I’m confused on calibrating my 12” D.I.Y. Polystyrene (“Foam”) integrating sphere, modeled after this build thread by djozz with pics in post #64. Few things about my sphere: it’s sanded with 100 grit sandpaper inside, luxmeter is the “good” LX1330B and set into the foam a bit (no baffle, but shouldn’t get any direct light). Entrance hole is approximately 90mm at the top and 80mm at the bottom inside edge. I did NOT paint or cover the outside. Lux readings were taken in nearly complete darkness (verified that any ambient light was negligible beforehand), and during each “measurement session” I started with my EDC Convoy S2 on a freshly charged cell to be sure I was getting /- the same readings as during it’s initial measurement. “Consumer” flashlights with ANSI ratings were all tested with brand new Alkalines, and I even tested some 12v MR16’s and 2 household LED bulbs. Readings were always taken with the light’s bezel/- even with the bottom of the entrance hole.
I don’t have many (if any) reputable regulated lights to calibrate to and so I chose to use a co-workers Fenix HL55 headlamp with a fully charged cell. This headlamp has been tested in a DIY sphere at CPF with it’s ANSI ratings pretty well verified. I tried using a Princeton tec Apex headlamp (again - not mine) running on NiMh cells as a “calibration check” light, but the manufacturer doesn’t list their spec as ANSI nor outputs for separate modes (I now think the Princeton Tec’s “275 lumen” rating is with both the spot & flood LEDs on high together (a mode I didn’t measure in). Other ANSI lights I tested were a 2D LED Maglite, a 2AA Xenon Mini-Maglite, a 3AAA Home Depot 100 lumen XB-D light, a 3C Defiant XM-L (light prior to the “Super Thrower” which isn’t in my spreadsheet yet), a Walmart “Ozark Trail” 2AA XB-D and Ozark Trail 3AAA headlamp, Rayovac Sportsman Extreme 3AAA, and a Thrunite Ti (though I could not confirm it’s ratings are actually “ANSI”).
So I determined an average “sphere multiplier” using the Fenix HL55’s Low, Med, & High modes and then used it to calculate lumens on the rest of the lights. As you can see in the spreadsheet below, some lights check pretty well, but others are way off. I don’t know if the ANSI ratings are false, were tested on different batteries, or what. What do you think? Do I need some more regulated trustworthy lights to check my calibration with? Is there a proper procedure like calibrate to a low lumen level and check a high level? Or average a low & high “multiplier”?
I didn’t clarify some things in that spreadsheet. My lumen calcs are based on the multiplier shown beside “Alternative Average (dropping Low & Turbo)=”. I then have new “multipliers” calculated down that right side for lights that I had some sort of lumen output rating for to get a comparison against the multiplier being used. I also carried out those multipliers to 9 decimal places because I want to be extremely accurate .
I used an original ZL SC52 as the primary reference point for the multiplier, since it was by far the most accurate light I had at the time. Since then, I’ve confirmed the calibration on a variety of other lights but have not needed to update that calibration.
However, when I make a new djozz sphere #4, I want to do the actual selfbuilt-style calibration in hopes of getting more accurate results.
Thanks for the link TK, so is calculating a single “multiplier” not the correct thing to do? Or was that because Selfbuilt was only using a milk carton and not a round polystyrene ball? I’m not looking for extreme accuracy, but I don’t want to post numbers that are way off base either.
The multiplier is the easy method, but it’s not the most accurate method.
I know my measurements are off, primarily because I’m using a multiplier. So my lux-to-lumens conversion function is a straight line instead of a curve. I happen to have a straight line which is pretty close to selfbuilt’s curve for most of the range I’m interested in… but it’s still the wrong shape.
However, calculating the curve is not a simple task and requires a lot of measurements and a lot of outside data. So I haven’t done it yet. Hopefully the new sphere will provide motivation for me to do it.
I just for the first time read selfbuilt’s write-up about the correlation between his milk carton and other people’s integrating spheres, and I’m a bit surprised for two reasons, 1) that the correlation with increasing outputs is not lineair, and 2) that he accepts this as a fact and compensates for it, but he does not ask himself why the correlation is not lineair? (I don’t have a direct suggestion btw, would have to ponder on that before having any clue)
@Garry, I think your results are as good as you can expect for this type of sphere, I see a working multiplier of about 0.37 in many measurements, and multiplier values very far off are probably because the specs of that flashlight are not right.
