The Legendary BLF Integrating Sphere starts here! (Delivered)

I need some feedback from the Pro’s :nerd_face:

I like .0060 for the sphere cal number.
And for thru wall measurements I came up with my own way of measuring. You can read up on it on the previous posts.
If your lazy to read, what you can do is take your thru wall measurement and multiply that with .238

Ie…your trustfire 3t6 is 18,000 x .238 = 4,284 lumens

Convoy L2, XHP-70 P2 1C on 20mm T-PAD, Zener mod MTN-20DD kit with SIR800DP FET and guppydrv universal firmware, bypassed (20awg) switch board and springs with copper contacts, 2 Shockli 3500mAh IMR26650's at 4.2v apiece. Sphere #6 cal .0060, meter cal 555

ML 3820 lux - 22.9 lm

1% 12220 lux - 73.3 lm

2% 20000 lux - 120 lm

5% 41700 lux - 250 lm

15% 129600 hole, 3700 wall lux - 778 lm

3700 x .238=880 lm

25% 6460 wall lux - x .238=1537 lm

40% 10300 wall lux - x .238=2451 lm

100% 25200 wall lux - x .238=5997.6 lm

Apologies guys, I just noticed the app was down, probably for the last week or more. I left a shared hosting environment and setup a new server. In the process I made a little typo in the Apache setup and neglected to test the BLF app. My bad, it’s fixed now.

Warning long post ahead, bullet points at the top for those that do not care how the conclusion was arrived at

Howdy, DB custom was nice enough to donate his sphere to me to see if I could figure out the readings, So I guess I have a reason to post now. :smiley:

Bullet points

- Got DB Customs sphere

- Tested every stock light I have

- Found massive variance between readings through-hole and through-wall and general unreliability of anything over 1000 lumens

- Also notice side issue of meter readings changing when set to 10x vs 100x

- Logged all the data to an excel file and crunched numbers

- Discovered that the light lost with strike bezels is the cause of most of the unreliability in through-wall measurements

- Discovered that heads larger then hole opening cannot be measured accurately.

- Spent hours testing materials for internal diffuser to increase the reading range of the meter beyond 1000 lumens with through hole measurement accuracy.

- Finally found that Styrofoam discs seems to work well and increases reading range to at least ~6000 lumens with much higher possible with denser/thicker foam.

- Found that a fixed multiplier seems to work great with internal diffuser and thus the sphere can be used like normal for any light that fits the opening. See pics later in post.

  • Realized what time it was and fell into bed after a long days work.

Before I start I want to say that Josh did an amazing job on these. I am a bit obsessed with some things, one of those is ingenuity. I positively love the overall design of the setup, the way it is self-contained, it uses the shipping box to house everything, the design of the 3d printed parts ect. It is all top notch and very well thought out. Props to you Josh!

Only changes I think I will make to the overall setup is making some kind of shelves/holder for the rings and meter on the inside of the box, so I can move it around without everything sliding to the back. Well at least for now, I tend to add bells and whistles more and more over time lol.

I want to thank DB Custom for the sphere, I would not have been able to do this otherwise!

I also need to say that I have only skimmed this thread, what I am about to say may very well already be known, although I never saw my possible solution mentioned.

I got it in on Saturday and started playing with it yesterday. I spent all of yesterday working with it trying to figure it out and I think I might have figured out the issue and a solution to the through wall / over 1000 lumen light readings that everyone has been having.

So now to the testing. Sorry for the lack of pictures, I honestly got sucked into this saga before I knew or planned on doing it and didn’t think to take pics of anything, although not sure if they would be been worth it anyways, you have already seen the sphere.

I will post a link to the Excel document with the testing from yesterday for ya’ll to look over yourselves, it is rough as I did it in a long marathon binge of flashlights flying everywhere and telling people to stay away from the box because it is “important” and not “just a box”. The overall data in it should be understandable, keep in mind that batteries were drained some by the end of the say, so numbers that are consistently lower across the board could be due to that, ask if you have any questions.

