A DIY Integrating Sphere for measuring total lumen output

DIY Integrating Sphere

For some time I've been meaning to put together a DIY Integrating Sphere to give me the ability to measure the output of a light and compare it to the specified outputs.

Well I've now finished Subwoofer's Integrating Sphere V1.0, and here it is:



There are many designs floating about for a DIY Lumen measuring device, but I decided to try something I have not seen posted elsewhere and take it back to basics.

In this V1.0 of the design, I've taken the following approach:

For the light integrator I wanted to use a sphere because it has no flat surfaces, so should help avoid any direct reflections skewing the results.

A hollow polystyrene sphere (30cm) appeared the simplest choice, but actually allows a lot of light to pass through it.

To counteract the high level of light transmission, the inside was given several coats of white matt finish acrylic paint. The bright white paint is opaque and reflects well, and the matt finish helps scatter the light and make the overall light level inside nice and even.

Then I needed a sensor…

I thought about using my lux meter, but ruled this out for two reasons. Firstly the sensor head is quite large and secondly, the lux meter does not allow me to log the results.

Finally I settled for a very simple circuit based around a photodiode. The photodiode was selected as it is specifically tuned to visible light (in fact it has an IR filter built in), and has a response very similar to the human eye. The reason for this is that I am measuring the output of a light which is supposed to emit visible light, so if the light happens to output a lot of IR or UV light, this is not much use, so becomes wasted output.

After working through various possible circuit designs from those that could measure sub-lumen outputs, but quickly overload, to the design I am using, which should respond relatively linearly up to many thousands of lumens.

Having cut an aperture into which the light will be shone, a small section was flattened just inside the sphere. The circuit board was fixed with hot melt glue such that no direct light, even from a full flood light, would fall directly on the sensor.

The output from the circuit can be measured with a multimeter in the mV range, and having a logging multimeter then enables me to check output over time.

Calibrating the sphere V1.0

Not having any 'standard' light source, I decided the best way to attempt some sort of calibration of the sphere was to use a selection of lights from reputable manufacturers including Fenix, Nitecore, FOURSEVENS and Zebralight. Each of the lights was set to all available output levels, with a reading taken for each, and all the results collated with the manufacturers specified outputs.

There was a general trend with a few results falling either side. I decided to use an average of all the results giving me a factor to use with the mV reading to convert it to measure lumens.


Linearity of response

To check how linear the sphere's response was (i.e. for each extra 100 lumens the output voltage should increase by the same amount), I set up the following test:

Choosing four lights with small enough heads to fit into the sphere opening all at the same time, I first measure the output of each of them individually.

Then starting with one, then another, then another and then the last, I noted the readings for their combined output.

The result was that the cumulative output at each stage, exactly matched the sum of the individual light's readings.

Due to the limitations of the aperture in the sphere, I was only able to perform this test up to 1500lm. As this portion of the response was completely linear, hopefully higher ranges will also maintain this linearity of response.



The results so far

Comparing extreme throwers and full flood lights has so far indicated that the sphere's integrating function is working very well and does not favour any beam profile.

Having had a few pre-production samples for review with unspecified outputs, the sphere has enabled me to provide estimated outputs for the levels, in lumens.

Following a review of a 2000lm light which came out at 1891lm using this sphere, I decided to measure a 65W HID. This was done with some concern as the heat and UV emitted by this light might have damaged the polystyrene and sensor. The peak reading was 3560lm, which seems very reasonable as HID's emit a lot of IR and UV, and it was an ebay HID. The sphere did start to smell a bit hot, but came out unscathed and after checking a lower output light again, found no damage had been sustained.


So for a first attempt I am quite happy with it. There are plans to add a sub-lumen element to the circuit, and to use a larger sphere (50cm), when I have time…

Thanks for documenting this! I'm planning to revamp the device I currently use, and I like seeing the approaches others are taking. I currently use a photocell in resistor mode, and I'm planning to do something similar again but with a better photocell in photovoltaic mode.

Sent from my mobile device. Please excuse brevity and typos.