As stated in the build thread of this type of sphere, it will not be really accurate because the reflectivity of the sphere (which directly relates to the multiplier) changes with whatever is in the entrance hole: a big reflector flashlight with ss bezel will increase the inner reflectivity and therefore measures relatively high, while a small reflector flashlight with black bezel will measure relatively low. In my other integrating spheres I compensate for that (by measuring the changed reflectivity for each measured flashlight, I use a constant output lightsource build into the sphere for that). Without compensation you can easily be a couple of percents off.
To get a feeling for how tricky light measurements are, and what variation you can expect:
*take a small zoomie in spot position (with partly drained cell and on low setting because then hopefully the light output will not change too much) and shine it (at the same position in the entrance hole) in different angles into the sphere, write down the lux-values and calculate how large the variation among the values is in percentage. This is an indication of how good the integration is, if this variation is lower than (say) 5% the sphere is doing its job quite well.
*shine your Thrunite ti into the sphere at the edge of the entrance hole, and while keeping it there, write down the changing luxvalues holding several different flashlights (switched off) in measuring position in the middle of the entrance hole. This gives the variation of reflectivity (and thus multiplier) caused by the diverse types of flashlights.
In the end, if you get within 10% of a real lumen value with your sphere, you are doing a very good job.
Thanks djozz, I’ll have to experiment with what you said next opportunity (free time is scarce). I also have my PVC light tube built (3” PVC), but never really used it since I only had the crappy luxmeter at the time. Perhaps I should try it out calibrating it with the Fenix HL55 and measuring some of the same lights again to see if I get Lumen values in the same ballpark.
Painting the outside of the sphere won’t really help in attaining a better multiplier, will it?
I wonder what effect there is from putting a white tight-fitting stencil around each light so that only the tip of the light is visible from the inside of the sphere, and there is reduced loss through the entry hole.
I expect the results will be significantly different than not using a stencil (I’ve already tested that), but I haven’t checked the inter-measurement consistency of a bunch of stencil readings versus non-stencil readings.
This is also part of why I don’t have useful measurements for my Blackshadow Terminator. I haven’t built a Terminator-shaped quad-bezel stencil.
You do not know how much a white stencil affects the multiplier until you measure it.
The widespread misconception is that you need to keep as much light inside the sphere as you can for the best accuracy. Keeping the light inside helps the amount of integration but with a well-build sphere that is not your problem, if your total of holes is under 5% of the inner surface (this is a rule-of-thumb number for a very good reflecting inner surface, our polystyrene surface is worse than that so I keep them under 2%) the integration is good enough to worry about other, bigger flaws.
What is really needed is not maximising the reflectivity at all times, but keeping the reflectivity constant. Then your multiplier is fixed. Wether you need the white stencil (or not, or any shade of grey) around a flashlight can only determined by measuring the reflectivity.
For improving accuracy for this type of sphere, you can either minimise reflectivity variance caused by entrance hole effects by reducing the hole/inner surface ratio (i.e. small hole in large sphere), or you can measure the specific sphere reflectivity for each to be measured light source before doing the actual light measurement (using a constant output light source elsewhere in the sphere) and then adjust the multiplier accordingly.
In my sphere #4 with build-in reflectivity measuring light source, I do not adjust the multiplier, but I have build in a simple device that for each flashlight optically adjusts the reflectivity of the sphere to a constant value.
The question about calibrating a sphere, I have answered several times since 2011. Perhaps it needs to become sticky’d.
There is a very simple, scientific way to do the calibration, without guesswork about some stock lights output.
All you need is a parallel beam of light (a thrower) and a luxmeter and the formula
A few concerns with that method (in no way implying that these concerns are any worse than the concerns with other calibration methods):
-the sun’s brightness changes per second (at least here in the Netherlands), so using the same luxmeter for the sphere and the lux-measurement in front of the hole can lead to inaccuracies
-the hotspot of a thrower flashlight is never uniform in brightness so you have to work with an average brightness of the light entering the sphere.
-the method is vitally dependent on a luxmeter that is correctly calibrated for the colour temperature of the light source (sun or flashlight), I would not trust any cheap meter for that.
Is there a link showing how you’ve done this? And/or can you go into some detail on it? Could I do something like create a hole where I tightly fit a Convoy S2/S2+ being certain to always insert to where the light’s bezel is flush with the inside surface, and run the light at a low to medium level on a AMC7135 regulated driver? I guess that requires getting a very good lumen measurement on that S2/S2+ light to begin with. Then my next question is, is that “hassle” worth it? Are we talking narrowing down our overall accuracy from say 10% to 5? Or is it more like 30 to 5%?
