Design of a maximum-efficacy general purpose LED light

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
With so many optics buffs on this forum I'd like to ask for some input on designing a LED replacement for my room lighting. Situation sketch:
- I'm an power electronics design engineer, I design AC mains stuff for a living. I'm designing this from scratch, no premade anything. I have access to all the large distributors.
- My 10x10' room is currently adequately lit by a 58W round fluoro tube http://www.amazon.com/dp/B003COHRTS/?tag=cpf0b6-20
- Typically this tube is run at about 2000 lumens, it's rarely run at the full 5klm, but it does need to be able to do this occasionally
- I want to replace this with basically the lowest power alternative possible
- No direct point sources, I like my area lighting
- 5000K CCT is preferred
- This is a hobby project, I want to see how far I can push things. But that also means funds are limited. Sub-$50 in parts.

I am currently thinking of doing something like this:
- Take 6 Cree XM-L2 Neutral White 5000K T6 LEDs
- Stick them on some aluminum profile
- Hang this from the ceiling with LEDs facing towards the (white) ceiling
- Try and find the most efficient AC LED driver circuit to drive this at 700mA nominal (2klumen), 2A max (5klumen)

This way:
- I can benefit from my white latex painted ceiling to do all the area lighting for me, I don't have to worry about optics
- I don't have to worry about thermals; the light is free-hanging and has maximum passive airflow
- I can probably get something near 150lm/W WPE nominal, and still in excess of 115lm/W with the lights on max.

But I'm really not sure of this. This is the first time I'm really attempting to attack general purpose lighting. I'm doing assumptions, like that white paint is a really good reflector (is it?), that this will create enough of an area light (will it?), etc. I'm also not at all informed about the different LED drivers around. I want to go for maximum WPE, so going for a nonisolated buck converter is probably my best bet. Something like NXP's SSL2109T seems like a nice choice. Are there better drivers for this application? Maybe even dimmable drivers that still retain 90+ % efficiency?

I'm also open to suggestions for alternative LEDs. Currently I only have Cree XM-L in my database as it seems to get the best efficacy, but if the datasheets are good enough I can add in any other make and model. I use this calculator to get efficacy and efficiency properly calculated: http://www.enijssen.com/publish/XM-L2 calculator.html (I also advertised this link in the other topic).

Of course, remote phosphor is also something I'm happy to work with, but I have no idea where to get things like that and how to design it. If it's doable for an electrical engineer to do this in low quantities I'm all for it, but otherwie I'm tempted to say it's out of the question for now.
 
Last edited:

Steve K

Flashlight Enthusiast
Joined
Jun 10, 2002
Messages
2,786
Location
Peoria, IL
at the risk of taking some of the fun out of this, how about two of these Acrich2 modules?
http://www.digikey.com/product-detail/en/SMJE2V12W1P3-CA/897-1151-ND/3719827

I'm using 3 of the 8W modules in a table lamp, and am pretty pleased with it. If you are sensitive to 120Hz flicker, you might not like it. Personally, I don't notice the flicker.

Other than that, the lack of a switching power supply, PFC circuit, etc, makes it very reliable, compact, and low EMI.
 

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
That's cute, but I can do much better than a paltry 100 lm/W. Also, that doesn't quite answer my fixture discussion points; I'm very divided about how to make this into an area light without having tens of percents of my light absorbed by a diffuser or matte reflector.

I am extremely sensitive to flicker, this has to be photo/video-proof. I'm doing PCB photography almost every day. However, I'm pretty sure something can be done about that with some caps on the acriche driver - I've seen other people do that. Might still be an option, it's certainly something I forgot to take into account in looking at drivers.
 

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
Hm... I like the idea of having part of the light being emitted directly into the room instead of all of it being indirect. I might even be able to use the existing lamp fixture in the room, although it's pretty small to be an effective area light (it's only about a foot in diameter).

Good stuff to think about. Unfortunately we don't have McMaster-Carr in the Netherlands, but we do have very cheap aluminum profile dealers (like almost bulk material prices for extruded profiles). I'm going to browse around to see what I can find.
 

Barbarin

Flashlight Enthusiast
Joined
Jul 30, 2001
Messages
1,305
Location
Pamplona- NA- Spain
I would not use just six emitters, because you will be getting six noticeable shadows, and the uniformity will be poor. You want this light for working and you don't want shadows that might lead to stress of confusion.