The only stock lights I have that would work for this that I could think of at the time are as follows:

Nitenumen NE01
Defiant 3c 850 lumen
Convoy L6
Supfire M6
Astrolux S1
WELL USED and abused US serial number with OP reflector Fenix P2D, carried every single day for the last 10 years. More neutral tint then HK serial
Fenix P2D HK serial number and SMO reflector, colder tint and hardly used in the last 10 years
Fenix P3D

All the rest are modded or in various states of assembly.

Batteries are HE2 or HE4 that started at 4.15v or better, most ended up around 4v +/- a bit by the end of testing.

I then tested via the normal through-hole method first for all but the Fenix, which I remember near the end of testing. I logged the readings on each mode, along with any cases of the meter reading changing on another range setting.

I then repeated the testing with through-wall measurements. I didn’t have any meter issues with these interestingly

First I started out testing some small lights and it worked great, consistent and the numbers were in line with what I expected. Exactly how actuate they are? That is for someone else to figure out which multiplier to use, I have virtually no stock lights to calibrate it with lol. The .006 multiplier seems to work well though.

The issues started when I moved up to my more normal lights which are 1000+ lumens and over what you can measure through-hole. Readings suddenly were all over the place and anything but consistent or accurate.

After an hour or 2 of messing with things I decided to get scientific with it and run the numbers. So I pulled out every stock light I have and started taking measurements. Logged everything and then studied the data.

About halfway through testing I figured out half the issue, which is that lights with “strike bezels” that let light escape from the front are impossible to calibrate no matter how you take the readings. There is simply too much variance over the output levels in light lost. The more light that escapes the worse it is to try to get an accurate reading. These became very clear when testing a SRK (not in the excel file) with 8x XM-L2 U2’s. The higher the mode, the further off the reading.

On the other hand lights with smooth fronts (such as the S1) read spot on with the through wall measurements. So this was the first major issue.

The next issue was for lights that are too big to fit in the hole, put simply, it is impossible to get a reading on these lights, it simply cannot happen. For example an L6 reads ~2500 lumens with a through wall measurement even though it is actually around 3500 lumens. This is due to the fact that the multiplier for the through wall measurement is skewed due to a lot of light being lost when taking the through hole reference. ~29% of the light is being lost to the edge of the foam.

This is impossible to account for, so unless someone has an idea on how to get all the light from a large head into the sphere and take a reading, a flashlight with a larger head then the through hole cannot be measured sadly.

Lastly the other issue that I have is that the lux meter disagrees with itself in certain ranges. For example on some lights I will get one reading in the 10x range but a completely different reading in the 100x range. I have no idea why this happens but it happened with my other lux meter as well. It only happens a few times but when it does, neither reading seems to be “accurate”.

It is repeatable, with the same light on the same setting. No earthly idea why this happens at all but it plays havoc on the averages and needs to be eliminated from the numbers in order to get an accurate multiplier IMHO. The only thing I can possibly think of is maybe the PWM of the light is somehow causing the meter to spaz? doesn’t make much sense though since it does it with my meter as well.

Once I knew these issues I then set about figuring out a way around them. We are always smarter than the problem, thats my motto and I want to know what my lights are making!

After a lot of thinking, testing and playing around I finally hit on something that seems to work quite well.

Basically the through hole measurements are great, they are reliable and besides the exact multiplier, they are accurate. The issue is measuring lights over ~1000 lumens which maxes out the meter.

The idea was to do a thorough wall measurement in order to diffuse the light enough that the meter won’t be overloaded. Great idea, one I experimented with on my own earlier this year with a PVC sphere (where I ran into the same issues, although since none of the readings were known accurate it was impossible to diagnose it). In practice though this simply does not work with most lights since they have strike bezels.

So we needed to eliminate the through wall measurements and yet still diffuse the light, so why not make it that simple? Lets simply diffuse the light inside the sphere.

