Building an Integrating Sphere ...

[Endeavour]

Mr. GMan: Labsphere, the company I linked to, makes coatings specifically designed for use in integrating spheres, they have very low absorption rates, and scatter light very evenly; "reflective" was probably a bad choice of words on my part.

I'll second Don's warning and add that it's worth trying any chemical to be used on something like this on a small and non-critical area first to verify compatibility. I've dissolved several things unintentionally in the past through chemical incompatibility. It is my understanding that most epoxy resins will not harm polystyrene.

-Enrique

 

[wbp]

Barnard says the polystyrene sphere halves can be painted. This makes me think that these might be a better choice for this.

Labsphere's 6080 coating would be a good choice. It's the same coating that is applied to the inside of my Hoffman sphere. It's not cheap though. Edmund Optics sells the Munsell coating, which is probably pretty close to Labsphere's:

http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=1325

$250 for enough to coat 200 square inches. For a 12" dia sphere you'd need 2 bottles...

For what we're doing I'm not sure it's necessary though...

William

 

[LuxLuthor]

What a great collaborative topic, and major kudos to PrecisionWorks. Very great to follow this.

Most reference light sources are incandescent, running on very well regulated moderate/low voltage so as to reduce the effects of filament wear, and to get them to output Illuminant A ( (x,y)=(0.44758, 0.40745) - around 2856 K). The problem is that inexpensive light meters will read VERY differently at Illuminant A, which is what they were calibrated with, than they will at 5400K or 6500K.

So even if you calibrated your setup with an accurate incandescent light source, you would not get accurate readings from an LED flashlight with any of the Extech or similar meters. You'd be better off with an LED reference, which shouldn't be too hard to build. Or just use a known LED flashlight with a regulated power supply...

To get accurate measurements at different frequencies you have to spend a LOT more than the price the meter you are using, the most accurate being a spectroradiometer (I own two).

William

It was good to see them taken into account. I found this pdf useful in addressing wbp's LED vs. Incan point. This other general Integrating Sphere pdf from Labsphere is also a nice resource.

Precisionworks: A bit of food for thought from my end - you may want to consider applying a coating of some kind (thick primer, an epoxy resin like Envirotex swirled around the inside), followed by a coating of highly "reflective" white paint (www.labsphere.com has very lambertian coatings, others may sell some as well), which may give you a nicer optical surface inside the sphere, as opposed to raw styrofoam

-Enrique

Again, good to see the issue of surfaces (& baffle) being brought up for even more accuracy.

Barnard says the polystyrene sphere halves can be painted. This makes me think that these might be a better choice for this.

Labsphere's 6080 coating would be a good choice. It's the same coating that is applied to the inside of my Hoffman sphere. It's not cheap though. Edmund Optics sells the Munsell coating, which is probably pretty close to Labsphere's:

http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=1325

$250 for enough to coat 200 square inches. For a 12" dia sphere you'd need 2 bottles...

For what we're doing I'm not sure it's necessary though...

William

With his meter utilizing the CIE Photopic Spectral Response (380 - 780 nm), like you say an expensive paint designed to reflect a range of 200 - 2500nm seems a bit of overkill.

I'm wondering about the quality and applicability of the sensor being used with specific narrow wavelength LED (still reading) and how well this sensor reads/converts various LED ranges to CIE Photopic Spectrum (based on defined 555nm green standard)--note section 3.2 in first LED pdf link.

Then I'm also looking at the proper use of lux, lumens, watts in this scenario. Note summary in XeVision's page under section titled:

"A comparison of the watt and the lumen illustrates the distinction between radiometric and photometric units"
​

I was also noticing the specific designs this Edmund line has for LED measurements that go into various size I.S. Too bad this is all so expensive.

I hope I'm not taking this too much off topic...again, really cool topic.

 

[precisionworks]

Enrique, I had thought about coating the inside of the sphere, but thought that the first one, in raw polystyrene, would be a good reference. My first sensor port was a simple hole bored through the sphere wall - the high reflectance of the cut foam + lack of baffle gave inflated readings. Then a tube (painted flat black inside) was installed, which gave undervalued readings. That was removed & replaced with a tube painted flat white inside, with a 45° slant cut to shield the sensor ... which worked out the best so far. My reflectance is still below optimum, as the radius squared correction factor is 25, but the actual factor is 18. Which doesn't bother me, considering the cost of the Munsell coating

teflon seems to be a very good material for light reflectance

Don, PTFE, and probably a number of other plastics (like HDPE) would make a nice, optically 'soft' surface. I don't know where to purchase these in sphere form, but they would be rigid enough to need no external support box. Split in half & hinged like a commercial sphere, the 'innards' could be configured differently ... like putting the source light inside the sphere, shining directly at the baffle, with the photodetector location behind the baffle.