I suggest you to use high power LED stripes, with high CRI (at least 80), and long "lighting devices" to avoid any shadows or minimize them. You want something for working and for this purpose and even more for photography the CRI and even light distribution is most important than the efficiency. What is more, the most efficient lighting systems are not the most confortable, high CRI is against high lm/watt.
If you are into electronics to be able to identify perfectly any color, character, rust, corrosion... is very important for you. So I would put my focus in color rendering index and uniformity, lack of shadows, than on the pure efficiency. Is your eye system efficiency what is important here, you accuracy, not the pure lm/watt. Think on this as an operating room, not as street lighting.

20120904143110160.jpg


I'm getting now 2500 lm at 24 Watt per meter in some units and 1300 lm/12 watt. In addition you can dim them with a simple remote controller you can find on ebay to 50% and 25% power. Don use the dimming for the pics, though. I would go for 4000ºK, neutral, not cool.

My two cents.

Javier
 
Last edited:

AnAppleSnail

Flashlight Enthusiast
Joined
Aug 21, 2009
Messages
4,200
Location
South Hill, VA
- Hang this from the ceiling with LEDs facing towards the (white) ceiling

I would take your efficacy with a grain of salt, then. Take a moment and swipe a clean white cloth across your doorjambs. That's what will coat the upward-facing surface of your lamp. You'll need to wash your lamp a lot to maintain output. Secondly, while ceiling-bounce is efficient, and makes a good area light, most white paint is 80-90% reflective when clean.

Otherwise, it's a decent project. Ceiling-mounting about a kg of aluminum would require some work in America, but houses are built differently in the Netherlands.
 

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
@Barbarin: It's not particularly hard to get good enough color rendering under neutral white. LEDs aren't fluorescent lights; they don't have entire missing swaths of wavelengths. That being said, my limit is at 5000K CCT with good phosphors, both in terms of color rendering and work lighting.

As for the shadowing: that's why I want area lights. Just strips aren't going to cut it, they need to be 2-dimensional area lights. Strips have other advantages, but not being able to properly design them to hit a certain efficacy is a big drawback. I'm doing science here :p I'm not going to buy anything that I can't have a datasheet with spectra and performance data in them.

AnAppleSnail: Now that's what I wanted to know. So white paint is probably not good enough. Is there a way to improve this? I know that LCD retroreflectors are specced to be 99%+ guaranteed albedo with nigh-on zero degradation over life.
 

SemiMan

Banned
Joined
Jan 13, 2005
Messages
3,899
Here is a thought. Since this will never see the inside of a sphere, you will have no idea of the absolute efficiency. Concentrate on quality not unobtainium or no idea what it really is ism :)

You need to be a distance from the ceiling such that it will not shade the reflection too much.

If you are doing non isolated buck then really need to shield the LEDs for safety.
 

Barbarin

Flashlight Enthusiast
Joined
Jul 30, 2001
Messages
1,305
Location
Pamplona- NA- Spain
@Barbarin: It's not particularly hard to get good enough color rendering under neutral white. LEDs aren't fluorescent lights; they don't have entire missing swaths of wavelengths. That being said, my limit is at 5000K CCT with good phosphors, both in terms of color rendering and work lighting.

As for the shadowing: that's why I want area lights. Just strips aren't going to cut it, they need to be 2-dimensional area lights. Strips have other advantages, but not being able to properly design them to hit a certain efficacy is a big drawback. I'm doing science here :p I'm not going to buy anything that I can't have a datasheet with spectra and performance data in them.

AnAppleSnail: Now that's what I wanted to know. So white paint is probably not good enough. Is there a way to improve this? I know that LCD retroreflectors are specced to be 99%+ guaranteed albedo with nigh-on zero degradation over life.

Well, high efficiency CREE LED are around 70 CRI, which is under regulation requirements in Europe for indoor working.

Regarding LED stripes, I have made few of them for demanding applications, and I do have an integrating sphere I use almost daily, so I know the real numbers. (I'm not selling them, anyway). But you will avoid the shadows depending on the number of lines and position of them you use.

Here you are an example of the reports I get. This case was a 200 mm disc retrofit, with 280 Refond 4014 LEDs. It is a 3000K, but you get the idea.




Regarding white paint... you can try using BaSO4 based paint, or MACPET from Furukawa Electric. I have used it on many products with EXCELENT results. My choice in your case would be a faced up LED stripe against a half pipe covered with MACPET. That would be near the optimal you can get.
 
Last edited:

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
@SemiMan: this will probably be tested with an integrating sphere. I'm not joking around.