I tried a bunch of things from white plastic lids to bowls ect but most of them did not diffuse the light enough to get a worthwhile increase in the measurement range to be worth it.

I then figured that if the Styrofoam wall was enough to diffuse it, why not use more Styrofoam inside the sphere?

I then found some 5” x 1.5” Styrofoam discs (well 2x .75” discs stacked) and I put that over the sensor inside the sphere. Being a circle it seals up nicely just laying it there but mounting it would indeed improve things.

Pictures of the internal diffuser:

With this internal diffuser in place I then took all the readings again. Did them just like the through-hole measurements.

All the sudden the readings on the powerful lights started lining up perfectly with the “open-hole” readings.

After taking all the new measurements and averaging the numbers, this particular diffuser setup (could easily be changed to allow for more or less lumens), a multiplier of .037 seemed to work across the board. Virtually all of the readings taken with the diffuser and using that multiplier came out to spot on what I would expect across the range (aka, the modes were spaced as they should be, if not exact on the lumen number). They also match up almost exactly with the non-diffused through-hole readings.

Every light I tried came out perfect and honestly the sphere is a lot easier to use like this since you would not have to change the test process no matter what the light output.

I am curious to hear what people’s opinions on here are about it. I welcome you to try it yourself and report back.

I have not mounted these discs yet as since these spheres were designed to be standardized I wanted to see if we could come up with a standardized way of doing this before customizing this one. Plus ideas on how to mount it while still allowing it to be removed if desired are welcomed.

Here is the excel file with all the raw data

Josh’s Standard Sphere tests Raw data

I welcome any and all comments.

It’s great to see more research being done on improvements. As for a standardized way to do that… perhaps the hole plug it comes with could be attached over the meter hole with some white caulk? White caulk would probably be needed instead of clear to prevent gap leaks and keep everyone standard. I wonder what the new lowest reading would be?

I actually tried using the hole plug and while it did work it only increased the reading range slightly IIRC, think the max was bumped up to around ~2000 lumens? I didn’t see anyway to increase that further easily so I started looking for other options.

Honestly something a bit thicker/denser then what I have would be ideal, should end up with a max around 10k lumens which should handle 98% of lights to be tested.

The minimum is actually still surprisingly low. I got a reading of ~9-10 lux with the S1 on moon mode.

Nice work, over my head to be sure.
Glad you got it and had the time to attack it with such vigor! :slight_smile:

You could also put a cheap neutral density filter directly over the sensor, sized to fit as closely as possible. Rotate the box so the sensor is at the bottom, and gravity should hold the filter in place when needed. A ND8 should increase the range to 8000 lumens, or a ND16 would let it handle 16,000 lumens. Or some people (*cough* Frankendale *cough*) might even need more than that, but the filters are available much stronger if desired.

That looks like a great idea because the effect is mathematically predictable, so there would be no unknowns in the lens calibration. One person could use the ND8 and discuss his results with someone using the ND16 and know that they are in sync.
Once we agree on the desired lens diameter, I can 3D print a lens adapter and mail them to everyone for free. Then the lenses can be removed or swapped, and still seal good.

Well, I only modded my TR-J20 to 13,500 lumens, unlike the 34,000 somebody else we know did… :stuck_out_tongue: I had to make an 11 ounce heat sink for him for that one, remains to be seen how much longer it’ll run…(Ok, so I’m THINKING about it! Geesh!)

I agree, the ND filters are a great idea, never even thought of that. They are pretty cheap on ebay as well.

The ND16 would most likely work fine unless you plan on measuring a lot of moon modes and the like, although you could simply remove the filter if needed for that.

I guess most people would never build/own a light over 8k lumens, particularly with a head small enough to fit into the sphere. So maybe the ND8 is the way to go, should handle the low end down to ~5 lumens based on my testing.

The nice thing is that you can self calibrate the filter pretty simply if you have a light that is regulated and consistent.