Is there a source where BLF users are posting lumen measurements for guys like me to reference and compare to? Or are they just scattered through posts everywhere?
I use a flat multiplier for my device… and it’s .26. I don’t have any of those lights you’ve listed, so I can’t tell you how close your measurements are. I know some lights are quite off, per their supplied specs. I calibrated my device at first with a Fenix HL50 and later a Zebralight SC600MkII. I have since lost my Zebralight, but I’d like to get one of their current generation lights to see how close the light stacks up to my old calibration light.
Thanks for your response. I was just finishing reading this thread where you were posting about figuring out a multiplier. Are you using a round “polystyrene” sphere or a PVC light tube? Are you finding your calculated “lumens” seem consistent from low values to high values and also from throws to flooders? Glad to hear you are also seeing some lights being off stated specs.
Hi, Djozz
I see what you mean about the sun’s brightness. I did not use the sun because the sun overloaded the sphere’s meter. The sun was just to illustrate the method of a parallel light beam. I used a thrower with a nice, even hot spot and was aware of the voltage sag so I took more than one sample pair of “lux in front of hole” and “sphere meter value”.
As you can see the sphere meter was occupied (build into the sphere) so the “lux in front of hole” has to be measured with a separate luxmeter.
It is impossible to say which is the best method, the imperical with an expensive flashlight or the scientific, by using the formula. That depends on the equipment used.
In my case there was no doubt about the method because I don’t own an expensive light other than an Zebralight H50. Fortunately it turned out to measure 60 lumens after my calibration which I think is the “correct” value. The sphere also confirms the binning of LED’s on stars.
I made that one calibration on the sphere in 2011 and have had no reason to distrust it since.
Happy measuring!
I really like the scientific method, going to give it a try sometime and see how it differs from my current calibration.
And I still have in the planning to someday spend some serious money ($400) on first building a dead-constant led light source and then have it officially measured at a facility here in the Netherlands that does certified light measurements with a professional sphere. At least I already own a very good luxmeter that was once officially calibrated (but the certificate is expired now).
I’m using a PVC pipe. I am happy with my multiplier. In all my videos, since summer on of last year I compare my outputs with the manufacturers ratings. Sometimes they’re close and sometimes they’re not. Like my recent slew of Nitecore lights, I’ve found on the top end my lights measure slightly more than advertised on most. But in other lights, say the Acebeam T20, my top mode was less (only by a little bit) then theirs. Here are a few observations that I can say about my measurements.
1. Sometimes my highs and turbo modes read a bit higher than others who measure the same lights I do. BUT those measurements are still lower sometimes than the manufacturer.
2. High modes and turbo modes drop the fastest, so I calibrated with one of the medium modes on the Zebralight.
3. I tried several modes on the Zebralight, and created a multiplier for each. Then I tested each one of those multiplier against higher and lower modes on the Zebralight. The medium and the .26 multiplier seemed to produce the most accurate results compared to other modes.
4. Fenix and Zebralight both seem close when cross calibrated and compared to one another.
5. Sometimes when I test a brightness and it reads higher than the advertised spec, the runtime can be shorter than advertised as well.
6. When I do beamshot comparisons on “high modes” my tested specs and the relative brightness to one another is often reflected in the beamshot section. Although it can be hard to tell visually between a 900 lumens and 1000 lumens.
7.I think generally overall my specs are slightly higher than other reviewers. Not by much, but with random sampling I’ve done. As far as I know, no one has ever compared my testings to other peoples. I’ve only made 1 or 2 transcription errors that people ave brought up in my reviews. Those were actually because I forgot to double check my notes or math… not because of a bad reading on my lumen device. I’ve talked privately to one or two reviewers who got slightly lower readings on one light than I did. But no real “Those are wrong” call outs. That might be because people don’t take my findings as seriously as other reviewers, or that people are generally satisfied with my findings. I can’t say for sure, but I prefer the later explanation.
Thanks for the response. Your method on the Zebralight sounds like what I was trying to do on the Fenix HL55, except that I didn’t run my “calibration multiplier” back against the other modes.
I like that and are seeing forward to the outcome!
The last 10 years of my working life I worked in a facility here in Denmark that does certified light measurements. We had a 2m professional sphere that got on sale. At first no one wanted to buy it but eventually it was sold for appr. 33000 EU, if I remember it right.
No wonder they charge really big money for trackable measurements.
.