Would PWM make an LED appear brighter than a lesser constant current equivalent?

In search of some brighter LED's for a solar application, I purchased this LED pack: http://www.amazon.com/dp/B0085MOXGC/?tag=cpf0b6-20
By measurement I have confirmed that each LED consumes 333ma at 3.4v. However, my solar power budget for the LED is at most 50ma at 3.4v. If I were to use PWM to limit the LED to 50ma at 3.4v (obviously diming it), would the LED appear brighter than running the same LED at a lesser constant current but using the same amount of watts? Are there any rules of thumb regarding this?

The reason I ask is that (it seems) all the highly efficient LED's (>100 lumens/watt) are higher wattage types, and so I'm wondering whether I can access the efficiency of higher powered LED's using PWM, or whether all their extra efficiency will evaporate if I use PWM. My objective is to get the highest brightness for a particular wattage (in this case, 3.4V x 50ma = 170mw). Or is there a different/better approach for achieving the objective? Would I be better off just finding whatever LED produces the most lumens running 50ma constant current at 3.4V?

As you expected, PWM does not give you an increase in efficiency. It is like running it at the higher current (just for short periods of time).

Those LEDs that you purchased are not very good or efficient.

What type of light are you trying to achieve? .. spot? ... area?

Semiman

I'd like to replace the LED's in some IKEA solar garden lights with ones that are brighter and more efficient. Presently 9 individual LED's on a string are drawing 450ma (50ma each). I aim to replace the original solar battery bank with something better, and I thought maybe advances in LED technology might allow me to snip off the existing through-hole type LED's and solder on better ones in their place. You may ask, "Why not just chuck the whole thing and buy something good?" Excellent question. Answer: I'm not sure anything good in this category actually exists that I can just buy off the shelf. So, this is more of a maker project, where the bollard-like lights will still look IKEA, but probably everything else about it will likelly be a totally new fabrication.

For instance, there are some Moonray's on Amazon that claim to be "25x brighter." The reviews seem to attest that they are at least brighter than what's typical. Presumably the performance improvement is mainly due to more efficient LED's? At least on the high end, LED efficiency has improved a lot over the past few years. Has any of that performance improvement trickled down to more basic LED's, like those used in solar garden lights? http://www.amazon.com/dp/B004IEBBIY/?tag=cpf0b6-20

Anyhow, I'm not even sure where to look for trustworthy, consolidated info regarding LED performance. When it comes to LED performance, much of even the quantitative performance info seems highly exaggerated (much like the claimed capacity ratings on certain rechargeable batteries such as Ultra**** seem audaciously false these days by claiming capacities of 5000mah for 18650 Lithium ion rechargeable batteries).

Cutting through the chaff of misinformation, the CREE MC-E emitter seems like it is perhaps one of the most efficient currently on the market. It produces 320 lumens at 350ma. So, are there any white emitters that can produce even 200 lumens/watt (or better) at 20-50ma and 3-3.4v? If so, it might be a good substitute for what's presently in the IKEA garden lights that I'm modding. If not, then although obviously running a higher performance, more efficient LED like the CREE MC-E or the like using PWM with less than 100% duty cycle would obviously be dimmer than at 100% duty cycle, would it still be proportionately brighter than more ordinary constant current 20-50ma LED's if the highly efficient LED was run at a 10% duty cycle (i.e. 35ma equivalent)? I'm guessing the answer is yes, because although the current vs. luminosity graph for the MC-E doesn't extend all the way down to 20-50ma, the plot for the current range that CRE charted appears to be very linear. However, that's just my guess. I'm hoping someone here actually knows and can comment. Anyone?

Maybe I should limit myself to buying only one brand of led's? That way, even if the datasheet info is exaggerated, at least maybe it will be possible to make comparisons as to relative brightness and efficiency within the same brand.

At low current levels (Compared to LED rating), PWM is less efficient than constant current.

Don't forget that the MC-E comes in a few flavors - With a forward voltage between 3.4v and 13.6v. If you compare 350mA @ 6.8v or 13.6v, the power consumed is a few watts.

