TIR Lens Design Help

[Optical Inferno]

I'm having a bit of trouble designing a standard TIR lens system (first real attempt actually). Is there anybody out there who has had experience designing these lenses for illumination, that could give me some pointers or general guidelines?

The design itself is a simple revolved TIR optic like ones that are available from Ledil, Carlco, etc.

I have a working Solidworks (SP1) model that I can share, and I use Optisworks as my simulation software.

Thanks

 

[bshanahan14rulz]

Generally, I'd say stick to what's available off the shelf if it's a simple design. Once you get past the air cavity and reflecting surface, it seems that many gains come from recalculating the front refracting face, which ends up getting all wavy-looking. What problem are you runnign into?

I got to play with a beta of LEDOpticsDesign that was pretty fun, helped me learn a lot. Helped me learn that I should leave TIRs alone and stick to off-the-shelf ones. You may explain your situation and see if you can't get a trial or something. It's a relatively small company, but they've just recently become commercial, selling their software.

 

[Optical Inferno]

I can do the calculations and create the lens...that is relatively easy. The problems that I'm having is more not knowing what the internal ray trace should be, or what the external ray trace should be to give a homogenous output.

Should the output be arranged such that the rays exit at the desired angle or should they be an equally spread "fan" over the desired total angle? Should the internal rays be collimated or I guess that can be corrected at the final surface.

I've tried the first method by having the rays exit the lens at the desired angle without fanning, and the RP of the lens was not at all what I expected. My next attempt will be to do the fan method. If anybody has any advice or could point me in the right direction, it would be appreciated.

 

[bshanahan14rulz]

I would like to preface this by saying that I'm just a hobbyist, and most of what I say is just stuff I've thought about. I'm really just hoping to flex your brain a bit while we wait for someone more knowledgeable than you to arrive.

You are using an LED ray source? I would guess that you wouldn't really look at how the rays exit the TIR, but rather how they would project at a couple of distances might be easier? Ideally, a collimator would have all exiting rays parallel. Of course, this isn't possible with something as big as an LED as a source, and nobody said that all TIRs must only be narrow angle collimators...

The way Carclo seems to do it is by using the "reflector" surface to create a hotspot, and then capturing the spill using an appropriately sized aspheric element in the center, whose output is overlayed on the aforementioned reflected hotspot. I think since the diameter of the aspheric element is much smaller than the "reflector" surface of the TIR, the output is a larger spot, so it adds to the hotspot, but also produces a corona around the hotspot, which some people note as a square image of the LED.

Internal rays I would think just to make sure that whatever doesn't hit the reflecting surface hits a refractive surface. Khatod uses a convex lens on the inside, while carclo appears to use a convex or aspheric on the top side. I think a lot of the final output shaping is done by the front surface, either by microlensing or just texturing.

When looking at the rays exiting the lens, I think the farther they are from the center, the more likely to be accurately collimated. In different words, a higher percentage of rays from the LED will be within a smaller range of angles if they reflect off of the back surface farther from the source, while the rays that hit the reflective surface closer to the source will be able to reflect at a larger range of angles (because the image of the source from that point is bigger, so rays can come from a larger range of angles). SO, with that said, I think the best visual inspection would be to look at the rays exiting the lens farther from the center to make sure they are going where you told them to.

This is an interesting project, and I really really hope someone who actually knows what they are blabbering about will come in and set me straight, but maybe for the time being, this can get your brain to find the real solution ;-) I seem to remember member RA hand-crafted some glass TIRs a while back. I think he focused mainly on the reflective curve than on the front or inside features.

 

[Harold_B]

Optical Inferno - What add-in packages within OptisWorks are you using? I have been a SW user since 2000 and an OW user since 2005. My current setup is SW2012 with OW2012 running the Optical Design, Colorimetry Analyzer, Lighting Simulation, and Fluorescent Materials Editor packages on a Boxx 8550 Exxtreme. I have the 2013 licenses but can't install until I complete some client obligations.

For your system you will need decide how uniform the output needs to be (typically a percentage over the beam profile) and at what distance(s). Nothing is perfect at all distances especially with LED sources.