Bigmac_79 said:

Thanks for documenting this! I'm planning to revamp the device I currently use, and I like seeing the approaches others are taking. I currently use a photocell in resistor mode, and I'm planning to do something similar again but with a better photocell in photovoltaic mode.

Sent from my mobile device. Please excuse brevity and typos.

Click to expand...

What do you use as the light integrator?

subwoofer said:

What do you use as the light integrator?

Click to expand...

Right now I'm using a box covered on the interior with white paper. Not the perfect integrator I'm trying to decide if the next one I build should be from a sphere or a PVC contraption. As of yet I haven't found a suitable sphere in my price range, as I want a setup big enough to handle large lights. I like your idea of coating the inside with matte paint. I was at Lowes the other day looking at paints, and I was thinking I'd really like some textured paint to scatter the light around a bit more, but I haven't yet found anything that looks quite like what I want.

Since you got the sphere set up, how long does it take to get a reading for a new flashlight?
I noticed that older reviews dont have any table with your lumens measurements, e.g. LD12 R5, TK21 U2, ..
Is it asked too much to insert lumens tables in each of your flashlight reviews and not only the newer ones since you got the sphere?

shelm said:

Since you got the sphere set up, how long does it take to get a reading for a new flashlight?
I noticed that older reviews dont have any table with your lumens measurements, e.g. LD12 R5, TK21 U2, ..
Is it asked too much to insert lumens tables in each of your flashlight reviews and not only the newer ones since you got the sphere?

Click to expand...

Good question as I have considered re-visiting some older reviews that were carried out before the sphere was finished.

My test bench is unfortunately not 'allowed' to remain set up as it looks 'messy' ;-) So between test runs it get dismantled and put away. What is not shown here is that the test meter used is connected to a laptop to log the results rather than rely on me and a stopwatch, and an oscilloscope as well.

Even once set up, there are the pre-test requirements as I need to prepare the test cells to ensure they are fully charged and in good condition, and then I have a set of three calibration lights which are used to monitor and maintain calibration of the sphere. These lights have been carefully selected as they have outputs which on the medium setting are very consistent regardless of the quality of battery used. Once all of this preparation is done, I'm ready for a test run, and ideally I would then test a set of lights (as for each set-up I have to do all of the previous steps). Each of these tests needs to be done several times with different cells allowing time to cool between tests to 'reset' their condition.

So a single light would take around 3-4 hours to test, and one as a part of a series is 1.5-2 hours. Considering that I fit in all my reviewing around a full time day job and the requirements of married life, it is unlikely I will find time for this as the new reviews are my priority. If I find myself at a loose end then I might do this.

subwoofer said:

Not having any 'standard' light source, I decided the best way to attempt some sort of calibration of the sphere was to use a selection of lights from reputable manufacturers including Fenix, Nitecore, FOURSEVENS and Zebralight. Each of the lights was set to all available output levels, with a reading taken for each, and all the results collated with the manufacturers specified outputs.

Click to expand...

Thanks subwoofer for your time and hard efforts!

Calibration. Imho calibration is the key point of published lumens data and to me the one which matters most because once your setup is calibrated to an agreeable scale, we wouldnt need so many lumens tables from the same reviewer in order to compare brands and lights.

From my research and experience, Nitecore Zebralight Klarus Selfbuilt all overrate the lights. Overrate compared to what, compared to which scale?? Good question. The answer: compared to Fenix published lumens. As late as in winter 2010/11, Fenix was the first in the industry to introduce the ANSI FL-1 rating in their flashlights with 100% seriousness and accuracy. No more generous overrating and no conservative underrating but precise serious data.

Clearly, a light rated-100lm-by-Fenix is much brighter than a light rated-100lm-by-Zebralight (or Nitecore Klarus Selfbuilt). I like Klarus and Nitecore build quality and i own a couple of their lights but their lumens ratings are totally off compared to Fenix lights.

