LED optics design

[Siwastaja]

Hi,

I've been tasked (or, to be fair, I have tasked myself ) with designing some LED optics, since off-the-shelf LEDs cannot provide the beam shape we need.

The requirements are not exact rocket science, but since I have never done this and Googling gives mostly noise, I need to ask, how do you do this?

Since this is a one-off design in a small startup, we cannot buy expensive software, and we cannot outsource the design work for a high-cost optical house, especially since our requirements are flexible and easy to satisfy, so nothing special required. Is there any free software for optical simulation you have found useful?

I have two options for the basic light source, a 150 degree LED emitter (http://www.osram-os.com/Graphics/XPic7/00204353_0.pdf), or a 90 degree version with a longer "lens". I guess the 150 degree version is a better starting point, since I can get my own optics closer to the emitter?

I need to illuminate a rectangle approx. 90 deg x 40 deg, doesn't need to be super uniform. It's ok if the middle area receives double or triple the light compared to the corners, but I want to avoid wasting led power in stray light outside this rectangle. Simple led optics producing 40 deg round shape would be perfectly acceptable, since I can aim multiple emitters to different directions (going to use at least 5-6 emitters anyway). If 30, 35, 40 or 45 deg emitters were commercially available at required power, price, efficiency balance, we would simply use them and ignore secondary lenses, but they are not available.

But, better than 40 deg round, I imagine that the optics could be designed as an anamorphic lens, so that the LED module could be manufactured as a single, straight PCB, without need to aim individual emitters.

In fact, once upon a time, I needed a light source for a line scanner, and only had LEDs with standard 150 deg round radiation pattern - so I hand-filed "1D" lenses from PMMA, and it was a great success, giving a bright collimated line out of any round light source.

But I have no idea how to properly design and simulate these, so that they could be CNC milled for prototypes, later injection molded (from PMMA).

Any ideas and pointers are welcome - you could even suggest a commercial LED lens manufacturer, as long it's low-cost, allowing interactive design, since we have no money for the typical "outsource to expert designer, 200 hours at $200/hour, get a product that perfectly matches the specs, even if the specs were not what you actually wanted since you didn't know" flow.

[Rerouter]

Try sketching up where the board is in relation to where the light source needs to be applied,

Automotive LED backlights so far sound quite similar to what your trying to accomplish, in those cases they use leds spaced apart about a cm along the longest edges, and just drive them in series with a boost constant current driver to keep the brightness mostly uniform. they then use a square of acrylic that they shine the LED light in to, and a white diffuse reflector adhered to the bottom side of it to bounce the light forward over a wide veiwing angle.

[ajb]

If you can approximate a point source, then you can start with some fairly basic lens equations to approximate the required curvature--since you appear to be using a single wavelength, you won't need to worry about making the lens achromatic, which helps.

But in the real world, high power LEDs aren't exactly point sources.  The active area can be fairly large relative to the optic, so you have to consider the light an area source, which will not have uniform emission characteristics with respect to angle over that area.  This can be further complicated by the encapsulant, which will have a different index from the die, which can complicate the emission model.  Not rocket science, but also not so simple. 

If you really want to do this yourself, you might use an off-the-shelf optic designed for your LED to get a head start.  Find one that's close to your needs, or maybe a few of different beam angles, and look at how you might adapt and tweak the design.  If you have the ability to iterate on a lens design, you might be able to home in on something empirically.  Or, even better, maybe you can use the existing optic as-is and add another lens or two afterward to correct the output to your desired shape. 

[kosine]

Well, it's as about as quick and crude as I could be bothered with, but just shining a LED though a couple of blocks of acrylic seems to change the shape without affecting the brightness too much. (Edges were given a sand and polish.)

You might be able to laser-cut some flat sheet to find a shape that works best. There's bound to be some loss of light whatever you do (IR according to the datasheet), but maybe you can get it to work. I doubt an injection moulded light-guide would give you much improvement.

