just a Newbie
DIY Integrated Sphere
I just did a DYI light project for measuring lumens and I want to share it here. Not too long ago, I bought a light meter from Amazon and started doing some ceiling bounce test with all the flashlight and bicycle lights I bought the last two years. I find it interesting to see the different result I find from various manufactor. However I did not like the fact that I had to wait for sunset before my room is dark enough to do the testing and the problem with different reading depending on the distant from the ceiling and angle I point it upwards. After some researching online, I found some folks who homade their own Integrated Sphere and thought I'll do the same. Well, it works although I'm not sure just exactly how accurate it is but so far it is not too far off after testing my light.
I saw some info of folks who build there sphere from foam. I look and found a hard rubber ball that I thought will work from Amazon. It is a ball made for pet dogs called Jolly Ball. I figure if it can withstand the abuse from a dog, then it probably will be tough enough. Beside that, the price was only about $24 and the size being 14" was about right.
When I received it, I notice there was a plug to allow the ball to fill with water, I figure I'll just cut that out with a hole saw as one of the two holes required for this sphere. I cut the ball in half and drilled out two holes. One has just enough clearance for the light sensor and the other is for the light source which I decide on a clearance hole for a 2" fitting. The two holes are 90 degree apart. I add a baffle to shadow out any spill I get foing directly to the sensor with a half circle plate. I had to rough up the surface with a wire wheel, prime it, then painted it with an Latex flat paint. The paint actually left a slight gloss sheen on it (old paint or mixed paint?) but I tried it as is to see if I had to repaint it later.
Next step is to add a fitting to the light sorce opening which I found and old magnetic base dial caliper holder. I drill and mount the arms to the ball and added a Twofish holder for holding flashlight. For bicycle specific lights, I mount those to a 1/2 inch PVC pipe and attach that to the Twofish strap. To hold the light sensor in place, I used velcro on the ball and on the senor. Here is a picture of the opening and also my Zebralight SC600 getting ready to be tested. The sphere sit on a training block which kept it in place.
I had several flashlights that some members at CPF has test and have some estimated measure lumens. Those lights were Zebralights SC600, Thrunite TN11, and Jetbeam BC40. I use these three flashlights to find my Corrected Value for each light and took an average of the three for my final Value. My Correction Value turns out to be 30.34
Here are some pictures of the SC600, MS808 P7, and Gemini Xera being tested:
I did some test of all the lights I had and most were as expected and some had proved my suspicion such as my KD C8 xml was only drawing 1.5 amps and my test result showed 593 lumens. My biggest surprise was my Yezl Z1 XML resulted with 662 lumens which explains the short runtime and the hot body.
What was dissapointing was the result I had from testing my Debree VBS V3 EZ900 thrower with an Aspheric lens. It resulted with only 200 lumens when it should have been more like 250-300. I heard of sphere having issue with thrower but not sure what I need to do to correct that. Maybe a different Correction value? Differnt Baffle? All suggestion are welcome for improving my sphere or even a name for it.
Anyhow, sorry for the long post, but just thought I share
Wow! Awesome job!
I would just make a mathematical correction for the aspheric thrower if you know the actual measured lumens. The roughness of the surface may affect the reflectance within the sphere.
This is just freaking awesome!
just a Newbie
Thanks. I"ll look into some known thrower info and tried to find some vlaue that works. I feel a little giddy with this sphere. Here is some data I got from my lights with the Sphere and they seem about on par (I think) but if they are way off, please advice.
Originally Posted by ThinkBike
MS 808 P7----------------------527 lumens
MS 808E xml------------------755 lumens
Titan P7--------------------------537 lumens
Bikeray IV------------------------966 lumens
Gloworm x2---------------------1212 lumens
Gemini Xera--------------------735 lumens
Gemini Olympia--------------1572 lumens
Zebralight SC600 xml-------------------751 lumens
Thrunite TN11 xml------------------------623 lumens
Jetbeam BC40 xml------------------------827 lumens
S mini xml-------------------------------------280 lumens
Yezl Z1 xml-----------------------------------662 lumens
KD C8 xml-------------------------------------593 lumens
MG X thrower xml--------------------------659 lumens
MG Rocket XPG R5------------------------303 lumens
Nice DIY setup. With any such setup, ESPECIALLY with a generic/typical lightmeter is that the lumen readings you get are going to be biased by the lightmeter.
