W x 100 giving 19.4 lu/W while the efficiency with pulsed light is about 850/36 giving 23.6 lu/W, almost 22% higher.

If the same order of improvement occurs with your XM-L you'd get more like 158 lu/W!!

Angus

I hear what your saying, in that the lumens "output" at the "same" 13A is higher in the pulsed condition on the Graph. This is correct, the graph clearly shows that the pulsed state is far more efficient, producing more "output". Reading this graph would make it easy to assume that the pulsed output produces more light, however I believe this isnt the whole truth. Seems too good to be true.

My point is that the that the datasheet shows that the lumens efficiency is LOWER with the 22.5A Pulsed 50%, than the 13A constant, consistent with what is expected, a 10% efficiency drop. So im not sure how the graph works...

I have no definitive solution to this puzzle.

Wouldn't it just be that the 22.5A pulse is outside the optimal efficient range of the circuit?