Is PWM more efficient for driving LEDs? 3

[geekEE]

Does anyone know of any app notes or other technical info that compares driving LEDs with PWM vs constant current in regards to efficiency? I have heard that PWM can use less power for the same apparent brightness due to nonlinearities in brightness vs current and in the eye's perception of that brightness, but I haven't been able to find anything that verifies this beyond hearsay. If possible, I'd like to know quantitatively how the efficiency varies with duty cycle and pulse current.

The reason for this is that I'm trying to reduce "vampire" power in our products and it's hard to put a number on the power saved (if any). If I can't find any info, I'll probably have to just do a quick experiment.

Glenn

 

[itsmoked]

I know where that data is but I'm not sure how you can find it. No one, but NO one, ever did as much hard research on the photometry of LEDs than Hewlett Packard.

In all the versions of their data books they had in-depth tutorials on the subject. It was partially done because they made lots of LED numeric and alpha displays that you would normally multiplex(blink). They went into detail on how to drive them for the 'brightest look'. Blinking them was always going to look brighter than steady state because you could safely overdrive them for a briefer time and the eye would integrate and persist the image. If you blinked them too brightly the ON time would be reduced enough to cause the eye to see the blinking trails if the head was moving.

Anyway if you can get your hands on an old HP data book you'd have the info you're after.

Thoughts on your dilemma. Personally I think the way to solve your problem doesn't lie with PWM and the noise this may create. There are some insanely bright and efficient LEDs out there. Just find a very, very, bright one and barely drive it. Drive it with sub mA current.

Other thoughts;
Don't drive it very bright. Not many people want to see a bright light telling them their widget is on.

Blink a dim LED 2s off - 100ms on. That's good for a 95% energy savings. Ever notice a battery smoke detector? They blink the LED ON every minute for about 30ms I'd guess.

Next time I'm at the office I will look and see if I still have the HP data books. I pitched well over a thousand data books a while back not sure if I pitched my HP ones for this very subject.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[geekEE]

Thanks Keith, that bit of info led me to search with some different terms and I found a couple more things that people had written about this subject, although they disagree to some extent. The first page has a dead link to an Agilent page that seems like it would be perfect.

http://www2.whidbey.net/opto/LEDFAQ/The LED FAQ Pages.html#Q22

http://members.misty.com/don/ledp.html

Neither page is the quantitative answer that I was looking for, but it seems to depend on the type of LED that I'm using. Keith, if you manage to find the HP info, that would probably still be helpful. Thanks,

Glenn

 

[Skogsgurra]

I guess that the book you are looking for is the HP Optoelectronics Application Manual. It is a beatifully made book with ISBN 0-07-028605-1. A couple of hundred pr so pages in large format and covering everything from LED theory to optical filters and mechanical issues. A book of a kind they don't do any more.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[geekEE]

Thanks Skogs. Used copies of that book seem to be readily available and pretty inexpensive, so I'll pick up a copy.

Glenn

 

[VEBill]

It would be an easy one afternoon project to try the experiment, either with a 555, or a uC, or some test equipment.

 

[itsmoked]

Turns out I kept every HP data book I ever had. It is so sad that that company sold off their LED devision to make a fast buck. (idiots)

And I have the aforementioned HP Optoelectric Applications Handbook.

After an hour I am not actually finding the aspect of human reception verses luminance.

There is definitely an aspect of a brighter light being easier for a human to identify as apposed to a dimmer constant light. I suspect the issue with signal to noise ratio affect this out come.

The really big gain comes from the quantum efficiency of the actual LED material. Certain materials have non-linear,(exponential), efficiency improvement with greater current. So you actually get substantially more light out of the LED for power in. This coupled with the human vision system works well, IF!, it is one of the LED materials that go non-linear. (I believe green does)

As I mentioned earlier if you blink it you need to do it at 100Hz or faster. If the machinery the LEDs are on vibrates then you must blink the LEDs 5x the vibration frequency or you will get a blurry mess.

This is the actual data sheet right out of the book related to this:

http://www.secomtel.com/UpFilesPDF/PDF/Agilent/PDF_DOCS/OPTOELEC/APPNOTES/LAMPS/AN1005.PDF

Keith Cress
kcress -

http://www.flaminsystems.com

 

[VEBill]

I was once in a rented car (made by a company now teetering on the brink) that had some LEDs scattered around the dash. Is was disconcerting that every time I moved my head, the LEDs left a trail of dots in my peripheral vision.