@Barbarin: Regarding CRI, there's two things at work that I at least treat as different problems:
- Missing spectrum
- Different proportions in the spectrum

CRI is literally nothing but the weighted difference between a blackbody radiator and the spectrum under test. There is no reason why a blackbody radiator is the golden standard. Most of the day, sunlight has significantly lower CRI than 100, down to sub-70 in the evening. Incandescents are essentially the only usable light source with CRI 100, but they are (IMO) terrible light sources, with a CCT way too low to be comfortable and very low light output in general. In summary, just quoting CRI is getting nobody anywhere. It has to be viewed in context.

Fluorescent lighting suffers from actual gaps in color rendering: some colors show up as either completely other colors or as black under (older) fluorescent lamps. New phosphorus coatings have improved this, but there is still the big issue of marked mercury emission lines in the spectrum that mess up metal colors, etc.

LEDs have none of this. They have a smooth spectrum with no missing colors. Blue and red have higher spectral power densities than green, but this difference is not more than a couple dB at most. For the human eye, colors aren't rendered as crookedly as, say, under a clear-day sky in the shadow.

The low CRI of LEDs comes down to the green gap, mostly. I personally don't care about this. I have not found this to impact light quality nearly as much as all other lights I am used to using. My current lighting is fluorescent with about 72-CRI. Every LED is better than this. I find the amount and distribution of light to be much more of a mental drain than the spectrum having slightly more red in it. Because let's face it: high-CRI LEDs don't actually have a better overall spectrum, they just mix in more red and/or selectively block the blue content. There is no magical extra green - which there should be for actual better color.

So this is why I am mostly dismissive of 'light quality'-comments. I know - to a certain extent, don't even try calling me an expert - what light quality means and what CRI does. I can judge from the datasheet spectra what the light quality will be.

edit: almost forgot, but thanks about the comment on paint. I think I'm going to try to measure the reflectivity of my ceiling with an 18% grey reference card and my dSLR and see how bad it is. I can live with 90 or 95% reflectivity if it turns out to be that.
 

SemiMan

Banned
Joined
Jan 13, 2005
Messages
3,899
For high reflectivity and highly diffusing surfaces you might consider these:
http://furukawa.co.jp/mcpet/english/index.htm
http://whiteoptics.com

You could paint the ceiling above the light with highly reflective/diffuse white paint. Look up Akzo Nobel paints that are designed for the interiors of channel letter and other signs. Reflectivity properly applied is >95% and diffusion is very high.

Unless you plan to sell it (which you will sell few of -- Kickstarter?), then why bother with the cost of sphere time? Enjoy what you make, but going over board takes the fun out, and while everyone things they are building a better widget, few are.

Semiman
 

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
Nah, it's not for sale, but I do more LED stuff on contract and I tend to use my 'free time' to develop stuff that might be of use later. There is no alternative for proper lumen measurements - an integrating sphere is the only possibility. If the resulting construction is practical in any way I'll probably release the schematics, layout, BOM and drawings open source (creative commons).

And there is no significantly higher design effort to this done properly compared to when I would just design something 'that works'. Most of my design time is research anyway. Designing and building the boards and assembling the mechanical stuff is nothing.

@Harold_B: One thing in the datasheets isn't clear to me: why is total reflectivity on some of the products more than 100%? I'm sure this is some quirk of how reflectivity is defined, because they don't produce light I imagine.
 
Last edited:

Hoop

Enlightened
Joined
Dec 18, 2011
Messages
316
Location
Spokane, WA
CRI only accounts for 8 color samples, of which RED [R9] is not included.... Most LED's are highly deficient in R9. Superior lighting grade LED's often list CRI values from R1-R15.
I would not use an led for interior lighting that 1) does not list the entire spectrum scores from R1-R15, and 2) does not score above 80 in all categories.
The readily available sharp zenigata's will achieve this. Yuji LED's, Bridgelux Decor, and xicato XSM Artist Series will as well. Couple these led's with a 90%+ efficient Meanwell or equivalent driver. Color quality is so much more important than absolute efficiency when your light source is already near or better than 100 l/w, in my opinion.

If you are going to ceiling bounce I would suggest using a glass or acrylic lens over the leds so that the led's will be protected from dust and from cleaning. Your heatsinks capability will be largely determined by its vertical surface area. This means you might need your fixtures to hang down quite a bit or extend out quite far in order to achieve adequate heatsinking.