[Swemarv]

This might do the job..https://www.osram-os.com/osram_os/en/products/product-catalog/infrared-emitters%2c-detectors-andsensors/infrared-emitters/high-power-emitter-gt500mw/emitters-with-850nm/sfh-4783/index.jsp

[Corporate666]

Use this:

http://www.bmoptics.com.tw/EN/productsPages.php?id=4254

I am not sure if your specific Osram LED is specified for this particular lens, but finding an off-the-shelf lens that is exactly what you need is so rare for an application like yours, that I would change the LED to one that works with this lens if need be.

Or you could start with something like this:

https://www.digikey.com/product-detail/en/ledil/FS15847_FLORENCE-3R-IP-O/711-1402-ND/6674591

and cut it to accommodate a single LED.

If you want a sharp cutoff at 90 and 40 degrees, you would need some sort of secondary shade spaced off the emitter/lens somewhat, because the LED isn't going to be a point source of light and therefore won't 'behave' across the curve of the full lens.  You could probably achieve the same with additional optics before or after the main lens, but that's getting into the territory of expensive custom setups which you said you don't want to get into.

[Siwastaja]

Thanks,

kosine, looks interesting, although I suspect that the optical efficiency is quite poor (similar to using black plastic lens hood) but need to test this out.

Swemarv:
Was considering that LED, but it's not for new designs. This is not a one-off project, so can't use it.

Re changing the LED:
I have found only about 3-4 suitable LEDs. I need a combination of high efficiency, high power, low price and fast rise/fall times, so basically there are not so many to choose from. Osram has some of the best. In any case, the options seem to be 90 deg or 150 deg, nothing else on the market satisfying power and price specs.

Re "sharp cutout"
The requirements here are quite flexible. I'm perfectly happy if I have 50% light output at the edges, and 30% at the far corners, compared to the middle. So similar to the soft cutout of most off-the-shelf 90 degree LED radiation pattern, but just squeezed to 40 deg in one of the axis, is OK. I'm going to use a secondary black plastic lens hood to absorb stray light anyway. I just don't want to waste 50% of the LED power in this hood, so some pattern shaping is needed.


But price is important, this is going to be manufactured in masses, and I know that it costs about $2000 NRE + about $0.10 each to injection mold these lenses - so I can't buy off-the-shelf $2 lenses with 2000% markup.

[Swemarv]

Have you looked at the Marubeni LEDs

http://tech-led.com/product/ir-nir-leds/smbb/

You may also have a look at Citizen and Stanley, they also make similar LEDs

[Swemarv]

Have you looked at the Marubeni LEDs

http://tech-led.com/product/ir-nir-leds/smbb/

You may also have a look at Citizen and Stanley, they also make similar LEDs

Princeton optronics makes vcsels with difffractive optics that gives a 45x60 square, realy nice.
http://www.princetonoptronics.com/product/large-divergence-vcsel-array-modules-1w-850nm/

Skickat från min D5803 via Tapatalk

[Siwastaja]

Princeton optronics makes vcsels with difffractive optics that gives a 45x60 square, realy nice.
http://www.princetonoptronics.com/product/large-divergence-vcsel-array-modules-1w-850nm/

VCSELs are very fancy, and really the best tool for the job, but it's still in a totally different price segment, at least two orders of magnitude too high.

I'm budgetting $7 for 10W peak optical power, which I get with four OSRAM LEDs. Add $1 for the array of four lenses, which should be easily achievable through injection molding at batch size = 10000pcs.

I think the best way would be to try out some commercial lens samples to get a "feeling" of the required shapes/dimensions, then 3D model some tests, CNC mill / hand polish PMMA prototypes, actually test&measure, since it appears there are no free / open source tools even for basic optical simulation / ray tracing.

[Swemarv]

Princeton optronics makes vcsels with difffractive optics that gives a 45x60 square, realy nice.
http://www.princetonoptronics.com/product/large-divergence-vcsel-array-modules-1w-850nm/

VCSELs are very fancy, and really the best tool for the job, but it's still in a totally different price segment, at least two orders of magnitude too high.

I'm budgetting $7 for 10W peak optical power, which I get with four OSRAM LEDs. Add $1 for the array of four lenses, which should be easily achievable through injection molding at batch size = 10000pcs.