I will leave the shopping to you, Texas_Ace. You are on a roll. I would recommend a lens with a white colored band though. Black may suck the light out of the sphere and convert it to heat, since black absorbs light. If that’s not possible, I may try to make the holder swallow the lens as much as possible, but it would be harder to remove.

That’s why the filter needs to be pretty close to the right size. :slight_smile:

My lux meter has a black area around its white sensor cover. I figured I’d try to find something about the same diameter as the black part it already has. Just gotta make sure it gets a decent seal so all incoming light will have to pass through the filter.

What I can definitely confirm though… is that you shouldn’t put the filter right in front of a bright flashlight, like a triple XP-L with a FET. It only took about 3 seconds to get melty. :frowning:


I agree that a grey or white lens bezel would be ideal but in my searching I could not find any reasonably priced filters with anything but black. So I think we might be stuck with that. Possibly use a white pen/marker to white it out but that would be hard to standardize.

So for the filter we need to pick a size, the lux meter is about a 36mm opening, I assume the hole is about the same (don’t feel like opening it up at the moment).

So in theory anything over ~40mm should work but I figure ~50mm+ would be better to ensure the bezel is clear of the meter. Any opinion on how much oversized it should be? Say 77mm+ or just a little bit at ~55mm? Which would be easier to make a bracket for?

By cross referencing some filters from ebay to amazon it looks like even the cheap china filters get good ratings with no real complaints on lens quality that I saw. So doesn’t appear there is a need to go with anything pricey/fancy.

In which case that just leaves picking a size and seller off ebay/aliexpress.

After looking around I think that ND8 is the best option, it should allow most people to install it and forget it and use the sphere as is for just about everything. Plus they are easier to find in a wider selection of sizes.

For those that need more they can order an ND16, I will try to find a size that offers an easy ND16 option as well.

So opinions on size, slightly larger then meter or much larger?

Any opinion on Ebay vs Aliexpress?

You posted while I was typing.

This is a good point, so do we go with a small 37mm filter or a larger oversized one?

37mm seems to be the closest common size to the meter opening, aka: http://www.ebay.com/itm/Tianya-37mm-37-mm-Neutral-Density-ND-8-ND8-Lens-Filter-for-Camera-Camcorder-/250820925349?hash=item3a661797a5:g:keUAAOSwVFlT8cRE

There are a lot more sizes and options as you get into the 50mm+ range.

Thats sad but funny with the melted filter, guess they were not designed to dim flashlights!

Maybe it would be easy to remove the lens from the metal it comes with? idk. My plan for the mount was just a simple cylindrical ring that the threads could be pressed into to hold it in place. The layers of the 3D print should grab the tiny threads excellent. Then pull to remove.

I kinda doubt removing the lens from the metal ring would be feasible in most cases. Unless it is a super cheap filter they are designed specifically to not allow that to happen.

I like your mount design, simple and easy. It is possible you could extend the mount up and around the outside of the filter, thus covering all but the edge of the ring facing the middle of the sphere. Maybe have a small grove to pry it out later, although that could get risky of cracking the sphere.

Personally I think that small amount of black would be a non-issue for calibration purposes. It should absorb a fixed amount of light and thus the multiplier could be adjusted for it.

FYI, I have access to a 3D printer if that ever comes in handy, although not sure that the filament would be the same as yours and might cause a discrepancy. Not the best at CAD design either, thats still on my to learn list.

Interesting thing of note in this picture. Notice how the filter appears to be melted more around the outside edge of the LED optic?

That demonstrates why the larger head lights can’t be measured, you lose a lot of light to the foam for the reference reading and that then skews any other readings you may try to make.

If you look at the excel file you will see that the through-wall reading matched the internal diffuser reading both at around ~2500 lumens on the L6 (should be ~3500). So about 29% of the light was lost to the foam in that outside edge.

I wounder if it is possible at this stage to make the hole any larger? or was that shot down earlier in the thread?