Run of the mill lightmeters work fine for setting up a camera to take a picture with natural lighting or full spectrum lighting. The problem with LED lights is that they are anything by WHITE. The phosphors they use are very 'peaky' and designed to convert the underlying LED to a facsimile of white light. Your typical lightmeter can be very inaccurate at responding to those spectrum peaks and so you could have 2 lights with the exact same optics and TRUE lumen output but with 2 different binned LEDs and you would get 2 different lumen readings with your lightmeter.
Your setup will work quite well for comparative tests between various optics and drive currents etc., using the same LED, but it won't provide accurate lumen readings across the board. Obviously to do that would require some real $$$ to get a commercial integrating sphere coated with the $$$ paint and also a spectrometer based radiometer for more $$$.
Anyhow, I'm sure you're aware of the restrictions/limitations of your setup. Just making sure you understand that even if you 'calibrate' your sphere with a couple of know lumen output devices to get a correction factor, that you still won't get accurate (absolute) results. It will work great for RELATIVE measurements as you vary your optics/reflector/drive current, heatsinking etc etc.
just a Newbie
This brings up a question about light meter. Are there any light meter that does have some OK accuracy with breaking the piggy bank? I seen a lot of cheapy sub 100 dollar meter and know the limitation of those but any recomendation for $100 to $300 meters that are better? I also have been looking for Barium Sulfate powder which suppose to help when mixed with paint. This idea was mention somewhere in the past but have yet heard of the result nor have I found any for a descent price.
I have two projects which for now which only needs releative comparison which 'm hoping the DIY sphere will help. One is changing the cheapy driver on my 3 x XML Dry direct drive torch used as a bicycle light. It was a torch which I mod it with a back cap with a switch and external wire for a battery pack. It was shorten without the battery tube or the OEM tail cap and switch. I did burrn it up from either heat (most likely) or the solder blob that were present of the crappy driver board. I'm planning on getting a better driver and test the light output at various current along with the housing temp where I will try to add a copper plating and paste where the three LED sits on. The other project is for a Maelstrom X10 which I will try to do a direct drive (use as a quick on/off thrower) on it and see what extra gain I will get.
Are there really any obvious advantage of a sphere vs ceiling bounce test? I did notice different result from the same light because it seem that a ceiling bounce test is more sensitive to things like the distant from the ceiling, anlge and even moving or having new furniture in the room has changed my reulst. Not sure is this just specific to me or does this haappens to others?
Lightmeters are just NOT going to provide accuracy for what you are trying to do. That's why you need a spectroradiometer to get absolute readings that are accurate.
Your setup and a lightmeter will be fine for relative measurements with the same LED.
http://www.usu.edu/cpl/PDF/Barium_Sulfate.pdf is a decent article. Barium Sulphate powder can be bought from chemical supply stores - even Ebay has some vendors.
You built the sphere and are now asking why a ceiling bounce test doesn't give good results?? Isn't it obvious that the ceiling will reflect light in different ways, distance will obviously affect the amount of light hitting the lightmeter, the rest of the room will reflect some of that light back up to the ceiling and to the lightmeter etc etc, i.e. you have TOO MANY VARIABLES. The whole point of the integrating sphere is to INTEGRATE the light over a KNOWN unit area that is meant to be 100% uniform, the sphere is designed to reflect a beam/source of light over and over again to create uniformity. Integrating spheres are not just used to measure light, they can also be used to create a uniform beam. You have googled for things like... how integrating spheres work etc etc?