The "most efficient" LED at high power levels is the XM-L2 right now. For your application, a Cree XB-D or XT-E will be appropriate. These are moderate-power LEDs that will last nearly forever in your planned application. There may be cheaper options, though.

In general, you might get a 50-80% output increase from using really choice LEDs. But even a crap LED will only be a few times dimmer than a super-amazing LED, at 50 mA. The path to higher brightness is higher power - And all the costs that come with that.

AnAppleSnail said:

In general, you might get a 50-80% output increase from using really choice LEDs. But even a crap LED will only be a few times dimmer than a super-amazing LED, at 50 mA.

Click to expand...

Thanks for the good info. In general, would running a larger number of lower current crap LED's outperform a single non-crap LED? I notice that in a way the CREE MC-E seems to operate on that principle by, in essence, incorporating 4 LED's into the package of a single LED. By doing so, it seems to get 50% more lumens/watt.

Yesterday I also received some 3.2v 20-30ma diffused white LED's from Amazon (http://www.amazon.com/dp/B0059H604O/?tag=cpf0b6-20). Would running two of those in parallel at 25ma each (equals 50ma total) likely outperform single instances of even the fancier CREE's (XB-D and XT-E) that you mentioned running at the same total amount of milliwatts?

WhiteRabbit said:

Thanks for the good info. In general, would running a larger number of lower current crap LED's outperform a single non-crap LED? I notice that in a way the CREE MC-E seems to operate on that principle by, in essence, incorporating 4 LED's into the package of a single LED. By doing so, it seems to get 50% more lumens/watt.

Yesterday I also received some 3.2v 20-30ma diffused white LED's from Amazon (http://www.amazon.com/dp/B0059H604O/?tag=cpf0b6-20). Would running two of those in parallel at 25ma each (equals 50ma total) likely outperform single instances of even the fancier CREE's (XB-D and XT-E) that you mentioned running at the same total amount of milliwatts?

Click to expand...

Well, there's some funny business going on here. There is no white LED in the world that reaches the efficiency Cree is implying on their datasheet. They are being VERY tricky with the way they are describing their testing. Have a look at the data sheet. Each single die has a forward voltage of 3.1V @ 0.35A. The whole package of all 4 dies together produces 320 lumens at the test current. So each LED consumes 3.1V @ 0.35A (about 1 watt) to produce 80 lumens. 80 Lumens per watt is much closer to the correct value. These 4 dies must be wired in series or parallel to create the whole MC-E package.

Running LEDs in parallel slightly increases output - Slightly. Since you have the Cree datasheet open, page 7 of my MC-E datasheet gives the Relative Intensity vs Current chart. If we look at 350mA=100%, and 175mA=55%, then two LEDS run at half-power each get about a 10% output boost. Odds are great that you can't perceive this difference - Cree's brightness bins are 7% wide!

Also, more LEDs changes the light distribution - Possibly making your light less useful than with the correct light distribution.

Edit: No, two of those wouldn't outperform an XB-D or an XT-E. These cost about $5 on a 20mm star. The little "bullet shaped" LEDs have a terrible problem: All the power you put in (50mA @ 3.4V = 0.2W) has to escape through insulating acrylic, or tiny pin-wires. The thermal resistance of these packages can be 200 C/W, so your little bullet-LED is going to be 34C hotter than the inside of your solar light. The good news is, solar lights only run at night. But 10C + 34C = 44C, a temperature at which the cheap plastic optic will degrade right over the LED. This blackened smear can be seen with high-power SMJLED mag-lite mods from 1998 or so. It directly cuts output of the LED, quite quickly.

Newark has a great Application Note about the MC-Es in general:

Click pdf.

AnAppleSnail said:

Don't forget that the MC-E comes in a few flavors - With a forward voltage between 3.4v and 13.6v. If you compare 350mA @ 6.8v or 13.6v, the power consumed is a few watts.

Click to expand...