The TIR curvature is going to be somewhat fixed by material choice. The refractive index of the material and the wavelength of the LED will drive the curve. Making it uniform will require some manipulation once you get it going where you want it to go. The easiest thing is a diffuser on the output surface but you will lose the somewhat sharp edge from the collimation.

Not knowing which packages you are working with I can tell you the method I use to design optics for LEDs no matter if TIR, reflectors, or aspheric lenses.

Assuming the optical performance requirements are established: First is to determine the physical constraints. How big/small does the optic need to be? Are there features required to mount the optic like flanges or locators? Is there a preferred material?

What is the source? Most LED manufacturers supply a surface model and a rayset for download in SPEOS format for use in Optis. That is fine for a rough approximation. There are files available in the Optis Library for download also. The issue with raysets is that they do not take back reflected light into account in the simulation so the result is less accurate. We measure sample LEDs on a goniometer and in an integrating sphere to capture the output profile and spectrum. I model the LED right down to the chip and wire bonds. The simulation results for the LED are compared to the measured data and the normalized cross correlation and root mean square errors are calculated. The NCC should be greater than 99% and the RMS should be less than 5%. An accurate source model is critical. Another consideration for source modeling is color uniformity over the viewing axis. This has been a topic on the forum for the first gen XM-L for example. The only way to accurately predict the color uniformity is with a mass fluorescent material editor package. That way the uneven path length through the phosphor layer is accurately modeled.

Once the source is defined and modeled I use a point source to rough in the optic. This will only tell you where the rays are going to go in a general unquantified way. Once I have a rough system in place then I will use a rayset for the LED to get the system performance close to spec. The final design and confirmation is done with the complete source model and as many rays as time allows. I also use GoToMyPC so that I can run simulations from my iPad at home.

It is important to minimize the simulation error. If you select the Interactive Simulation pull down in the Optis tree under Simulations you can select Rays Without Error. Select "No" and run a ray trace. The remaining rays will show you where the errors in the model are located. Accurate material definitions, complete material and surface interface definitions, and consistent file locations are important.

There is an optimization package for non-sequential ray trace available that would allow you to define the output you desire and the system parameters that you prefer to change. I do not have that package. The Optical Design package however does have an optimization feature for sequential ray trace and is much less expensive.

I am not sure if what you are doing is an exercise or if you have a product in mind. If it is the later I can't stress enough how important the accuracy of the source is to the simulation results. We recently designed a TIR for a client and while the tool was being made the LED PCB supplier changed to another LED. It changed the output from a nice spot to a bright ring unless it was purposely de-focused! Fortunately it was just a demo. Here's one we did for Dow Corning to demonstrate their moldable optical silicone material: http://www.dowcorning.com/content/publishedlit/11-3371-01_Insert_b.pdf

Here's a few papers that might help:
http://www.suss-microoptics.com/med...v_Schreiber_LED_homogenizing_SPIE_5942-20.pdf
http://www.breault.com/resources/kbasePDF/wp_spie_032_axisymmetrical_concentrators.pdf
http://www-atom.fysik.lth.se/QI/las...t, Light Guide Techniques using LED lamps.pdf
http://www.innovationsinoptics.com/...ptimizing_High_Brightness_LED_Performance.pdf

 

[Optical Inferno]

Thanks for the help guys. I think I have a handle on it now. I was able to get a trial version of the software that bshanahan suggested, and was able to develop a working model. That software is pretty impressive for how much it costs. I also got in touch with a more local optical guy and they mentioned that doing rapid prototypes of optics in an SLA style, likely won't yield the results that I'm looking for.

Harold...I'm trying to create a revolved TIR that will confine a Cree XH-G to about 40degrees. This is the first TIR that I have had to design and likely won't be the last, so I'm looking for a general confirmation of the design process for the future. I only have 2013 LM1 for OW and 2013 for SW so I'm kinda limited on what I can do (no simulations, only ray traces).

Thanks for the help.

 

[Harold_B]

Any chance you will be at the SWUG Summit in Detroit tomorrow?

 

[Optical Inferno]

Any chance you will be at the SWUG Summit in Detroit tomorrow?

Unfortunately my predicament requires me to be here to discuss details with a local optical designer. Hopefully they will be able to fill in the gaps that I'm missing.