I'd suggest an evening in an armchair with your laptop as you google and read about integrating spheres and in particular the challenge of accurately measuring lumen output of an LED.
Your setup is a fine DIY project and I commend the effort and creativity that went into making it. Just don't expect it to give you absolute Lumen measurements of various lights.
Excellent job !
I don't think , you need to improve it in any way , for the intended purpose .
the ceiling bounce , has much more variables.
a wall test, where you shine it on the sensor, be good , to determine center brightness.
you can use the cree calculator to get into the ballpark.
and setup a barebone calibration, but need to measure Vf and current.
a sphere /sensor, is good for a certain range.
if you take a 5000k lumen monster, it won't be comparable , tolerances going to be too high,
to be meaningful.
time and heat play a big role too.
first 10 sec, 1min, 5min, 10min .
for me 5sec, 30 sec, since I don't like to wait that long.
oh, one improvement for the sphere, a translucent cover for the entry point .
shining the light inside, the readings change,
- when handheld,
- when different optics / reflectors are used.
the other though I have, is to have a large enough entry point, to be able to measure,
multiple lights in combination. say 2x X is still better then 1x Y
for your results, just add a note, something like this,
c-lumens, non-calibrated , lux (+/- 7%) times fudge factor = c-lumens (+/- x %)
as usual, just some thoughts,
only valid in my time-space continuum ,
for the mili second it occurred, if it did.
First I have to say that I applaud your effort to make something like this. That fact that it gives you a reading that can be usable in some fashion is a big Plus. Like someone else said I'm sure there are multitudes of various factors that come to play with the actual measured output. Even so, I'm sure an actual Integrated sphere has to deal with the same variables. Regardless the one thing that can't be measured with an I.S. is "Useable Beam Pattern". This only your eyes can do. Beam photos are very helpful but nothing compares to what your eyes are telling you.
Originally Posted by colleen c
Colleen I like what I'm seeing with the measured values on the torches and bike lamps. Particularly of interest to me is the KD C8 and the Yezi Z1 ( both XM-L torches that I own ). The measured output on the C8 doesn't surprise since the current draw is only about 1.5-1.7A on high with fresh battery. Even though the measured reading on your set-up is ~600 lm it is an "amazingly long thrower" and beats out everything else I own for pure throw. The Yezi Z1, on the other hand, amazes me every time I turn it on. Despite it's small size ( reflector and body same diameter ) it has an amazing output and very usable beam pattern for the bars. As such it continues to hold it's place on my MTB bars for back up duties or for "torch only" rides. ( I love a torch with a small footprint ) I think your measured value for this torch is spot-on ( 662 lm ). My eyes generally told me it was in the ~750 lumen range with fresh battery. ( * possible mine has a brighter emitter than yours since mine has a slightly bluish/colder tint )
I'd like to see you do the test with the MS 808E and Gemini Xera again. This time let both run on high for about 5 minutes with a fan on low and then give a reading. My bet is that the 808E will drop significantly. My 808E is very bright when first turned on but I have a feeling the output drops after it heats up a bit ( ** and it does heat up a bit! )
Anyway, I look forward to seeing more reports using the " Colleen illumination Sphere" or "CIS" if you will.
Last edited by Cat-man-do; 04-15-2012 at 02:36 AM.
just a Newbie
Originally Posted by rschultz101
I did a test with the 808E and Gemini Xera with my "CIS" thing. I took initial readings and got reading at 5 min and 10 min. As I recall Cat, my MS808E has a more yellow tint than your Magicshine has because we got our from different vendors. Here are the result.
Originally Posted by Cat-man-do
Initial start: 732 lumens
after 5 min 671 lumens
after 10 min 657 lumens and still show some sign of dropping although much slower than the first 5 min.
Initial start 727 lumens
after 5 min 697 lumens
after 10 min 694 lumens and the reading is stable
Good work on the Sphere Colleen.
The results certainly shows that thermally the Gemini is getting more heat away from the LED die or the components are of a better quality and therefore more efficient that the magic shine.