Considering the way I am into lights measurements, I do not need an integrating sphere. I am perfectly alright with my usual 'claim' that the light outputs of flashlights and emitters that I obtain using ceiling bounce measurements are within 5% of reality. But I was very curious about how an integrating sphere would perform and thought that it was a really neat project to build one, but it had to become an as proper IS as I could make using budget materials. So there I went, for the fun of it :-) .
I started reading, Labsphere has a very informative pdf on integrating sphere building (here), I read some recommendations from Saabluster on CPF about how he used bariumsulphate on the inner coating, and even found an article of some weird folks who extensively tested wavelength dependant reflectivity of coatings using various fractions of BaSO4 in latex wall paint (here).
First I chose a luxmeter. And at the very start of the build so much for building a truly proper integrating sphere: a calibrated luxmeter that follows the official human eye response used for the lux definition goes from 500 dollar upwards. So instead I bought a budget luxmeter that has been widely used by flashlight enthousiasts (including match, for use with his sphere), a Ceto ct1330B. I made a post on how I received the meter here :sick: , there's also a few images of how the inside of the meter looks over there. And believe it or not, a good graph of how the budget meters respond to the light spectrum is impossible to find on the internet, the only graph I have is this very chinese one that was on the leaflet that came with the meter :tired: (CIE=human eye response, but what are nonometers?):
Next consideration was size. The bigger an integrating sphere is, the better, because the various holes in the sphere interfere with the integrating quality of the sphere. The rough recommendation is that holes do not occupy more than 5% of the inner surface of the sphere, less is better. On the other hand, my hobby workspace is 1 meter square of the computer desk and will never extend until my son leaves home (he is 3 now :tired: ), so a compromise was inevitable ;-) . I decided the light input hole would be max 50mm, it would allow emitter testing, and all the flashlights but one that I own fit in it, but it is not big enough for the really big reflector lights (that I do not have anyway, and if I do in the future I will make do with good old ceiling bounce testing). I also decided to use only part of the sensor device of the luxmeter that I bought for the sphere, to save on surface use. I found a 1 euro playing ball that was 24cm diameter, the sphere would be quite small but with this size the holes still occupy less than 2% of the inner surface so that should be ok. I started with the 'match' method: paper mache, and to make the sphere a little more solid I made a 25cm wooden box for it.
These few pictures do not justice to the number of boring hours that went into the paper mache work (making the wooden box however was fun :-) ). So far so good, but then I started painting the inside with latex wall paint and I was really unhappy how that turned out: layers of paper bubbled up during the painting and the end result was nowhere as smooth as I wanted it to be. It would probably even work ok as an integrating sphere, but I really wanted to make a better job of it, this would not do.
Now I bought a 25cm styrofoam ball for cheap and started working on that. There were some cavities on the inside that I filled with wall repair stuff, I made a 52mm hole for the light sources and cut a bit of the outside of the sphere flat to make the hole less deep for ease of use, I thought out a neat way to mount the luxmeter's sensor at 90 degrees compared to the hole (bored out the screwholes of the black sensor-cover to allow M2 bolds go through, and made new holes at the side to screw down the white dome of the sensor), I made a baffle out of cardboard, and taped it in the middle between the hole and the sensor (to stop direct light from illuminating the sensor).
I now sanded the styrofoam to really matt with 800 grit sanding paper, made a corresponding 52mm hole in the wooden lid of the box, made a nice finish to the outside of the box with transparant floor lacker, and the sphere was ready for the first tests. First observations:
-the sensor was too sensitive: the maximum measurable output was 720 lumen. There is an aperture needed in front of the sensor.
-shining a focussed zoomie into the sphere at different angles had only little impact on the reading, so the integrating properties of the sphere were already good; it should be quite insensitive to different beam patterns :-) . Only shining directly on the baffle had considerable impact on the reading (10% lower).
-the readings were really sensitive to the size of the hole :-( , when my small K10 flashlight was surrounded by a disc of white paper, the reading was a whopping 10% higher than without the white disc. For the Convoy M2 the difference was 7%. This is partly due to the small size of my sphere (the hole is relatively large and contributes much to the total light loss in the box), but in bigger spheres it should still be quite a problem. Just adding the white disc does not solve the whole problem, the measurement error will be different for every size light, even having a reflective versus black bezel makes a measurable difference, I found. The only way to solve this would be using a fixed 'adjustment light' together with the flashlight to measure the error caused by the presence of the flashlight before switching it on.
Next phase was a coating on the inside of the sphere, it should improve the reflectivity (good for integration=beam shape insensitivity), and when using barium sulphate in the paint it should also reflect most of the visible colour-spectrum equally well (good for 'true' luxreadings, that is: if you ignore the low-quality luxmeter :-( ). The first layer of paint was plain latex wall paint. I was given 100 grams of lab-grade barium sulphate from work (worth 4 euro's), the next two layers of paint were a mixture of 40% (weight) latex paint and 60% (weight) barium sulphate, with a little bit water added to make it better brushable. When all was dry, the whole surface was carefully sanded with 1000grit sanding paper to make the surface really matt and expose the barium sulphate as well as possible. Here's a picture somewhere during the painting:
Now the sphere let less light through the wall, suggesting better reflectivity, but it still had visible light loss. I now left it that way, not wanting to apply an extra four layers of coating, I decided that it was good enough.
Now the sensitivity of the sensor was adressed: a piece of duct-tape with a small hole was sticked onto the sensor window. Low-tech but effective: the sensitivity went down, the maximum measurable light output went up to about 4200 lumen, which is fine for me. Bonus was that the angle-sensitivity of the sphere was even lower afterwards, even shining a zoomie directly onto the baffle made hardly a difference to the reading.
I also made a wooden ring to narrow the hole to 31 mm, that should be better for emitter measurements, but it needs a separate calibration of course.
I made an adjustment light from scratch (no, not my entry for the OL-from-scratch-contest ;-) ). It had to shine into the hole on the side, occupying as little inner surface of the sphere as possible. Here's the ugly thing that I came up with, soldered together from copper, running on a 2x7135 board on a 14500 battery, the led is an Oslon Square 5000K that I had spare and is even a bit smaller than a XP-G. It was reflowed on a trimmed 10mm Sinkpad. The 'head' was simply made white using Tipp-ex correction fluid.
I thought it was nice to make a video of the reflow of the led on the board and the board on the copper pillar, so here's some reflow porn ;-) :
So now I am able to measure the impact of every flashlight on the reading and adjust the conversion factor accordingly :-) .
Here's the finished box once more:
The last little project was a new mount for led-testing. I found an all-copper CPU-heatsink with fan on a flee-market, and soldered a 30mm diameter copper pillar on top (lots of solder paste, large blow-torch) , with four brass holders. the top of the pillar was made really flat and two M3-holes were tapped to accomodate led-stars. This should do the job, the pillar fits nicely inside the 32mm hole of the sphere.
This is all for tonight, next thing will be a set of measurements figuring out how well the box works (not easy, hence the title of this thread!!!), with comments, this will be in the following post, but not tonight.
I hope you liked this report sofar :-)