Luxmeters: thoughts, findings, comparison

Nice work djozz, once again!

I suppose we are assuming the Class A/B Mobilux A meter is pretty much the gold standard. It's revealing and somewhat disconcerting to see the variations, and also disconcerting to hear that maybe the NIST calibrated Extech meter in the $100-$500 price range so praised about in our community, doesn't even meet a class C rating of errors up to 20%. This could explain why some have seen very little difference in the LX1330B and an Extech.

By comparing the meters side by side as you did, eliminates a lot of variables - temp, pressure, variations in battery effects over multiple trials, etc., so the results seem very real and should be reproducible. Of course I would like to see more common lights/LED's in the test, but obviously you have limitations.

Don't feel the data, testing and info you provided here is under appreciated from the lack of much response. It's going deeper than many want to go. I subscribed to this thread from the first day but held off posting.

Also, all those discussions going on about universal standardizing our meter calibrations seems out the window now, or radically changes it - we'd have to take tint/emitters into account in calibrating our readings, and that's not always well known.

Personally, I was looking at a data logging model of an Extech to buy, but now, I'm not sure it's worth it, accept for the better features, the precision may not be much improvement at all.

I read the post on Monday but was too busy scratching my head to post replies :stuck_out_tongue: .I’m sure I’m not the only one.

After a second reading today,comprehension is dawning.I’ve got a HS1010A from Gameboy sailor which,I presume,

is the same as the Tondaj(I can send it to you for “calibration”if you like :wink: )

Would it be feasable to add weak red and blue filters in front of the Tondaj to bring it in line with the lambda curve or

doesn’t it work like that.

Spotted this thread earlier this week,might be worth a $10 experiment

Thanks Jos for the tests!

That must have taken up a lot of your free time...

My conclusion from all those tests is..... what the heck, I will just keep using my cheapo meter and be happy with it.
Seems like you just never know if the results are100% correct.. and in most cases its only a small % difference.

There are many variables... and even the more expensive meter could over/under measure in those different circumstances, who knows which one was actually more correct?

It`s great to have a reference though!

I`m happy with the lux meter I have, as I mostly just use it to compare numbers between the lights I have, and it`s good to be able to see if there is actually an output drop....

When I measure the "throw" and it`s in the same ballpark as the specs tell... I`m confident the meter did its job ;)

but that kind of rules out when you start to mod that light :(

Thanks for your effort Djozz
 Thought provoking!

A question

I use a Ceto CT1330B and it appears to have been quite reasonable in findings until now. Its results have been about average for a range of lights tested on BLF eg A60, STL-V2 etc. But in testing some lights recently it maintains the same readings as in the past for lights in the 100+k lux range (eg Maxtoch 2X), but strangely now varies 18-20% low in the 30-80k lux range.

For example - Two 2Xs just tested both read 112kcd at 6m, fairly consistent with Maxtoch’s factory read of 120k at 10m. I tested an example of the same model light 3-4 months ago and it also read exactly 112kcd at 6m. That’s consistency! But in testing some new Maxtoch M24s they dropped from 83k lux on my meter 3-4 months ago (Maxtoch’s recent factory test 83.5) to just 65-72kcd now on my meter. Odd. I then retested my Jacob A60 which has always read about 52-53kcd, only to find it now reads 43kcd - all tests with fresh pana 3100s. Cleaned A60s threads/contacts and no difference - 18% loss in metered lux reading.

Question - can a light meter go bad in part of its reading curve - lose 18-20% in part of curve but remain ok in another? Seems to be the case. Batt in meter fairly new.

Oh my, that is exactly the kind of headaches I get when measuring light, and why I started the process of getting more grip on light measuring (which caused more headaches).

But, assuming a consistent luxreading set-up, it sounds more like a flashlight thing than a luxmeter thing. My experience with cheap luxmeters (and your Ceto, which I use as well, in my integrating sphere) is that with all their flaws of which I tried to adress one in the OP, at least they are very consistent over time: if I take a reading and do the same a month later with the same flashlight, it reads exactly the same, and for most of the drainage independant of battery level. I have no idea what is going on electronically in those things but in the sensor's point of view I can not think of a reason why it would be off just in a certain range.

EDIT: I did not test it (yet), but luxmeters apparently can have a very bad temperature consistency. As can be read in the first table of the OP, a class C luxmeter is allowed to have an error of 2%/K, and who knows about our chinese meters? I guess that 2%/K means that if the ambiant temperature is 10 degrees (Kelvin/Celsius) different (which can easily happen when measuring outside), the reading can be 20% off (my measurements were always indoors, with temperature not varying more than 2 degC, but still 4% error is big!). It may be not that bad, but we just don't know until it is tested: higher end meters come with spec.sheets with the characteristics, the chinese are not telling anything.

