Characterization of optic and LED combinations (S2+WF1, S21A+WF1)

I will be characterizing various optic and LED combinations to determine the focusing characteristics and optimal centering gasket thickness. I will update this thread with the tests as I do them.

At present this is not a test of the ultimate performance of the optic or LED, but this information may be added later.

About the test setup:

The focal distance test is conducted on a granite surface plate, with the sensor of an Extech EA30 Lux meter affixed 1.33 meters above it.
A plumb bob is used to center the lux meter sensor above the reflector.
The optic is affixed to a 12” Mitutoyo height gage.
The LED is clamped to a fan cooled CPU heatsink. (Intel D60188-001)
The LED is powered by a benchtop power supply at a constant current.

The test starts with the bottom of the optic being flush against the top surface of the MCPCB, i.e. there is no gap between them.
The height gage is moved upward in .001” increments and the lux value is recorded at each interval.
The results are graphed.

Any points on the graph which appear errant are measured again along with the neighboring points. The variance between the initial test values and the new ones is accounted for, e.g., if the new set of measurements are reading 0.03 lux above the former set, .03 lux is subtracted from the results in order to correct the errant value from the original measurement set.

Weights are measured with a Vertias S303 precision balance. Calibrated with Troemner 7017-0 UltraClass 100g weight.

Test #1:

Convoy S2 SMO reflector (not s2+) with an OSRAM KW CSLNM1.TG (1mm2) LED, also commonly known as WF1, mounted on a Convoy 20mm star. LED was installed by Convoy.

OSRAM KW CSLNM1.TG is a domeless 3030 LED with a 1mm x 1mm emitter area.
Measured width of package: 3.00mm x 3.00mm
Height: top surface measured to be .028” [.71mm] above the top face of the MCPCB. This may vary slightly depending on solder thickness.

S2 Reflector measurements:
Diameter: 0.813” [20.65mm]
Height: 0.755” [19.18mm]
Clear aperture: 0.725” [18.42mm] (approximately)
Vertex opening diameter: 0.197” [5.0mm]
Focal length of the parabola: .04” [1mm] (estimated)
Weight: 6.663 grams

LED was powered at 700mA constant current.
Lux meter sensor was affixed 1.33 meters above the reflector.
Tested September 4th, 2022.

( Here is a graph with more range in the positive direction resulting in a larger image size. )

Notes and observations:

The highest recorded lux value did not occur at a single point, but over a range of three points: 0.025”, 0.026”, and 0.027” [0.635mm, 0.66mm, 0.686mm].

The optimal gap between the bottom-most part of the reflector and the top surface of the MCPCB was .026” [.66mm].

+/-.002” [.05mm] from optimal focus results in a lux reduction of ~0.5%.
+/-.004” [.10mm] from optimal focus results in a lux reduction of ~1.5%.
+/-.006” [.15mm] from optimal focus results in a lux reduction of ~3.3%.
+/-.008” [.20mm] from optimal focus results in a lux reduction of ~6%.
+/-.010” [.25mm] from optimal focus results in a lux reduction of ~9%.

The spill angle is approximately 55 degrees.

Beam characteristics:

This reflector + LED combo is rather forgiving. The beam looks acceptable at all points, without major artifacts, even when the reflector is sitting flush against the MCPCB or when positioned too high above it.

White wall observations @ ~1 meter:

  • Height/gap of 0.000” [0.00mm] - The beam has three distinct parts: spill, a wide angle corona, and a hotspot. The hotspot has some minor artifacts and looks bit like a nebula, with slight color and brightness variation. There are no apparent rings or major issues.
  • Height/gap of 0.010” [0.25mm] - The angle of the corona decreases as the reflector becomes more focused. The hotspot now has no artifacts and appears nearly round. All parts of the beam have distinct boundaries with sharp transitions: spill, corona, hot spot.
  • Height/gap of 0.026” [0.66mm] - Median focus point. The transition between corona and the spill is smooth. The hotspot is still apparent but is not a distinct round spot with a sharp edge, the transition into the corona is softer.
  • Height/gap of 0.048” [1.22mm] - As the reflector height increases past the median focus point, the hotspot shrinks in size. At this height it converges and disappears. The center of the beam has a small yellowish tinted four point star shape in it.
  • Height/gap of 0.055” [1.27mm] - A dark spot begins to appear in center of beam. The dark spot enlarges in size as reflector height increases.