From the limited data of your review series i can tell that your published lumens data never coincide with the Fenix scale but tend to be higher than Fenix specs. At the same time your numbers are lower than Nitecore promises. So to speak, your numbers fall right in the middle between the truthful exact Fenix numbers and the self-overpromised Nitecore numbers. If a light model 'ABC' has got the Fenix logo and Fenix specs it with 100 ANSI lumens, the identical flashlight with a Nitecore logo Nitecore would spec it maybe with 140 "ANSI lumens we are member of PLATO blablah", and in your review of either identical light you would "measure" maybe 120 "ANSI lumens". So how is it possible that 3 parties publish 3 different ANSI lumens data for the same identical light?

Now it's too late. You established a calibration according to your assessment and the shifted scale remains as it is: your numbers will always make Fenix lights look self-underrated and Nitecore/Klarus/Zebralight/.. lights look self-overrated.

I agree that Fenix lights should be used as singular source of reference to calibrate a sphere, and I dont agree that Nitecore Zebralight Klarus (or Thrunite for that matter) should be included in the calibration process because they self-overrate their lights to varying degrees and inclusion of any of their lights would distort and shift the scale upwards beyond control.

So probably the best way for calibration of a sphere, since we dont have a 'standard' light source, imo is to focus on a single manufacturer, Fenix, and measure say 10 popular lights of them and learn if their published lumens specs are consistent with each other once the sphere is calibrated to the 10 lights. If a reviewer's sphere can reproduce the Fenix specs of all 10 Fenix lights, then it would mean that Fenix publishes consistent data and that the sphere is PERFECTLY calibrated to Fenix standards, Fenix lights!

By the way, i really like and appreciate your and mev's lumens scale so far. They clearly show that Nitecore Klarus Zebralight overstate their lumens ratings when compared to lumens ratings by Eagletac, Fenix or Rofis.

EDIT:
btw, which model/product of photosensor do you use, from where did you get it and is there any datasheet about it available?

shelm said:

Thanks subwoofer for your time and hard efforts!

Calibration. Imho calibration is the key point of published lumens data and to me the one which matters most because once your setup is calibrated to an agreeable scale, we wouldnt need so many lumens tables from the same reviewer in order to compare brands and lights.

...
By the way, i really like and appreciate your and mev's lumens scale so far. They clearly show that Nitecore Klarus Zebralight overstate their lumens ratings when compared to lumens ratings by Eagletac, Fenix or Rofis.

EDIT:
btw, which model/product of photosensor do you use, from where did you get it and is there any datasheet about it available?

Click to expand...

Thanks for taking such an interest.

I agree with you, that calibration is a very important factor. In fact I have been considering a plan to unite all the reviewers who have ISs and get together a set of standard sources (maybe a regulated AA light), measure them all with my sphere, and then send them out to all the other reviewers to see what their measurement comes in at. This may then allow us to compare across all reviewers. However this will require some investment which I do not have spared funds for at the moment.

I would disagree slightly about using a single manufacturer. I considered this, but wanted to work on a more balanced view. So I used everything I could lay my hands on at the time and plotted a graph of readings and specified outputs and having discarded the readings that were way off the trend, created a line of best fit to the scatter of readings. It was this best fit factor that I have used for my calibration. I now use a set of reference lights (which don't get used for anything else) to monitor any possible sensor drift. So far there hasn't been any.

Even taking my middle ground calibration, there have been many disappointed manufacturers. Lupine for one were not happy with my output figures and they apparently take the lumen measurement as seriously as Fenix.

As I stated, I only present my output figures as an estimate, but one which all my reviews can be compared with, I do not have the money to 'certify' my output measurements. Within the limitations I have, I am very pleased with the performance of my IS.

It is by no means perfect, and I know I have some losses, but at least these are consistent for all the measurements I take and allowed for in my calibration factor. I also have plans for a few improvements when I get the time to build the V2.0 sphere with dual sensors.

The photo-sensor is a Sharp BS120 and its data sheet is online.

LOL

So if you use Fenix as your calibration standard, any lights deviating from Fenix are considered inaccurate.

In my own measurements, including some done with TurboBB at PF17, etc, I can't say I agree that Fenix is the only maker to be 100% accurate, or even that they ARE 100% accurate.

I'd postulate that NO ONE is 100% accurate, as the LED's are not even that consistent in output. So if a few lights (3 is all the specs you referenced call for, a statistically small number) out put 100 lumens, later batches of the exact same light might put out 80 - 120 lumens, and so forth.