Wasn't the claimed 303 lumens/watt at the 3.4v voltage though? http://www.cree.com/News-and-Events/Cree-News/Press-Releases/2014/March/300LPW-LED-barrier

AnAppleSnail said:

The little "bullet shaped" LEDs have a terrible problem: All the power you put in (50mA @ 3.4V = 0.2W) has to escape through insulating acrylic, or tiny pin-wires. The thermal resistance of these packages can be 200 C/W, so your little bullet-LED is going to be 34C hotter than the inside of your solar light. The good news is, solar lights only run at night. But 10C + 34C = 44C, a temperature at which the cheap plastic optic will degrade right over the LED. This blackened smear can be seen with high-power SMJLED mag-lite mods from 1998 or so. It directly cuts output of the LED, quite quickly.

Click to expand...

Thanks! You just saved me from what would have been a fatal error. For LED's with through-hole packaging, how best to get a fairly uniform light distribution? Use a clear straw hat? I guess it's implicit in your recommendation that I should just forget about through-holes and embrace LED's that are surface mounted onto aluminum disks?

BTW, my preceding post (immediately above) was cross-posted.

With an LED like a top binned xp-l at 50ma you will be close to 200lumens / watt but the same would be true of a bunch of good 5630 leds at low current.

Diffusing a through hole led reduces efficiency so only do if needed.

SemiMan said:

With an LED like a top binned xp-l at 50ma you will be close to 200lumens / watt but the same would be true of a bunch of good 5630 leds at low current.

Click to expand...

If that's true, then maybe the 5630's are a better fit for this application? These are the IKEA lights that I'm tasked with modding:



I have no particular attachment to them, but my wife does... :ironic: They do look better installed than the photo would seem to imply. They cost around $30/set when my wife bought them originally. IKEA no longer sells this particular model.

We have 3 sets, and the mod I'm contemplating would drive all 3 sets (9 lights) from a single solar bank consisting of four 2200mah 18650 Caldwell Lithium ion batteries (http://www.amazon.com/gp/product/B00IZT28QC/?tag=cpf0b6-20). So, if I had to use PWM to lower the duty cycle, I could do it from a single source.

Yesterday I did a drain test of the four fully charged 18650's in series using a buck converter that was only 73% efficient (as measured by one portapow premium before the buck and another one after the buck), and that configuration supported an actual load of 450ma at 3.3v (roughly 1.5 watts) for 12+ hours (I can attest that it was less than 16 hours, but I'm not sure exactly when between the 12 and 16 hour marks the protection circuit terminated the test. They still had 14+ volts at the 12 hour mark). When the LED issue is finalized, I'll see what sort of solar panel I'll need to back my way into so that the battery bank can be reliably recharged. To increase the battery cycle life, I'd like to operate them in the 20%-80% capacity range. So, tentatively that's the maximum energy budget that I'd like to find brighter LED's to fit within. If I can get everything working, there's a chance I might later switch to a lithium titanate battery chemistry, with the aim to making these garden lights a "once and done" project.

I'm not sure exactly what LED's the solar powered WakaWaka uses (http://www.amazon.com/dp/B008G6282I/?tag=cpf0b6-20), but the WakaWaka has a large number of Amazon reviews that average out strongly positive, and according to http://www.themodernsurvivalist.com/archives/3251 the WakaWaka uses two 0.5W LED's that are 120lumens/watt from Seoul Semicon (same manufacturer as the 5630). I guess that doesn't really prove anything per se, but it does give Seoul Semicon at least some amount of credibility in the realm of solar powered LED's.



http://heracolights.com/2014/03/10/3528-vs-5050-vs-5630-led-smd-diodes/ compares the brightness and wattages of the 3528, the 5050, and the 5630, and it looks as though either two 3528's or possibly one 5050 would likely fit the power budget and not require PWM to fit like the 5630 would.



So, getting back to the topic of the original post, I guess in this particular case the 5050 would be better than the 5630? Though the numbers presented in that article seem a bit loosey-goosey, if I grind the numbers using the high end of the given ranges they imply the 5050 might be about 10% higher lumens/watt than the 5060, and about 38% more lumens/watt than the 3528.