 

[Harold_B]

That is unfortunate but I can certainly understand. I'll be there and it would have been good to meet you. If you don't get the answers you need from the local OE feel free to get in touch. There's just so much I can do for free since we are an engineering consulting company but I can do a lot when it comes to coordinating with another Optis user since I am one of the owners. I can learn from another user that approaches the software from a different perspective so it's always good to exchange ideas. You might also keep an eye on the Optis site for events in the Detroit area or get yourself on the email list if you are not already. Good luck in your meeting.

 

[Optical Inferno]

OK...little update here and another question. We are going a different route with the design now, however we are still investigating TIR lens design. So thanks for the help guys in pointing me towards some helpful tidbits.

As for the question, does anybody know how TIR lenses are prototyped??? I've done SLA prototypes with no success, and I've contacted a plastic machining company that declined as well. Can anyone point me towards a company that can prototype these parts for a reasonable cost...

 

[Harold_B]

I'm assuming you have chosen PC or PMMA. In that case you should be able to do a Google search for plastic optics rapid prototype. Most likely you would be looking at a diamond turned optic which will take a few weeks (4-16) and can cost anywhere from $400 to $2000 each depending on size and complexity. You might get it for less per part if you have a SLA made and then use that to make a silicone mold. Again, a rapid prototype shop should be able to help you out.

Can you say anything about the new design direction?

 

[RoGuE_StreaK]

I too have had extremely little luck finding any info on possibilities for small run optics. To my way of thinking a "simple" lens (as opposed to TIR, I'll get back to that) should be reasonably well replicable using high-definition silicone and PMMA, but I've found maybe two blogs in the entire interweb where people have done this. For "good enough" custom lenses, I can't see why a laboriously polished rapid-prototyped simple lens can't be replicated for SFA, if you have the moulding materials on hand?

As for TIR, the "T" and the "R" bits I can see as an issue with the above method; you'd have to get a really good finish on the reflective walls otherwise they ain't going to reflect, which as Harold says probably means diamond-turned is the only viable way to go. Maybe there's a university out there somewhere who'd be willing to churn a surface out for not much cost?
To me (™) the "easiest" way of machining/moulding it would be a three-piece mould;
(a) turn the outer reflective curve from solid stock; silicone mould is taken from it's outer surface
(b) turn the internal "cavity/lens" as an inverse, ie. a solid post; use this as a plunge tool, or take a mould and cast a new plunge tool
(c) turn the front face lens (if it's turnable), and take a silicone mould off it. This bit could be made to chop'n'change, so you could "simply" replace this silicone face to give a new output profile (eg. frosted)
Of course this is all completely theoretical, I do have a silicone mould taken off a cheapie TIR sitting around somewhere but I never got around to trying to cast it; still have to get my vacuum degassing rig set up (among other things)

If the above description was clear as mud, I might get around to drawing some diagrams sometime next week

 

[Harold_B]

Cost of tooling and / or a diamond turned prototypes is one of the down sides of PC or PMMA optics. Here are a lot of advantages but if it is an option I highly recommend looking into using the Dow Corning injection moldable optical silicone(s). There are several advantages, one being the low cost of prototype tooling. If the optic is smaller than a fist then tooling can be had for a couple thousand dollars or possibly less depending on complexity. Then a few hundred or more parts can be made costing a couple dollars or so each. That would provide engineering samples and preproduction parts until a production tool is ready. A bonus is that several geometries are possible with silicone that are not possible with plastics. And no, I don't work for D-C.

 

[bshanahan14rulz]

All the more reason why I'm amazed that Ra simply hand-made and hand-polished one out of glass. I still think I must have missed a crucial piece of info in his threads...

 

[SemiMan]

FYI,

Got an email recently that reminded me of a company that does diamond turning. I don't remember them being cheap.

http://www.gsoptics.com/

Semiman

 

[RoGuE_StreaK]

I'm sure I came across an Australian university's site that mentioned they had a diamond lathe lying about for optical turning, but I'm buggered if I can find it now (don't remember the name, and can't find anything in my history). Would be interesting to see if such places would do one-off turning for next-to-nothing, as they aren't trying to recoup the machine costs, and it's often sitting unused for most of the time. For example in trying to find the site now I came across another uni that seemed to indicate you could hire the laser cutter for $10 for half an hour!

