I found the problem is the direct path. After put a baffle between the flashlight and the lux meter sensor, the reading reduced a lot. Right now, my maximum reading (XM-L2 U2 at battery current of 1.5A) is about 10000.
By definition, if total light is 1 lumen over an area of 1 square meter, the lux reading should be 1 lux. I think the relationship of lumen and lux reading can be derived in theory. Of course due to various factors, the result will vary greatly.
Now I'm getting curious how it looks like :-)
I'm glad that with some troubleshooting you are getting it to work. What is the lumenrange you are getting now?
When I use a big baffle made by carton board, there reading is about 10000 for 1.6A battery current and XM-L2 U2). However, when I use a thin carton photo paper as the baffle, the reading increase to 35000 again. Maybe some light leaks through. I need to test some more before post photos.
I cut the baffle to be smaller. Now the reading is about 15000lux. The key is I need to find the correction, or dividing, factor.
I took a few photos of the sphere but don’t knw how to attach them.
There's a thread about how to post pictures. You can not upload pictures directly to BLF,you have to upload them to an image hosting site and paste the URL of the picture into the insert picture menu:
https://budgetlightforum.com/t/-/15452
Thanks!
The photos are attached. To have correct lumen values, it looks like I need to have a large dividing factor (>20). I don’t know if it is normal.
Ah, now I understand how you have build it. I'm not sure if it is done intentially, but your build is essentially different from the sphere in the OP in that you made a hole for the luxmeter. The sphere of the OP does not have that, the luxmeter sees the light through the sphere's wall, and that was for reason of simplicity:1) no baffle was needed this way, 2) some extra integration takes place, and 3) and the light reaching the luxmeter is decreased, getting the range useful without extra grey filters and such.
Now that you made it the way you did I think the best is to make the baffle as small as possible (and white coloured) , just big enough to stop the two holes 'seeing each other', and lower the luxreading in another way: a grey filter before the sensor, or making a longer black coloured tube with the luxmeter at the end of it. Using a dark coloured baffle to reduce the reading, like you do, also reduces the integration of the sphere, making it more sensitive to beam profile of the light source (the whole point of using an integrating sphere is to measure the amount of light of a source independant of what shape the beam is).
Thanks. What you said make sense. Now I am not sure my baffle is really needed. It does reduce the luxmeter’s reading but may or may not because the directly light was blocked. In any case, the reading was very high previously and I like it to be a little lower so what I measured will not exceed my luxmeter’s reading. I just want to compare different flashlights anyway so accuracy is less a concern.
Now, what I need to do is find a way to calibrate it.
BTW, just curious, what is your correction factor, i.e., the ratio of lux readings divided by estimated lumen?
You still need a baffle between entrance hole and detection hole so that no direct light from the source can get into the exit hole. It helps integration to great extend. A small roundish blockage in between can do the job.
The conversion factor (strictly spoken it is a conversion instead of a correction) differs a bit between different spheres that I made. As you can read there, the sphere of the OP has a conversion factor (I use lumen/klux for ease of use) of about 83, so 83lumen/klux measured. The max of the used luxmeter is 200klux, so the maximum output the sphere can measure is 16600 lumen.
Thanks for your clarification.
I read from one posting somewhere about a calibration method. I think it would be a good one. He suggested to measure the lux value of direct sunlight. Then pointing the IS’s inlet port directly face the same sunlight and measure the lux value at the sensing port. The lux value (directly from sunlight) times the port opening area gives value of lumen. The conversion factor can be computed from the computed lumen and the lux value measured form the IS’s sensing port. One may need to adjust the lux meter and the inlet port such that they are both at the highest values so they are measure the same thing.
I will need to paint the outside of the IS to be black so no light will come from the ball directly before I try it.
What do you think about this approach?
Calibration with the sun like this will work, but your calibration will not be independant from your luxmeter's calibration, if your lux is off, your lumen will be off. And luxmeters cheaper than several hundred dollars have big tolerances, even if they are correctly calibrated (which the cheap ones will not be).
Perhaps the best 'budget' calibration is using a household reflector led bulb of a trusted brand (Philips, Osram come to mind, they usually come with a (multiple) TIR), switch it on and let it settle for half a hour before using it for calibration. Your entrance hole must be big enough for it (50mm)
Do not paint your sphere black, with its thin walls it is even advised to paint it white for improved reflectivity. For a sun calibration event, put it in a cardboard box with a hole or something like that.
Thanks for your comments and suggestions. I don’t know how large is the error of my luxmeter. I will keep this in mind.
As for using household lamps for calibration, the problem is that my inlet light port is not large enough. In any case, I just want to have an idea about the flashlights that I have anyway.
I will let you know once I have more test results.
Shipping costs more than the price 
.
Glad I stumbled on this thread. Great work Djozz! Looks like I can locally pickup (2) 12 inch half balls @ $8each. Question, with a 12 inch ball (30.5cm), what's the largest size hole I could use and be within reason for decent results? Can I go up to 75mm? (i.e. for a Courui?) My use would be for bike lights which could also involve shapes that are not round (more oval).
Thanks,
Garry
Hi Gary, a 30.5cm outer diameter is about 26cm inner diameter. You're fine with a 75mm hole in that, it will affect the amount of integration quite a lot, but even up to 116mm is within 5% of the inner surface area (5% of the inner surface taken up by holes is a rule-of-thumb-maximum for proper integration, according to the Labsphere theory, but that is with a very high reflective inner surface, and sanded polystyrene is pretty white but not that white). But you could also make an insert in the 75mm hole for a smaller hole, say 42mm (a C8 would fit in that) by re-using the cut-out of the 75mm hole, if the cut is neatly made with a scalpel or exacto-knife. You need to use a separate multiplier for the two holes (so calibrate the sphere for each hole separately).
(for real accuracy the multiplier should be adjusted for each measured flashlight, but then you need to make use of an multiplier measurement light and that was not the focus of the 'cheap and easy one' :-) ).
I repeat this because someone above made a hole for the luxmeter also: do not make a hole for the luxmeter, the light is measured through the styrofoam in this design. The design of the OP assumes a wall thickness of around 2cm, this reduces the amount of light reaching the luxmeter enough to get the lumen-measurements into a workable range.
Thanks for the detailed response djozz!
-Garry
Thanks to this thread I now have my own little sphere…
Thank you djozz!
Cool! What did you use for the base and the “rings” for the light entrance and light meter? Looks like the base is thin cardboard? And the rings thin PVC?
My new lux meter should be coming tomorrow! (Curious to check it against my HS1010 to see if my HS1010 is off as badly as I suspect.)
-Garry
Looking immaculate! 
I see that your entrance hole is quite long, so could you do a little experiment for me that I’m curious of?: could you check how (in percentage) the luxmeter reading differs while you move the flashlight up and down the hole. The reflecting mirror surface should make that the difference is not much (all light should be directed into the sphere, and the reflectivity should not differ too much either), I would like to know how much.