 

[itsmoked]

Yes, that is ex tr em ly an no yi ng!!

I can see lights flickering that most people can't it's often very distracting.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

Hi Soked!

In my book, it is in section 5.2.2.4 "Operational Curves for Strobing an LED Device" (Weren't they just wonderful at HP - than...). Three diagrams showing pulse duration vs refresh rate and such things.

That section is followed by 5.2.4 "Relative Efficiency and Light Output" with still some information. The book is full of it.

I think the present HP company wouldn't sue me if I scan a few pages and put them here.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Skogsgurra]

Oops. Should be Smoked, of course.

Could have been an insult. But, considering your root beer, I guess it isn't

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[geekEE]

So far the info that I've found has been inconsistent. I've found some webpages that say that there is no difference. I've found some that say that there is only an advantage if the frequency is low (around 60Hz). Ugh. That could be really annoying if it were that low.

http://members.misty.com/don/ledp.html

http://techon.nikkeibp.co.jp/english/NEWS_EN/20080407/150114/

I've also been looking at the graphs on LED data sheets that show the luminosity vs current and so far the ones that I've seen show either a constant slope or a slight decrease in efficiency at higher currents.

I didn't realize that this was a subject that had so many different opinions!

By the way, Keith, you spent an hour looking through the HP book? Thanks! But, didn't you have anything better to do on a Saturday?

Glenn

 

[Skogsgurra]

And, pray, what better could there be?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[itsmoked]

Indeed.
You think I'd rather be trimming the top of my hedge??! 12ftH x 8ftW x 120ftL

Keith Cress
kcress -

http://www.flaminsystems.com

 

[MacGyverS2000]

Glenn,

It's not so much a difference of opinion so much as a lack of knowledge and dependence upon hearsay.

As Keith mentioned, some LED materials lend themselves to higher pulsed-current efficiency than others. The reds, oranges, and yellows usually fit this mold, whereas the blues, greens, and whites usually do not. Notice I use the word "usually" as there are some exceptions. This makes creating full-color displays that are pulsed-current efficient across all colors difficult to construct.

The human eye sees peak light intensity, rather than average, which is why pulsed-current triggering works so well.

As far as efficiency tables, I'm not sure I can be of much help on that one, I can only offer food for thought. Try pulsing an LED at about twice the rated current, but with a 10% duty cycle (1ms gates seem to be common among most manufacturers). You'll see a brighter light at 1/5th the average current draw.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[geekEE]

Keith, do you still have rats in your hedge? Maybe you can train them to graze on the top of the hedge so that you don't have to trim it.

Dan, I would like to avoid hearsay, but there seems to be quite a lack of direct reliable info. Most of the links that I posted were just personal webpages. I would have thought that LED manufacturers would have made this info available on their sites, but I haven't found any so far, although the out-of-print book from HP that I ordered sounds promising. If you have any source of reliable info, that would be great.

I certainly respect and appreciate the info that everyone has posted here. This subject now has me really curious. I'll have to program a micro and do some experimenting. I'll order some LEDs with known characteristics in addition to my junk drawer of mystery LEDs.

Glenn

 

[MacGyverS2000]

Glenn,

Send me an email (see my sig). I have a good book or two that will give you the necessary "proof", though it won't include efficiency tables. Consider the info a "trade secret" ;-)

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

geekEE; Rats? Yes. But now that I attempted to take back some of our sidewalk there are NO leaves on the front of the hedge so you can see the panicked looks on their tiny faces as you walk along the sidewalk. I try to train them with 1000fps lead. They don't train well. Rather unruly.

Anyway, I think you are now onto the right track. Just experiment. It doesn't matter what a chart sez if you don't like the result.

Measure now, screw around until you like the result, measure again. The biggest savings will be with the most efficient LED.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

Manufacturers are generally loath to do that, since their reliability numbers are based on steady state conditions. Anything that they publish that suggests a more aggressive drive will result in more dead chips, and more tech support, and more warranty claims.

TTFN

FAQ731-376

 

[treez]

http://www.eng-tips.com/viewthread.cfm?qid=172223

...this post investigated PWM 'ing LEDs