Interesting comparison djozz. Sorry for replying so late. Im more interested in the practical stuff, but I appreciate the effort on the theoretical part as well. :beer:

Ill share some thoughts/questions regarding the typical cheap budget meters (may not be wise to assume they are all fairly similar). What you think about them?

-Trying to find out if certain emitters are the bin they state are hard. Example. XM-L2 U2 1A vs XM-L2 U2 2C If you compare several of each emitter, and find out that 2C does read a bit lower, it does not necessarily have to not be U2 bin. It may just be the (cheap) lux meter showing lower number due to the wavelength.

-When comparing domed and de-domed emitters, the loss may not be as large as measured by typical cheap lux meters. As an example, my latest de-dome job (3x XM-L2) showed 10% loss, but it may have been less if measured with a better lux meter. (Im mostly using my LX-1010B, which I assume is quite similar or the same sensor as your LX-1010B)

-Your measurements on de-domed XP-L showed basically zero loss right? Could this be due to having less of a "strange change" in tint compared to XM-L2? Im not much familiar with de-doming those emitters yet, so feel free to fill me in. Just wondering if some of the before/after changes that we expect when de-doming are more exaggerated due to cheap lux meters.

-Difference when measuring the color emitters were quite disturbing.

-Getting people`s calibration closer to each other others are IMO easily doable if people (who buys a lot of lights) are willing to put in a little bit of effort and calibrate their numbers a bit closer to various other peoples calibrations.

Sending out "calibration lights" would require several lights with different beams and tints in order to better get some accuracy. For some it may not be relevant to use 4C as a calibration reference, since some people mostly stick to emitters in the 4800-7000K range. On top of that, if there is only a single light, and that light is measured with a lux meter with an "old calibration" that may or many not be that accurate, what would be the purpose then? We know different meters are likely to have different readings with different tints. And we are not sure that a meter that was calibrated a long time ago me be perfectly accurate today. Basically, the calibration may not become more accurate, and you would have to get "everybody" in on the same calibration. And that is the main issue, and not very likely to happen. Also, IMO if some project with sending out calibration lights at some point did happen, they would also have to be reference lights for lumen in order to be worth the effort and to attract more people.

Maybe it’d be easier to pass around one meter, everyone takes readings with the same device we’d all be on the same page. In theory.

RaceR, you are right that it is debatable if a 4C tinted light would be the best light to send out, it depends on what you want to check. It is a tint that measured close between the Mobilux and the Tondaj, being less sensitive for errors in the blue region it may be a good tint for calibration comparisons. To reveal wavelength response errors a cooler led could be more useful. I am thinking now of sending two lights, one is the 4C-tint light and another could be 1C or so. But I may be building a new integrating sphere this fall/winter, one that incorporates the Mobilux meter and is a bit bigger so that calibration of the conversion factor is less important with every light measured. Then measure the lumen and throw of the two lights before sending them. The relative differences between the two lights should be correct then.

The XP-L was tested in my integrating sphere, so that was not the Tondaj meter but the Ceto meter. The Ceto corresponds better with the Mobilux in the blue region, so I would expect the effect of a tint-shift on the luxreading smaller than with the Tondaj meter.

The colour filters of the Tondaj and Ceto meter are clearly different, these pictures (that I posted before) show left the Tondaj sensor, right the Ceto. The colours of the filters show quite correct on the photo, the tondaj green/blue-ish, the Ceto very green:

(I am not going to open up the Mobilux sensor!)

Sorry for not reading it right sooner. I skimmed it earlier but had nothing to add of value. Not sure I do now. The inherent issues of measuring lux or even lumens are daunting to say the least.

@ mudgrips: Try changing the battery. My meter underreads when it has anything but a fresh-ish battery.

Yep will do ledsmoke. Bought a new battery for the meter yesterday to try as its the only variable I can really check.

Will test it tonight.

Well 
 did retest last night on 2 lights with new battery in lux meter - but no difference in readings.

Meter remains accurate over 100kcd but still reads 18-20% low in that 30-80kcd range - have no idea why.

Dam! - meter must be b*ggered!


How will make make sure that your lumen calibration is more precise compared to others?

How will you make as many people as possible join in on using your lights/numbers as reference?

The goal is not to dictate my calibration to other people, I have no calibrated light source either (I call my calibration 'djozzlumens' for a reason), but to get a pretty exact idea what differences there are, so that you can perhaps define a standard 'BLF-lumen' ;-) . I must admit that the incorrect wavelength sensitivities of probably many luxmeters very much complicates the idea of a standard calibration.

I just hoped that people find it a neat idea and want to participate. The costs are no more than sending a small package to the next in line.

BTW, my lumen-numbers are uncertain, but I think that my lux-numbers, coming from this very good once calibrated meter, are indeed more precise than most other lux-numbers around.