I think the best way would be to try out some commercial lens samples to get a "feeling" of the required shapes/dimensions, then 3D model some tests, CNC mill / hand polish PMMA prototypes, actually test&measure, since it appears there are no free / open source tools even for basic optical simulation / ray tracing.

Getting a custom lens molded for 1 usd in 10000 pc volume will be a challenge. I work in optronics and we are currently designing a custom lens that will be injection molded and will be produced in 1miljon volume, we are struggeling to get under 1 usd. Injection molding lenses is not a trivial task if you want to have any decent quallity of the lens.

Have you checked Ledil?  They make lenses for a number of different leds including osrams oslon packages.

One software that Iremember not beeing as expensive as ZEMAX or lighttools is OSLO, I dont have any recent prices but it could be worth checking.

Skickat från min D5803 via Tapatalk

[Corporate666]

VCSELs are very fancy, and really the best tool for the job, but it's still in a totally different price segment, at least two orders of magnitude too high.

I'm budgetting $7 for 10W peak optical power, which I get with four OSRAM LEDs. Add $1 for the array of four lenses, which should be easily achievable through injection molding at batch size = 10000pcs.

I think the best way would be to try out some commercial lens samples to get a "feeling" of the required shapes/dimensions, then 3D model some tests, CNC mill / hand polish PMMA prototypes, actually test&measure, since it appears there are no free / open source tools even for basic optical simulation / ray tracing.

The injection mold costs alone are going to be likely $20,000.  Molding lenses is tricky - you have extremely narrow dimensional tolerances, and the surface finish of the mold must be perfect.  It's not easy (or maybe even possible) to remanufacture a worn out mold, and they wear out quicker than normal molds because you have very little room for drifting dimensions/surface finish as the mold ages.

If you are expecting a custom mold *and* 10,000 parts for that $10k, I think that is far into the realm of "never gonna happen", unless you find a part that a manufacturer can customize for you without changing a mold.

The BM Optics lens I linked above is already 90 x 40 degrees, and they cost more like $0.30 or so or less in the quantities you're talking about.  It's not specified for your exact LED, but they do have lenses for the Oslon LED's.  I'd get in touch with them and see what it would take to get your needs met with either a custom lens or something off-the-shelf. 

[Someone]

since it appears there are no free / open source tools even for basic optical simulation / ray tracing.

One software that I remember not beeing as expensive as ZEMAX or lighttools is OSLO, I dont have any recent prices but it could be worth checking.

Blender is a little slow but for rough estimates its usable for ray tracing simulations where you don't care about dispersion, the challenge will be modelling the source with some level of accuracy.

[Siwastaja]

Thanks,

LEDIL quoted a price tad too high, even though it could have been worse.

Yeah, BM optics seem to have a few suitable lenses, thanks for the hint. Still waiting for their reply. I expect they are in the right price range, as the lenses look smaller and simpler than at LEDIL. It's very likely that they have a suitable off-the-shelf solution for acceptable price.

[Corporate666]

Thanks,

LEDIL quoted a price tad too high, even though it could have been worse.

Yeah, BM optics seem to have a few suitable lenses, thanks for the hint. Still waiting for their reply. I expect they are in the right price range, as the lenses look smaller and simpler than at LEDIL. It's very likely that they have a suitable off-the-shelf solution for acceptable price.

I'd be interested to hear what they come back to you with, if you're able to share.  I use their lenses quite a lot and I have been very happy.  They are much less expensive than Fraen/Ledil/Carclo/Polymer-Optics and others I've had quotes from.  I've had a couple of issues reaching their sales person on occasion but they were always super helpful when we got in touch.

[TerraHertz]

It might be worth making a poor-man's optical table, like at the end of this: http://everist.org/NobLog/20150628_LED_lighting_mk3.htm
If you need diffuser film, pull apart some old LCD screens. There's always an assortment of diffuser films in the back.
And the laser pack in laser printers has a variety of small and large cylindrical lenses.