Your idea for machining reducing ports is great, and I have some scrap HDPE solid round that will work. Just need to make one for each diameter light, or figure out sizes so that they can nest inside one another.

LuxLuthor, thanks for those links! Even with reading pages of information on spheres, the first Labsphere article never popped up, even though I'd read the second one.

I'm wondering about the quality and applicability of the sensor being used with specific narrow wavelength LED (still reading) and how well this sensor reads/converts various LED ranges to CIE Photopic Spectrum

That has concerned me from Day One. General purpose (inexpensive) light meters are heavily biased to look at all light & read out the 'sunlight equivalent'. The spectral response of these meters is much more attuned to the incan portion of the spectrum vs the LED portion. If both incans & LEDs were tested side-by-side, the results would probably be unacceptable using an identical correction factor. But a correction factor for incans could easily be developed that would allow dual use.

I believe that the Edmund design would not be terribly expensive to build in the home shop ... the challenge is to find a CCD array (Charge Coupled Device, like a digital camera sensor) that reads only the 360nm to 1000nm spectrum. Either there are limited spectrum CCDs, or optical filtering is used to allow passage of only the desired wavelengths. Linear CCD Array Cameras, like the LARRY 2048 (Linear ARRaY 2048 element) are available, but couldn't find a cost.

I hope I'm not taking this too much off topic ...

I appreciate your input. If it wasn't for half-watt & MrGman, this design would have stalled early, and been of limited use. Now, with William building his versions, there will be a basis for comparison. Which is really neat

 

[precisionworks]

RE: constant current vs constant voltage, this is from the first link that LuxLuthor posted, in Section2.2.1:

In the lab, LEDs are usually operated in a forward bias direction from a constant current DC power supply. At low currents, the slope of the radiant power (luminous flux) verses time rises faster than the slope of the electrical power (start-up range) verses time. At high currents, the slope becomes flatter (saturation area), which is mainly caused by heating of the LED chip. Under normal operating conditions (between the start-up range and saturation area), the optical radiation emitted by LEDs is strongly correlated to the electrical current, which is why constant current is recommended for measurements intended to characterize the optical properties of an LED.

 

[wbp]

I believe that the Edmund design would not be terribly expensive to build in the home shop ... the challenge is to find a CCD array (Charge Coupled Device, like a digital camera sensor) that reads only the 360nm to 1000nm spectrum. Either there are limited spectrum CCDs, or optical filtering is used to allow passage of only the desired wavelengths. Linear CCD Array Cameras, like the LARRY 2048 (Linear ARRaY 2048 element) are available, but couldn't find a cost.

The Edmund instrument referred to uses a CCD Spectroradiometer (or spectrograph) to measure the light. (Note also that it also uses an integrating sphere if the light being measured is an array or larger than the specified size.)

This instrument does not use limited spectrum CCDs or filtering. It uses a diffraction grating to spread the light out into a known spectrum. Each column of the CCD array receives light from one frequency band. The CIE Color Matching Transforms are then applied to compute the levels of Red, Green, and Blue. The math to calculate x,y coordinates and Luminance values from there is straight forward and well documented.

You could build one (I have) but it's not cheap, and you still need to calibrate it. I built mine from a Shelyak Instruments spectrograph (about $3500 with calibration reference built in) and a Quantum Scientific Camera ($5000). It's quite comparable to my Gretag Lightspex and much more sensitive. I already owned the QSI camera, which I use in my business, and I use the Shelyak for astronomy as well. Otherwise doing this is not very cost effective...

Processing the data from the CCD array is also not trivial. The data will be numbers representing the light of each spectrum band, but the bands will be arbitrary. Most instruments use a windowing algorithm to compute the 5 nm bands needed for the CIE Transforms (or 1 nm if you want to go to that level of accuracy but that would be overkill for most uses). Once that's done the rest is, as I said above, pretty easy.

The camera also has to be calibrated for background noise (Dark frames) and uniformity (Flat Field). Dark frames are easy, Flat Fielding is not. The QSI camera I use is very new, and has a Kodak chip guaranteed to 1% uniform and actually measured better than 0.5%; this is good enough that we don't really need to worry about it. Normally such a chip comes in a camera costing about 50% more than I paid but QSI had a run of Class 1 chips at Class 2 prices, so I lucked out. (The current generation of scientific CCD chips from Kodak are amazing.)

I don't think you want to use a linear CCD array - isn't that a single row of pixels? That would have very low sensitivity.

William

 

[precisionworks]

isn't that a single row of pixels?

It is, as the photo above shows.