Maybe you could talk with say 2-3 meter suppliers as we do with torch manufacturers, ask for some sample product, and test those particular lux meters for best accuracy. Then its a matter or referring blf people to those meters/suppliers. Could be basis for group buy and would also standardise readings for alot of us who use meters.

Would be matter of thoroughly testing 2-3 meters rather than sending ‘calibrated’ torches around alot of people. I’d certainly buy a tested and recommended meter - especially if it has been referenced against good expensive meter equipment. You might end up with two recommended meters in different price categories eg $0-50, and $50-150?

I’ve got to buy a new lux meter and would be keen to buy a make/model that someone has carefully calibration tested. Could help bring alot of blf readings/reviews into line.

Hoping this is an all right place to say this.....

It seems that in order to get a good idea of what a flashlight will do in terms of light output (quantity vs beam pattern) a better way to show this would be to use a 3D graph such as like the following describes:

This kind of graph showing beam spread and the colours representing intensities as from a flashlight shining upwards the highest portion being an arbitrary fixed distance from the emitter

added to this graph showing the beam intensities as on a flat wall for example at a certain distance

resulting in something along the lines of this graph (but other side up and much better and clearer)

These things can be plotted by any amateur physics hobbyist who can do simple trigonometry and calculus. Additionally specifying the light in the area under the cone would result in the quantity of light within the beam at a specified angle and distance.

This is just a thought as the light meters mentioned above are being used to obtain a smaller portion of what is mentioned here. If we can create these kinds of graphs then only the emitter colour temperature measurement will be left to solve.

Any ideas? If there is a better place to post this please advise. This was posted as a thought having read about the light meters above and additional thoughts were solicited.

There's not many rules at BLF, usually you can post things where you like and this thread is quite appropriate I think.

I'm not sure exactly what you mean but it looks an awful lot like the first flashlight test that I ever did a few years ago and that I posted on CPF :

I did use the geometry of the flashlight beam but did not do any computer work, just simple calculations, to measure the lumen-output of a light directly with a luxmeter. I now think it was a bit of luck but the actual output of that light as I later measured was within 10% of that measurement

That post was also a disappointment about CPF, the test was a bit of work, I had not read of anyone else doing it that way on CPF, but it attracted only one reply. It was a kind compliment from HKJ by the way

Because it was a quite easy test, I did a rough test on temperature effects on the reading of the Tondaj and Mobilux luxmeters. Method: I put the sensors (not the body) in the fridge, next to each other together with a thermometer. After 3.5 hours (the thermometer read 12degC) I took them out, attached them to a cupboard and measured the lux of the hotspot of my k10 constant output reference light at an arbritrary distance (1.2 meter or so) clamped in a stand. Then I took the readings of the meters the next 40 minutes with the sensors warming up to -hopefully- room temperature (in this case 24degC).

time after taking sensors out of fridge

Tondaj reading Mobilux reading
0 201 201
30s 201 201
1minute 201 201
2 200 201
3 200 200
5 199 200
7 199 199
10 199 199
15 199 199
20 199 199
25 199 199
30 200 199
35 201 201
40 85 85

Now that was a lot less exciting than I expected. Both meters did not care less about the temperature difference. I can think of two reasons:

1)The reading of both luxmeters do not suffer significant from a 12degC temperature difference. This is good news!

2)Or, the actual sensor of the Tondaj and/or Mobilux luxmeter is so well thermally insulated from the surroundings that they were not enough cooled down in the fridge/warmed up during the measurements to show the errors that would arise from a temperature difference. This scenario is also good news, it means that if you start a measuring session with the luxmeters at -say- room temp, it does not matter much if the temperature fluctuates, the insulated sensors will not notice that.

Conclusion: Warned by the DIN-norm I expected a significant worse temperature performance from the Tondaj compared to the Mobilux, but I do not see it in this test.

(And I never did a runtime test on my -two7135,XM-L2,0D- K10 before, 40 minutes runtime on an Efest 14500 of which 35 minutes flat regulated: you gotta love those 7135 chips for their constant performance )

Thanks for the continued testing djozz.

While the test equipment itself might not show much change due to temperature variation, do the lights perform the same? I know a light that makes a lot of heat at ambient 90 degrees F can’t make enough heat at 40F to warm the hands. So a chilly or cold night might allow the lights to run brighter, longer.

In the end, the hot rod I build here in Texas will behave far differently in Amsterdam. And this would tend to skew results as well, even if we all tossed our meters and jumped on a Group Buy to standardize.

So, RaceR86’s idea of standardization might we unattainable due to the wide degree of variances from multiple source points
emitter, star, driver, host, ambient, cell, meter, paint, lightbox and even individual testing variances from person to person.

Good idea, but it’s not looking good for making it work.