These people have a wide range of odd shaped lenses.  They say they offer a design service too. I haven't bought from them, just saved the link for future need.
https://thoptics.en.alibaba.com/

Incidentally, if you use LED sources constructed with multiple LED chips, and your optical path is all lenses, you're going to end up projecting an image of the LED chip array. Whether in focus or out of focus, it will still be objectionable at the destination. Using diffusers early in the path is the only way to avoid this.

[f5r5e5d]

https://en.wikipedia.org/wiki/Nonimaging_optics

maybe some variation of the Compound Parabolic Concentrator, possibly deep drawn stamping?

some more ideas: http://www.archlighting.com/technology/leds-understanding-optical-performance_o

[Siwastaja]

I'd be interested to hear what they come back to you with, if you're able to share.

They quoted me $0.30 @ 100 pcs and $0.20 @ 10000 pcs. This is in the range I was expecting from such a simple plastic part.

[kalel]

I can't really recommend this as I don't know enough to judge if it can be used accurately. It's a tool for artists.

But I have here some experiments with LED models not crafted to any actual specification/size, but just to look "similar" to existing LED.

https://www.eevblog.com/forum/chat/some-led-renderings/msg1243451/#msg1243451




It uses Blender and Luxrender, so it would take a bit of getting used to, and I cannot say if the resulting images are accurate representations at all. In my case it was just for fun, and no specific requirements were necessary.
They are open source, however.

[cdev]

Excuse me if this is an exceedingly dumb question.. but I've never used a 3D printer..

Do people ever 3D print lenses? Using transparent plastic? Is it even possible without bubbles?

[LaserSteve]

As some one who has been down the plastic optics road, just call Fraen or another vendor and get it over with. The cost of an optical finish mold is shocking.  You need close to an A-1  (1-10 uM) Diamond Buff finish on the molds, and need to maintain that finish over the production run, while maintaining symmetry with a de-bubbled polymer  during injection.  Its tough, and you might find yourself going to casting rather then injection.  Due to the high cost of a mold, you need your shape right the first time.

 Most mold making shops have never done anything like your asking, so you need a local mold shop with a high level of expertise.

   In my case, going to a reflective optic was easier then designing our own double  TIR optic. We did start down the double TIR road.  Good modeling software started at 7,000$. We did not buy the software, the main reason being the quality of the LED model when ran on the demo version was crap in the far field.  However four years ago the available LED model files were crap in the far field, but OK for the near field (less then 2-3 meters) I hope that has changed now, but there is a big difference in models made from live scans on a 4 or 6 axis goniometer versus factory source models that are sometimes simply done in a modeling CAD.

Low cost reflective optics are not known for their ultra-uniform illumination, unless made in a mold made by electro-deposition or a very high end diamond tipped CNC used for optics.


Metalized plastic reflective optics are far easier to make then internal refraction based optics. Most areas of the country (Here in USA) have a vacuum metalizing shop within a day's driving distance.  At least for reflective we could try to CNC the proto optic out of a polycarbonate and polish it by hand using really high end sandpaper and chemical polishing using solvents.

The hand carve and polish method using PMMA sheets actually works fairly well for some refractive optics, provided you do not mind testing using one dimensional stripes. .  We laser cut our shapes and solvent  vapor polished them after a wet sanding.

CDEV:

As for 3D printer prototyping of transmissive optics..

Depends of the 3D printer, a high dollar Stratasys running Opticlear resin would be moderately useful, a tabletop less then 3000$ printer will not help you much.  Even then, the .004" layering with 100% fill is a strong detriment to evaluating your optic, the scattering will be very high from index of refraction variations during cooling.

For reflective optics, if your making a bicycle headlight you might get away with a ABS or similar material for protos.

Call Fraen... The part you need will not be on their website, but I can attest that trying to do ABCD matrices in Excel vis a vis calling a plastic optics professional  will set you way back when trying to start a production schedule.  Mold design via iteration is painful.

Steve

[cdev]

What about Blender?

[LaserSteve]

Run the Blender results against Oslo Lt For some simple lenses in the far field and see what you get. Otherwise I don't know, I've never used it.

Steve