Centering ring design:

With the S2 reflector + WF1 combination, the top surface of the LED is basically flush with the bottom of the reflector. The top of the LED sits ~.012” [.3mm] below the vertex opening. This means any centering “tabs” or bosses that extend up into the reflector in order to center it will be extending above the LED and may block some light.

The test gaskets were machined out of G10, which I think may be a good gasket material in some situations. It is rigid, strong, non-compressible, and won’t melt. In most cases Nylon 66 is good enough though.

Model 1: This gasket resulted in a lux value of 99.02% of the optimal lux value. With the gasket in place, some light is being blocked.
Model 2: This gasket resulted in a lux value of 99.16% of the optimal lux value. A slight improvement.
Model 3: This gasket resulted in a lux value of 99.29% of the optimal lux value. Another slight improvement.
Model 4: This gasket resulted in a lux value of 100% of the optimal lux value. There is no loss on account of the centering ring. The tabs were eliminated entirely. The tapered geometry of the bottom part of the reflector is used to center the reflector. The LED pocket was made circular instead of square to improve manufacturability.

Test #2:

Convoy S21A SMO reflector with an OSRAM KW CSLNM1.TG (1mm2) LED, also commonly known as WF1, mounted on a Convoy 20mm star. LED was installed by Convoy.

S21A Reflector measurements:
Diameter: 0.910” [23.11mm]
Height: 0.452” [11.48mm]
Clear aperture: 0.787” [20.00mm] (approximately)
Vertex opening diameter: 0.276” [7.0mm]
Focal length of the parabola: .075” [1.9mm] (estimated)
Weight: 4.534 grams

LED was powered at 700mA constant current.
Lux meter sensor was affixed 1.33 meters above the reflector.
Tested September 18th, 2022.

Notes and observations:

The highest recorded lux value occurred at a single point, .017” [0.432mm].

This reflector seems to require more precision to focus than the S2 reflector. The resulting lux value (with the same test parameters) was less, even though it has a slightly larger clear aperture. The longer focal length of the S21A reflector makes for a shallower parabola, collecting less light than the S2 reflector.

The graph is less symmetrical than the one from the previous test. A test with the meter at a greater distance might make for more a symmetrical result.

+/-.002” [.05mm] from optimal focus results in a lux reduction of ~1.0%.
-.004” [.10mm] from optimal focus results in a lux reduction of ~4.3%.
+.004” [.10mm] from optimal focus results in a lux reduction of ~3.0%.
-.006” [.15mm] from optimal focus results in a lux reduction of ~11%.
+.006” [.15mm] from optimal focus results in a lux reduction of ~7%.
-.008” [.20mm] from optimal focus results in a lux reduction of ~20%.
+.008” [.20mm] from optimal focus results in a lux reduction of ~12%.

The spill angle is approximately 85 degrees.

Centering ring design:

I have not made it yet, but the following design should work well:

great info
I hope you get a chance to add a couple beam profile photos.

reflector height is the same w both LEDs, but the dedome makes a shadow ring in the spill of the beam:

to make the left beam look more like the right beam, do you think the reflector needs to move closer to the mcpcb?

> The optimal gap between the bottom-most part of the reflector and the top surface of the MCPCB is .026” [.66mm].

how tall is the LED?

checking whether the ideal height of the reflector, corresponds to the height of the LED dome.

Neat. It looks like a 3rd order polynomial would fit that curve pretty closely?

I’ve got something for you that could greatly reduce the amount of work needed to create these. My Synthetic Runtime project has expanded into a general purpose bluetooth data acquisition tool.

Its got support for 2 luxmeters so far: the UT383BT and (my personal favorite) WL81B. The WL81B is temperature compensated and has 0.1 lux resolution. Not quite as good as your EA31 but better than the typical luxmeter. Though you are measuring kilolux so that precision doesn’t matter.

I don’t have any BT calipers but the available <$100 models appear to act as a bluetooth keyboard instead of needed a special decoder.

