5 luxmeters compared (april 2018)

The cause for this test was a 4-year old luxmeter comparison on TLF from member ' forest' (link) that I had never seen before and that was recently referred to on BLF. In that test, the noname orange-coloured LX1330B luxmeter (many people on BLF have one too) gave surprisingly good results in comparison to a high-end Gossen meter. Since I never had a look at that specific luxmeter, and that I needed a decent but not too expensive luxmeter for a new to be build integrating sphere, I bought one on dx.com, in the hope that it is old stock enough to be of the same build as the one in the german test. At least it looks exactly the same .

But before using it I wanted to know a bit more about it. The german test compared quite a few flashlights, but the variation in tint and CRI of that collection was quite limited: all cool and neutral 70CRI flashlights, and an incandescent light source. But I'm not so much interested in the overall calibration, which when checked is easy to correct with one correction factor, I'm more interested in a correct wavelength response, or in other words: is the same calibration valid for all tints and CRI, or does it need different corrections for every different tint or CRI that is measured ? (I discussed the importance of a good wavelength response in my lengthy first luxmeter thread 4 years ago ( https://budgetlightforum.com/t/-/28689 )

So I made it into a new direct 5 luxmeters shootout. I have done some testing on all of them before, but it is nice to test them all together now :-) I chose 6 different flashlights with different tints and CRI that all have a good constant output setting so that I could conveniently measure the throw at 7 meter distance.

Sofirn SP33

led: Samsung LH351D

4000K 90+ CRI

Sunwayman D40A

led: Cree XM-L2

6500K? 70 CRI

Kaidomain K2 host

led: Cree XM-L2

3000K 90+ CRI

Supwildfire 'Mitko thrower'

led: Cree XP-G2 3D dedomed

±4500K 70 CRI

Olight M10 'Maverick'

led: Nichia E21A

6500K 95+ CRI R9080

Ultrafire C10

led: Epileds 5050 powerled

400nm UV

MobiLux Class A

(2nd row: calibration corrected by +3.4%)

171

177

281

291

133

138

1544

1596

31.6

32.7

2.25

2.33

Tondaj LX-1010B 155 263 116 1273 32 73
UNI-T UT-383 162 282 116 1344 30 104
Tasi 632A 179 274 142 1554 31.5 0.42
noname LX-1330B 176 -0,6%
306 +5,2%
129 -6,5%
1428 -10,5
37.8 +15,6
43

For discussion of this table I focus on the two main possible causes of a different reading: overall calibration diferences and differences in wavelength response.

You may notice that for all light sources, the LX-1330B and the Tondaj meter differ by the roughly the same factor: the LX-1330B measures between 12% and 16% higher. So they likely have a similar wavelength response but have a 14-ish% different calibration (I hope this does not already disturb people ). If you choose one to be right, you can easily correct the other with one number.

In contrast, if we compare the LX-1330B with the Mobilux meter, the LX-1330B measures the 4000K 90CRI Samsung led quite the same, the dedomed 3D Cree XP-G2 is measured 8% too low, and the 6500K R9080 Nichia is measured 20% too high. So what is going on is not a simple calibration difference but a fairly different wavelength response, almost impossible to correct.

If I have to choose the luxmeter that performs closest to my high end MobiLux meter, that will not be the LX-1330B but the Tasi meter: two readings are within 1% of the Mobilux, and the the most extreme reading deviations are just 9% apart (compared to the LX-113B: 28%, UNI-T 15%). So the Tasi meter will be attached to my new sphere!

As in my first luxmeter thread, I took the time to get my trusty Zeiss monochromator with regulated tungsten lamp out of the cupboard, and recorded the tungsten spectrum response of the 5 meters. For details of the set-up see over there. Here are the results, but below the graph is a needed discussion of what you see.

Discussion of the graph:

*It is tempting to accept the graph as an absolute wavelength response curve for the luxmeters, but for that the radiometric power of all emitted wavelengths of the light source should be equal, instead the tungsten spectrum increases 8 times from 450nm to 750nm. I could correct for that (the increase is pretty lineair in that region) but I have no data on if this type of monochromator has a fixed percentage power loss over the entire spectrum, probably not because that is not a requirement for the spectrometer what this thing was part of (measurements are not absolute but always relative to a blanco). So because I do not know the added error of the monochromator, I will leave the graph as is.

*However, the curves relative to each other are correct. For example, at 450nm the LX-1330B measures 7 times as high as the MobiLux meter, this would be no different if the light source was equally powered over the wavelengths. This makes the graph still highly interpretable, the differences between the wavelength responses are well visible. Do correct in your mind that the further right in the spectrum you go, the lower the real numbers are compared to what you see .

*The readings of the different luxmeters were quite different despite that each time the same light source was measured. This is a results of the small light projection (5x5mm) on the varying diffuser sizes over the detectors (and distance of the detector behind the diffuser). For a fair comparison, the diffusers should be filled out with the light source, which was not possible. So the readings were corrected to each other, the best way I could think of is making the surface area under the graphs equal.