Continuing the monolog (please do jump in with comments at anytime): So, if that's right, I just need to find the 5050 in a warm white on a suitable platform. It's mostly on tape, which I have no experience with. Can I just cut them off into singles and light them from the tape individually? Or is there a minimum number per cut?

Otherwise, I'll need some other platform, maybe an metal disk or similar. Is there a term for that? The image below shows scale, but the plastic simulated bulb packaging around the metal mount would probably get in my way...



Is there a CREE that's comparable to its 60 mA @ 2.8-3.4 Volts at 60 mA @ 2.8-3.4 Volts?

After some digging I'm guessing the WakaWaka uses the MJT5630 (http://www.digikey.com/product-highlights/us/en/seoul-semiconductor-acrich-mjt-leds/3251). 110 lumens/watt. Nominally it's 20ma at 22v forward voltage, but the datasheet says it can be driven even at 5ma. So, I won't need to resort to PWM. Because of how I'll be configured, I could either boost to the higher voltage or, alternately, add more 18650 cells and just have a bigger reserve to carry through cloudy days and probably last longer overall. The MJT5630 doesn't seem to be very commonly available, but digikey and mouser have the dies. Is it especially difficult to solder SMD LED dies?

Wurth Electronics has what looks like a really good one: http://katalog.we-online.de/led/datasheet/158301227.pdf, and it looks like I could solder leads to the bottom of the SMT package (if I'm understanding the datasheet). Digikey and mouse carry it, but neither seem to stock it. :shakehead At 30ma and 3.2V, it produces 10 lumens of warm white light. As near as I can tell, that's best of class for 96 milliwatts. The top of its range is 15 lumens at 45ma. Digikey sells them at 41 cents each.

Actually, Wurth has another one that's even better fit for my power budget: http://katalog.we-online.de/led/datasheet/158302230.pdf At 3.2v and 50ma, it produces 17 lumens. Cost is 46 cents each. Unfortunately, it's not stocked either. I have no idea who does stock it, or if I have to order from the manufacturer in Germany. :shakehead

WhiteRabbit said:

but the plastic simulated bulb packaging around the metal mount would probably get in my way...

Click to expand...

I've no relevant input on the rest of the stuff here, but just noting that with those T10 5050s you can just straighten out the wires at the end and the plastic slides right off. Can't recall precisely, but I think there's typically a through-hole type resistor in there to bring the 12V(ish) to an appropriate voltage/current.
Might be worth checking out the multi-5050 T10s, often works out a much cheaper way of getting individual ones, and if you desolder the board joints they'll all be on their own little PCB (not sure if MCPCB or not)

RoGuE_StreaK said:

I've no relevant input on the rest of the stuff here, but just noting that with those T10 5050s you can just straighten out the wires at the end and the plastic slides right off. Can't recall precisely, but I think there's typically a through-hole type resistor in there to bring the 12V(ish) to an appropriate voltage/current.
Might be worth checking out the multi-5050 T10s, often works out a much cheaper way of getting individual ones, and if you desolder the board joints they'll all be on their own little PCB (not sure if MCPCB or not)

Click to expand...

Thanks for your input! I ordered some to try out.

Earlier I thought 5050 was a particular manufacturer's model number, but now I realize it refers just to the SMD dimensions (5mm x 5mm). Do they all perform more or less the same, regardless of die manufacturer? In product descriptions the die manufacturer is virtually never even mentioned, and that's leading me to wonder if it doesn't matter. Likewise, earlier in this thread SemiMan referred to "a bunch of 5630's" and not "a bunch of 5630's from Manufacturer XYZ."

Speaking of 5630's, this assemblage got really high reviews on Amazon:


http://www.amazon.com/gp/product/B00H07V3YS/?tag=cpf0b6-20

I'd have to run it at really low current to fit it into the power budget. Most of the current vs. "luminous intensity" curves I've seen have been strikingly linear all the way down to zero current. The larger number, even if at lower current, would obviously help in in spreading the light around. Generally speaking, would the aggregate lumens emitted equal the lumens emitted by a single die at the same total power? Or would it be less?