A published lumen or cd spec might be for the whiter emitter and multiple RCR's, but the sample received for reviews, etc, might be a neutral white, and might be tested with an 18650 instead of two RCRs and so forth.

One might use ICR123's, another RCR's, one cell combination might not handle high current draw as well, and produce a dimmer result, and on and on...all of which can change the output.

When you add up all the LED variables, all the light variable, and all the cell variables, and the fact that the official test only uses 3 lights, you can see that ALL specs are to be taken with a grain of salt, regardless of maker.



I'd say the specs published are essentially a ball park rather than a seat number.

subwoofer said:

to 'certify' my output measurements. Within the limitations I have, I am very pleased with the performance of my IS.

Click to expand...

+ 1
Me too i am super-pleased with the performance, as reported by you, of your IS: your numbers are in the middle ballpark between low Fenix Eagletac numbers and high Nitecore Zebralight numbers. And since you take great care, efforts and time to calibrate and recheck its calibration status, your balanced approach of getting a best fit factor throughout a set of premium flashlights of various carefully selected flashlight brands —i dont know about Lupine but please omit all Klarus from the best fit/least squares calculation— is 100% acceptable, commendable and accepted. Congrats to this methodology!

It's a pity that there are no standard light sources for setting up the calibration of a DIY sphere.

Even more a pity, there exists no independent institution which does certify the truthfulness of manufacturers' published lumens data. In theory and in practice, manufacturers can write anything on the websites and retail packages and claim that the numbers are their own measurements taken with an ANSI FL-1 sphere. Nobody certifies that the sphere really exists, or that it is perfectly calibrated, or that the published data are indeed the numbers from the measurements. A perfect system of data publishing would require the existence of a single independent authority or institution to which wealthy flashlight manufacturers must submit several samples of a production light prior to releasing the new product to the market. Then the authority would publish the certified test results on their own website. This would guarantee that the lights were measured with the identical sphere by the same certifying institution.

Manufacturers can print retail boxes with their own "ANSI specs" and there is no independent certification for this particular flashlight model, really? What a disaster.

Well, there is PLATO. But PLATO doesnt certify submitted flashlights neither are they independent.

I do trust Fenix efforts and subwoofer efforts. Looking forward to any further lumens measurements you can provide. I didnt know that it takes that long to take a reading until it is secured for cpf publication. Insane work and time investment!!

Thanks so much for all!!

Great exchange Shelm and Subwoofer.

I too find frustration with the different manufacturer interpretations of "ANSI" lumens and have resorted to using a DSLR as my own ambient light meter because I could not believe what I was seeing from certain manufacturers. While the DSLR its no integrated light sphere, and has relatively wide increments, I still find it far more accurate and objective than what my eyes can perceive. Most of my light collection is composed of manufacturers that are on the "conservative" side of the scale and so it is easy for me calibrate, and draw a best fit line, with 80% of my lights.

While I find it unfortunate that some reviewers seem to have adopted a liberal lumen scale (simply based on the quantity of lights significantly "under-rated" vs "over-rated"), I'm not sure there is any alternative. Unless the "official reviewers" use their own funds to buy all their lights independently, just like any car magazine reviewer, they'll need to walk a fine line, and on egg shells, to avoid upsetting anyone, as mentioned in the Lupine example above. Upset enough manufacturers, and they will not remain a supplied reviewer for long - it's always easier giving good news, and great news, than giving any bad news.

Unfortunately for all of us, the big downside of reviewers using liberal lumen scales is that it becomes a form of de facto certification for most folks and probably results in sales movements which actually reward bad behavior. As long as a reviewer's information is consistent and comparable (and this does seem to be the case), the information should be available to re-calibrate the "conservative" manufacturers on an apples-to-apples basis.... but let's be honest, not many know to do this, or bother doing this, and there are so many models that have not been reviewed.

If honest manufacturer's sales are being hurt with their specification claims, and we have review systems in place which may actually reward exaggerating manufacturers, then this whole spec sheet thing seems destined for eBay/Deal Extreme -like advertising.

Consumer Reports needs a flashlight section.