If you aim the led's down, there are several diffusion products on the market which offer varying levels of diffusion vs transmission. Some diffusion products include Lucite Diakon Frost, Makrolon Lumen XT, PlexiGlas Diffuse, luminit LSD, and ACRYLITE Satinice. These are mostly only available in sheet. You can form these sheets into accurate domes and other shapes with a simple DIY vacuum forming setup. There are many instructional videos on youtube for DIY vacuum forming.

Edit: Khatod has a full range of protective lenses in both clear and diffuse which do not shape the beam in the "Universal Diffuser Optics" category.
 
Last edited:

Harold_B

Enlightened
Joined
Mar 10, 2011
Messages
410
mux - nope, 100% plus. Amazing stuff. Kidding. I'm not sure about the measurement methods that yield 100% plus (not the first time it's come up in published data) but for both when we've measured samples they've been around 98%.
 

SemiMan

Banned
Joined
Jan 13, 2005
Messages
3,899
There is no alternative for proper lumen measurements - an integrating sphere is the only possibility.

Unless you plan to put it in a sphere for thousands of hours, then lumen maintenance and a sphere are not coming into play. Lumen output from a sphere yes, but lumen maintenance can be done with cruder methods and since you are looking at in application lumen maintenance, room measurements would be suitable once it is running.

Of course you will want to look at board solder point temps and extrapolate to junction temps and compare against LM80.

Making the "most efficient" light though is mainly a purchasing exercise. Most efficient bare emitters you can buy, running at the optimum current for light output, coupled with the highest efficiency LED driver you can purchase or make. That is not really pushing state of the art. I would challenger you more on making a really good light with high output which is quite a bit harder.

Semiman
 

mux

Newly Enlightened
Joined
Dec 5, 2011
Messages
27
Location
Netherlands
Lumen maintenance can be done very accurately with just a dSLR. CMOS sensors in there are basically better than 1% accurate photon counters as long as you stay far enough away from noise. That's fine. But to get an actual proper absolute lumen measurement I don't see any way around the integrating sphere, so that's why I said that.

I'm also not really claiming to be pushing the state of the art with this project, I just want to experiment with... well, putting theory into practice. It's clear that it's theoretically possible to get much better efficacy than what's available on the market (even using the same emitters) through design. So... why not do that? And as this is my first real foray into that realm (all other LED projects I've done have not even been near general purpose lighting) I just want to make something that is of use to me personally. I'm not necessarily interested in solving a general question or problem.

Eventually, my garage lighting will need replacement as well. That's a good 600W of fluorescent lighting that needs to be replaced with at least 40klumen of LEDs. If this project proves successful, I'll come back to make a 'really good light with high output', promise!

Just out of interest, what is it that makes this problem harder to solve? Just economy, as in, it's not feasible to buy 500 XM-L2s and run them at 350mA to get better efficacy? Or is there something more at play?
 

AnAppleSnail

Flashlight Enthusiast
Joined
Aug 21, 2009
Messages
4,200
Location
South Hill, VA
Just out of interest, what is it that makes this problem harder to solve? Just economy, as in, it's not feasible to buy 500 XM-L2s and run them at 350mA to get better efficacy? Or is there something more at play?

Right-sizing. Why buy 500 XM-L2s at $5 each, hand-solder them to copper (I hope you vent your shop!), mount a few dozen stone of metal to your ceiling (oops! Dropped a section with $500 in LEDs on it), invest in a half kilowatt of controlled drivers... Just to stay too far from the LED's limits?

An XP-L gets about 200 lumens per watt at 350 mA, or 180 lumens per unit. There's no need to buy 230 XP-Ls, though. You can run them, within reason, at 2 amps each, on adequate heatsinks (60C heatsink metal is plenty hot, and well within rating). That gives you over 600 lumens each, only 70 LEDs. Much cheaper!
 

WhiteRabbit

Newly Enlightened
Joined
Sep 11, 2014
Messages
57
The last time I checked (5 years ago) untinted Behr Premium Plus Ultra White had one of the highest measured reflectance values for regular house paint. Luckily, it's not a very expensive paint. Not sure though if the formulation has changed since then due to zero VOC paint and what impact that might have had. IRRC, there may also be a high reflectance paint for use on a wall intended for movie projection.

I agree with the principle of bouncing the light around as much as you can by using the whitest paint you can. Floor color can also have a big effect, as that's the first surface most recessed lights hit. If it's dark, you're immediately wasting much of the light you make.
 
Last edited:
Top