Great idea, Hoop, and thanks for taking the time (and for what sounds like a proper test rig setup!). This will be a great reference and time-saver.

Could you share a photo of how you’re attaching to the gage arm?

Great thread Hoop!

One of the most important things in flashlight hobby.

A seemingly simple request, but my testing area has an unpainted wood ceiling with beams in the way. If I move everything into the house I could take ceiling photos. The granite surface plate is small so I can move it without much trouble. I think it’s worth doing to better characterize the optical behavior, especially since this kind of testing has not been shared on BLF before, as far as I know. Hopefully I can get a camera to portray the beam characteristics accurately.

I cannot answer this yet but it will hopefully become clear once I test a domed and dedomed LED of the same type with the same optic. It will be interesting to see the differences.

I measured the top surface of the OSRAM KW CSLNM1.TG to be .028” [.71mm] above the top face of the MCPCB. This suggests that the S2 reflector geometry terminates essentially at the focal plane, aka latus rectum.

I was not expecting the lux readings to be so symmetrical or for focus to be rather easily achieved across a range of points. There is a bit more loss to the right side of the graph, as the reflector gap increases.

As for your program, nice work there! I appreciate it, although writing down the lux values doesn’t take much time. It takes more time to adjust the height gage in .001” increments, as it’s easy to overshoot the mark or hit a .0005” value.

Also I was mistaken, my meter is an Extech EA30 model, not EA31.

I machined a clamp out of aluminum.

IF, as a rule of thumb, the base of the reflector should be at a height slightly below the top of the dome.

as you said, at the latus rectum line:

Then, to improve the dedomed beam on the left,

I should bring the reflector closer to the dedomed LED, by an amount equal to the height of the dome removed.

thank you very much for the brainstorming… really appreciate your measurements data… super helpful

I can't believe you posted a pic of someone's latus rectum!


I don’t know about that. I can’t yet say how an LED’s silicone dome affects the focus. It will hopefully become clear after some more tests.

It seems that a spherical dome shape would capture and bend the light from horizon to horizon (+/- 90 deg or more from center).

If the dome is removed or shaved then a dead spot out the sides and edges (less than +/- 90 d) leaves the dark ring.

A round emitter seems the best shape; square emitters, even with a spherical dome, seem to have dark rings and artifacts.

But it could be that all might be “fixed” with the shims.

This is a great test Hoops, many thanks. Plus it gives me a good use for my old vernier height gauge.

Nice to see this

Added a study of centering ring geometry for the S2 + WF1 combination.

I had expected to achieve 100% lux from the initial centering ring design. When that didn’t happen, I experimented with different geometry in order to reach it.

Great data, thanks for doing those measurements.

Thanks for the clamp photo. I know it’s fairly simple for you to do but that’s really some nice extra effort to go to. Really interesting on the ears and geometry as well…hadn’t ever thought about that aspect. And then how the heck did you hold the G10 stock for that tiny machining! Machine the top and then slice off from the rod (assuming rod stock is available…I’ve only ever seen it in sheet). The #4 configuration…I suppose that depends on mass production accuracy with the reflector bases, or can it accommodate a little variation? Never measured but all but the cheapest reflectors I’ve held have seemed to be very well made.

The clamp took about about 3 hours to make in total.

The gaskets were machined from .062” thick G10 sheet in a cnc mill. The fixture base was an aluminum plate, machined flat in-situ. The G10 sheet was clamped to the aluminum plate with Kant-twist clamps at four corners, close to the machining area. I would not use this method for production, but instead would use a matrix of screws across the entire sheet, machining the gaskets between the screws.

Manufacturing variability of the component stack (reflector + LED) is an important consideration and it would require more samples and testing to evaluate.

The reflector is presumably machined in two operations. The repeatability between the ops should be pretty good but not perfect. It depends on how they are machined and held and transferred between ops. A dual spindle lathe which finishes both ops without human involvement might be used. This would make the most sense considering the quantities being produced.

The vertex opening of the S2 reflector is pretty much exactly 5mm, (.197”- PIN GO, .198”- NO-GO) which suggests that care is being taken in their manufacture, despite how rough the outside looks. The finish on the inside parabola is good.