Talk about speaking before you understand the problem, I grossly underestimated the effort required for a limited spectrum meter

That's OK, since I only slightly underestimated the work required to get the IS to function properly

 

[LuxLuthor]

I think one of the most useful things about this thread is it begins to expose the standards & requirements for accurate descriptions & measurements. This has been rarely exposed in the many threads where Meterman LM631's & other CIE Photopic Spectral meters are used with impunity, and as accurate source information regarding LED Lux, Watts, & Lumens.

 

[wbp]

OK, don't laugh - saw this in my local Leslie's Pool Supplies. It's a sphere, all white inside, in two pieces with hinges. Hmmmmmm....

http://www.lesliespool.com/browse/H...-Cooler/D/30100/P/1:100:10000:1000010/I/43645

William :tinfoil:

 

[wbp]

OK, don't laugh - I was in my local Leslie's Pool Supplies and saw this - it's a sphere, with all white liner - hmmmmmmmm...

http://www.lesliespool.com/browse/H...-Cooler/D/30100/P/1:100:10000:1000010/I/43645

I suspect this one might be too reflective, but it's intriguing...

William :tinfoil:

 

[wbp]

An interesting article about an alternative to an integrating sphere:

http://www.laserfocusworld.com/disp...s-serve-as-alternatives-to-integrating-sphere
​

 

[precisionworks]

I like it ... no boxing needed, and one light port already installed

Good article on the baffled tube.

 

[wbp]

update on the styro spheres

I now have styrofoam spheres from both Barnard and Plasteel. In spite of their different descriptions I cannot see any difference, both appear to be ordinary styrofoam just liike what you would see in a cheap cooler. I was told that "polystyrene" was a smooth styrofoam, but apparently that's also what Plasteel is selling.

Warning: Barnard's 10" sphere is not completely hollow inside, it has flat areas at the end, apparently for strenght, but this makes it unsuitable for our purposes.

I got an 8" sphere from Plasteel which looks very promising. I'm going to start with that.

I just HAVE to pick up one of those beer coolers to try. I really suspect it will be too reflective, but if it doesn't work out, I can always use it for what it was intended for in the first place...

William

 

[wbp]

Inspired by Precisionworks, I built a 12" Isphere using a styrofoam ball. I've uploaded a picture here:
http://www.meier-phelps.com/temp/IMG00024r.jpg

The instrument is a Gretag Lightspex spectroradiometer, current calibration, using the fiber optic probe with cosine receptor attached.

Here are my first pass results, readings are in Lux:

Fenix L0D Q5 Hi 5803
Muyshondt Aeon Hi 10814
Lummi Raw NS Hi 22404
Bitz Pocket Hi 10984
Olight T10 Hi 13545
Nitecore Defender Hi 11446
Nitecore EX10 Hi 14620
Jetbeam II IBS Hi 17975
Milkyspit Creemator Hi 27396
Dereelight DBS V2 - 17970
MTE SSC P7 Hi 40874

If I divide by pi * radius**2, I get the following Lumen readings:

Fenix L0D Q5 Hi 74
Muyshondt Aeon Hi 128
Lummi Raw NS Hi 285
Bitz Pocket Hi 140
Olight T10 Hi 173
Nitecore Defender Hi 146
Nitecore EX10 Hi 186
Jetbeam II IBS Hi 229
Milkyspit Creemator Hi 349
Dereelight DBS V2 - 229
MTE SSC P7 Hi 521

To me, most of these seem to be within range of what I would expect, except the little Lummi Raw NS. The Fenix and the Jetbeam are spot on. The readings are repeatable to better than 2%.

I repeated the test, this time with the fiber optic probe without the cosine receptor. Readings are in cd/m2 (candela per meter squared).

Fenix L0D Q5 Hi 1915
Muyshondt Aeon Hi 3520
Lummi Raw NS Hi 7430
Bitz Pocket Hi 3617
Olight T10 Hi 4944
Nitecore Defender Hi 3932
Nitecore EX10 Hi 4845
Jetbeam II IBS Hi 5620
Milkyspit Creemator Hi 9180
Dereelight DBS V2 - 5800
MTE SSC P7 Hi 14394

Divide by radius squared to get lumens:

Fenix L0D Q5 Hi 77
Muyshondt Aeon Hi 141
Lummi Raw NS Hi 297
Bitz Pocket Hi 145
Olight T10 Hi 198
Nitecore Defender Hi 157
Nitecore EX10 Hi 194
Jetbeam II IBS Hi 225
Milkyspit Creemator Hi 367
Dereelight DBS V2 - 232
MTE SSC P7 Hi 576

These are consistent with the first set. Changing angle of flashlight does not change reading significantly. No baffle, but I'm considering adding one.

The opening in my sphere is not quite large enough for the DBS to fit, so it's resting on the outside. It's such a narrow beam I'm not sure how much this matters. I'm reluctant to make the opening larger just for this one light.