If turned metal was polished and buffed, could the finish be "good enough" for TIR properties? I guess a related question could be if a metal reflector was turned, polished and buffed, would it outperform a TIR whose mould has been taken from the same degree of finish? Obviously the reflector and the TIR can't use the same profile as there's refraction in the TIR

Sorry if this is getting too far into the "hijack" zone

 

[WmArnold1]

Fwiw, I can't find anyplace to prototype my hobby-grade TIR either. Is this a club? :) My apologies for somewhat hijacking this thread too..

Quickparts recently eMailed me to say they do low-volume injection molding now. (ref) But, I haven't contacted them yet and I don't know if they offer Dow Corning's moldable optical silicone, COC, or whatever.. (beware; Polycarbonates melt around 155°C)

Backstory: the great Ra motivated me with his posts ( #1, #2, & #3 <== best ) to craft a TIR to keep my basic spot & spill geometry, but, direct most of the Lumens into the spot-cone. IMHO, I'm trying to improve lights like the ZebraLight S6330 that spill an excessive "wall of light". So much so, that one's eyesight is compromised for seeing distant, dimly-lit, items. (i.e; forget ANSI 0.25 Lux; you'll be lucky to process 5 Lux at 100 ft because your pupils & retinas adjust to the nearby brightness!)

To address RoGuE StreaK's question: No; reflectors cannot collimate or reproduce TIR beam-shaping flexibility outside of the spot, unless, the TIR designer is intentionally trying to mimic the poorer options one has when using reflectors. In a nutshell; reflectors have two basic parameters: spot-cone angle and spill-cone angle. Further, the spill-cone angle also determines what percentage of the typical spatial distribution is reflected into the spot-cone. And, most disturbingly for me anyway, spill-cone beam-pattern is simply the raw emitter spatial distribution. (see graphic and my meager attempt to improve spot/spill Lumen ratio for more throw and a dimmer "wall of light")

 

[WmArnold1]

... To me (™) the "easiest" way of machining/moulding it would be a three-piece mould;

(a) turn the outer reflective curve from solid stock; silicone mould is taken from it's outer surface
(b) turn the internal "cavity/lens" as an inverse, ie. a solid post; use this as a plunge tool, or take a mould and cast a new plunge tool
(c) turn the front face lens (if it's turnable), and take a silicone mould off it. This bit could be made to chop'n'change, so you could "simply" replace this silicone face to give a new output profile (eg. frosted)

Of course this is all completely theoretical, I do have a silicone mould taken off a cheapie TIR sitting around somewhere but I never got around to trying to cast it; still have to get my vacuum degassing rig set up (among other things)

If the above description was clear as mud, I might get around to drawing some diagrams sometime next week

If you use something like my design, above; you can turn both the reflector & socket and close your mould with a flat plate. YMMV

 

[Optical Inferno]

We ended up using the software that bshanahan suggested early on in this thread to create a lens that we later modified in Solidworks. Due to time constraints on this project it was the only course of action that we had, but I would still like to revisit it later.

We are however, still using the SLA rapid prototype process to evaluate the lens. As it turns out, this method can be used, to a degree, to verify the lens performance. We were able to assess the fill of the lens and the general final spread, using rudimentary systems, to ultimately arrive at a more informed decision then just relying on Optisworks software.

On a related note, if anyone knows of an optical engineer or engineering firm that designs TIR lenses, that would be willing to arrange a training seminar, I would still like to learn how to design these lenses. Perhaps if enough of us in this community are willing to get together on this, we could do a group web seminar and spread the cost out...

 

[Harold_B]

I can understand where you are underwhelmed by the results from OptisWorks if you cannot quantify the simulation output. The LM2 version with the Optical Design add-in is a very powerful software and will save a lot of money in the long run over making prototypes and testing then tweaking the design. Another consideration will be that the native bspline surfaces in SolidWorks are not very accurate when it comes to optical quality surfaces. That is where the add-in comes into play, especially when it comes to conic sections. As for rapid prototypes, I did receive an email from these folks (I have not used them, but they have optically clear material): http://www.versadyne.net/Index.cfm You might have a hard time getting a consultant to train you to design optics since you are basically asking them to train you to take away their business. A local college course would be a better path.