Compressibility of the centering gasket material matters in this case as well because the surface area on the bottom of the S2 reflector is very small. The compression force would depend on how tight you install the pill into the head. I think Nylon 66 wouldn’t compress much more than G10, meaning very little, but a series of gaskets could be made out of different materials, with slight height variations, .001,.002, etc., then a pill installed in an S2 several times with each gasket, and lux values recorded and averaged, then all results compared.

If I were making my own light I would go through all of this trouble…. :sunglasses:

In the case of gasket #4, the tolerances should be tight on the thickness of the center pocket area, I’d shoot for .0266” to .0272” probably, unless install and compression results proved otherwise. The outside ring of the gasket which centers the reflector matters less and is designed to be an interference fit. The edge of that ring has no chamfer and is kept sharp. With the forces involved in the install of the pill, the gasket is forced to stretch a bit around the taper of the reflector. This is why in the photo of the OP, the G10 ring is stuck to the base of the reflector. It bit into the aluminum a little. The height and ID of the outside ring of the gasket should be held pretty close. I don’t think there would be enough variability of the reflectors to affect the centering function of the gasket. Due to the interference fit, a little variation can be accommodated.

If the S2 reflector + WF1 were to be used in a device where there is no pill, no real compression force on the stack, a style of centering ring with tabs would be more appropriate than #4. A design which has 3 tabs 120 degrees apart instead of 4 at 90 degrees would still center the reflector, but block even less light. Based on gasket #3, but with three tabs instead of four, the resulting lux value should be 99.47% of optimal. An alternative way to center the reflector in this situation would be to utilize the outside diameter instead of the vertex opening, which would also be a “tabless” solution and so 100% lux should be achievable.

Interesting bits to chew on here. (speaking of that…is G10 rough on the cutters?) How do you suppose linen phenolic would do for this? It might have a little more compressibility than G10 but it’s still quite low. Machines well and isn’t terribly affected by humidity (would imagine that in a light head it would be a non-issue)…raw cuts and initial abrasive finishing sometimes leave it very slightly fuzzy, though. I don’t know how many people make the linen variety anymore but we used to have at least two mfrs here in the US. The canvas and paper are much more common (and both are cheaper, too). I’ve wondered about mylar, as a shim, too…never seem to find the mylar washers in quite the right sizing but I keep my eyes peeled (printers often have some of those on various guide bars and rollers)…not sure if it could handle the heat, though. Ceramic sure would be nice but based on the pricing of plain flat washers I don’t think most of us would spend the money on a fancy ceramic mini-widget if it could be made. :slight_smile:

I like the idea of micarta and what not but have not worked with it or tried to find specs. The specs will vary depending on what is used to construct it, both the cloth / fiber and the type of resin. Fabric cloth would make for less tool wear and better operator acceptability than fiberglass, I’m pretty sure. G10 is very abrasive to carbide tools and also to my sinuses. It contaminates the machine sump (coolant) a bit as well but doesn’t ruin it. The 10 in G10 means 10% glass content, I have read. There are grades of Nylon and other plastics that are 30% glass filled. The GF Nylon 66 has a higher working temperature than regular Nylon 66. Notice the “heat deflection at 264 psi” value. I have not worked with the stuff but it would be interesting to see how it machines and irritates compared to G10. G10 completely powderizes when machined. If the Nylon 30% GF made actual chips it would be nicer to machine.

I have not put much time into comparing gasket material properties but am starting to do so now.

Edit: Looking into material properties more, FR4/G10 is far superior to even 30% glass filled nylon 66 in its mechanical properties. It’s also much cheaper. FR4/G10 is what I will use to make gaskets going forward. I actually have a lot of experience machining G10, having made parts for the electrical industry. The downsides to working with the material can be reduced with certain machining strategies.

Great tests!

Would measuring the lux at a distance greater than 1m be better for focusing? With the luxmeter close to the reflector you could have a converging beam (fig. C) that at 1 meter reads higher lux than a parallel beam (fig A) but would perform worse at greater distances as it would start diverging past the convergence point. It probably doesn’t matter much with a small reflector but I think that’s why people like to test larger throwers at 10+ meters