This is fun! But I've got to get back to real work now... :twothumbs

William

 

[MrGman]

Some of the numbers seem a little high to me for through the front lumens. Especially the DBS that doesn't fit into the sphere opening. You would be surprised at how fast the numbers fall off when I pull the flashlights away from the port by only 1/4 inch and those are ones that do fit inside.

When you get a chance can you post what the vendor's published lumens ratings are for each of those lights next to your numbers for comparison for those of us who aren't familiar with each of those models.

Also I don't suppose you got a picture of what the detector port area looks like with your detector at the port?

However, in general this also looks great and I am sure we can get it in tune with others. Its too bad you don't have a Fenix T1 to measure because I know that they read 230 lumens at turn on and about 225 real through the glass lumens on a good set of batteries when warmed up.

Your making this look too easy.

 

[wbp]

The DBS is supposed to be 210 lumens in this configuration, so that's pretty close. The opening is just barely too small for this one, most if not all of the actual light coming from that reflector is getting thru.

The Fenix is supposed to be 75, and the Jetbeam 225 on hi, so those are damn close.

I don't have time to dig up ratings for the others now, sorry.

Still scratching my head over the Raw NS - I knew it was bright, but that's a lot higher than the rated 160. It also gets hot pretty fast.

"easy"? Don't forget I'm starting with a $9000 instrument. Paying for that wasn't exactly easy (although it has paid for itself many times over since I bought it).

 

[MrGman]

I said you were making it "look" too easy. And definitely in regards to building a sphere and just throwing an existing meter into it that's true, since you didn't just run out and buy the $9K meter for this. But since you have a $9K meter on hand all the more better. (Take it as a compliment by the way, that's what it was )

However you readings are still on the high side of vendor specs. I am guessing that the baffle really needs to be in there to chop down some of that reading. The only flashlight that I have seen so far actually meet published specs was the Fenix T1 which they say has double sided AR coatings on the lens. My Fenix L2DQ5 doesn't have 180 lumens out the front, its more like 150.

The Malkoff M60 and M60F may be 235 lumens without a flashlight host's lens in front but what I measured from my Surefire 6P host was 220. So I am guessing that the DBS should not be reading 229 or more.

Just trying to help you fine tune it, it is very close, but I would still be more conservative knowing that the baffle isn't there.

It will be very interesting to see what precisionworks comes up with for the pill I gave him.

 

[wbp]

I'm not claiming that this setup is perfectly accurate by any means, but I do think the results so far are encouraging. I worry though that the primary issue is the quality of the reflective surface (styrofoam). There's a reason that they make spheres out of Spectralon or coat them with barium sulfate. It's quite possible that the reflectivity of our styrofoam ball will never give accurate readings for different types of flashlights without adding a coating. It's tempting to order some coating from Edmund and try it, but if their coverage specs are correct, it will take $1000 worth...

I just bought an L1T on eBay so when I get that I can see how it measures with this setup.

I've got some thin styrofoam sheeting - I'll make a small baffle and hot-melt that in place and see what that does, but from what I can see there isn't any direct light falling on the sensor probe. While I've got it opened up I'll take a couple of pictures of the inside. I'm adding a white PVC fitting for the sensor port to keep the probe placement consistent. It's threaded so I can change the sensor fittings as needed. I drilled out the inserts to hold the sensor snugly. The sensor itself is installed such that it is flush with the inside surface of the sphere. It has to be easily removable though because every time the Lightspex changes ranges I have to conver the probe to take a dark frame (Lightspex does not have a shutter). My lab is capable of total blackout though (which is not trivial), so in that case I can do dark frames with the probe in place.

This is actually leading to something I've needed in the lab for some time - the ability to accurately flat field my CCD cameras. The 12" sphere is large enough for the smaller camera; once I've got the kinks worked out I'll order a larger sphere for the big QSI camera.

William

 

[MrGman]

Yes this is encouraging. You answered my question about the sensor being flush to the inside wall of the sphere housing, that's what I wanted to see in the pics.

Tell my you don't have a Malkoff M60 or M60F? How can this be? Especially when they have been on sale all month.

Now all we need is to see Precisionworks update with the pill I sent him.

 

[wbp]

Tell my you don't have a Malkoff M60 or M60F? How can this be? Especially when they have been on sale all month.

Well, for one thing I don't have anything it will fit in, at least according to Malkoff's web site: "It was designed specifically for use in SureFire 6P, 6Z, C2, M2 and G2 flashlights." So what would I do with an M60 if I had one?

The only Surefire light I have is the E2E with a Milkyspit head, which is a killer flashlight. In general I find the Surefire lights overpriced for my tastes.