*Necessarily because what comes out of the monochromator is not much light, the responses of the luxmeters were recorded at the very low end of their ranges. So I accept the results to be true assuming that the wavelength responses do not alter at different light intensities. I never noticed such a thing, and have not read anything about it, but I also have not extensively reasearched it.

*The graphs underline that the Tasi meter's wavelength response comes closest to the MobiLux (that in itself may not be perfect but, being class A, should have a response closest to the luminosity curve of all). The LX-1330B has a very similar wavelength response to the Tondaj meter, quite far away from the MobiLux graph and I suspect that both optical filter and detector of the two luxmeters are identical. The UNI-T meter does not do a bad job actually both in calibration and wavelength response, especially considering of the low price shipped (I payed 13 dollar!)

The LX1010B always seem to be on the low side. Interesting to see. Maybe thats why I cant get my BLF GT pass the 1Mcd ? :)

No it is not, it (at least my copy) is reading high for a number of light sources, and low for others .

My GT reads under 900 kcd btw, I suspect it is underdriven but I have not checked that out yet.

Thanks for the test djozz!
I’m glad that I waited for it.

Thanks.

I wonder if some uncertainty could stem from measuring the lux from a slightly different spot in the beam? Did you happen to do some spot checks in a sphere which would probably result in more consistent readings?

I see an extreme inconsistency now, according to the wavelength response, the UNI-T meter should read nothing for the 400 nm led, but instead gives the highest reading of them all. Will check tonight if that reading was correct.

I tried to find the highest reading in the spot every time. My findings are that if the output of the flashlight is constant, measuring like this is pretty consistent. I did not do it this time but in another occasion comparing luxmeters, for each flashlight, after measuring with the last luxmeter I did a second reading with the first luxmeter and got virtually the same result.

Yeah, I guess at 7 meters the spot is quite big which makes it a bit easier.

LOL, here is part of the solution, the five detectors illuminated with 400nm light: two of the 5 diffusers actually show bright blue fluorescence!! :party:
(if I were to design a luxmeter :nerd_face: …well, never mind)
Luckily the common white leds we use do not emit close to 400nm!

Btw I didnt double check the pic. But mine isnt the same to the one I just mentioned. Name looks similar though. In your results the LX1010b has in 3 out of 4 (excluding the UV) tests the lowest numbers.

Funny how similar this is to Sound pressure meters and respective frequencies. I have been in audio for decades and this discussion is almost word for word the same.
The difference being sound meters all come with the response curves and different weighting for relevance in different environments. The color spectrum might be a little more complicated.

Thnx Djozz. Its bin a while since I did cd measurements. But I believe I am doing something wrong. I use the rapid tables converter for lux to cd. But for instance when I fill in the ± 1400lux of your Mitko thrower at 7m, I get 69 kcd, wich should be much higher.

Probably having a brainfart, can someone help me out. :weary:

No, nothing wrong with your calculations, none of the flashlights that I used for these tests were at maximum modes, but on a well regulated lower mode to obtain a flat output.

Thnx! Guess my mind can only think in max lumens and max kcd. Thanks again for the test. I only have the uni-t, but I also find it to measure 10%+ to low.

Thanks djozz :beer:

Thanks djozz. I wont mention invaluable but will say useful information. :stuck_out_tongue:
What would a rough price on each meter be?

You’re welcome. It isn’t invaluable to me either, I just find this measuring stuff lots of fun, especially when antique optical instruments can be used :slight_smile:

The newprices for the meters at the time I bought them were:

MobiLux Class A : 1400 dollar (but bought 2nd hand)
Tasi 632A: 135 dollar
LX-1330B: 27 dollar
Tondaj LX-1010B: 22 dollar
Uni-T UT383: 12 dollar

.
It is good to hear that my struggles are not unique :slight_smile:

Yes, there are complications in light measuring that are probably not present in measuring sound (but I have never been into measuring sound, it will for sure have its own pecularities). So much that an overall measurement error of 10% is already considered very good. What does not help is that even the most expensive luxmeters do not come with detailed wavelength/response data, if you are lucky you get a maximum overall error for that and that’s it.

And with the way we use luxmeters we do not even have to deal with how different luxmeters respond different to the incoming angle of the light, because with throw measurements the angle is always close to zero.

Great tests Djozz! I’m glad I found that TASI632. Now I’m more confident with all my measurements. This is a lux meter with decent accuracy across the full visible spectrum without breaking the bank.
I played with a DIY crude goniometer using my TASI632 but I’ll need to make new slit hole to it. 1mm seems to large for close range input.

- Clemence

What I did find out with the Tasi meter is if you measure at the very low end, under 10 lux, the reading gets inaccurate by static charge or capacitance or whatever something on the cable between meter and sensor, it can give an off-set error as large as 1.5 lux. If part the cable is resting on a metal surface (in my case the chunky monochromator) that error is gone.