I thought that after my test of 5 luxmeters in april 2018 I was done with luxmeter testing (not exactly a cheap pastime if you want to test anything else but very low-end meters), but then early this summer Enderman came with an unusual proposal: he would buy the Tasi 632A luxmeter that I tested in april this year off me for the new-price, so then for the money I could buy a couple of other new luxmeters that he found interesting, to be tested by me. In the end both he and I spent even more money than that and I ended up with 4 new luxmeters to test, among which a rather pricy Extech meter. Thanks Enderman for supporting the advancement of BLF-knowledge!
Links to earlier luxmeter posts:
https://budgetlightforum.com/t/-/26151 (april 2014, construction of a Ceto luxmeter sensor 
, see post #2 of today's thread for more chinese luxmeter-sensor construction fun!)
https://budgetlightforum.com/t/-/28689 (sept.2014, general information about luxmeters, test of Mobilux ClassA, Tondaj LX-1010B, Ceto CT1330B)
https://budgetlightforum.com/t/-/33532 (may 2015, an attempt to use the ambient light sensor on my phone as a luxmeter)
https://budgetlightforum.com/t/-/51032 (april 2018, test of Mobilux ClassA, Tondaj LX-1010B, Uni-T UT-383, Tasi 632A, noname LX-1330B)
The four new luxmeters:
Extech LT45, bought on Amazon for total 370 dollar shipped plus import fees (it is much cheaper in the US, buying US stuff in Europe is very expensive)
Tasi TA8132, bought on Aliexpress for 18 dollar
Uni-T UT382, bought on Aliexpress for 44 dollar
SmartSensor AS803, bought on dx for 16 dollar
*The Extech LT45 meter I had high expectations of because Extech meters are used by many flashoholics who seem to take their hobby serious. I bought one without NIST-calibration certificate, because I was more interested in the complete spectrum response than a piece of paper proving a calibration against a 2700K tungsten lamp (I'm sure that the factory calibration of a manufacturer like Extech will be done well enough). A useful feature of the LT45 version is that it has the option to enter your own calibration, which comes in handy in integrating sphere use, to incorporate the sphere multiplier directly into the luxmeter read-out.
*The Tasi TA8132 meter is one of the newer models by Tasi, since the expensive Tasi 625A tested quite well in april it is interesting to know how their very cheap models behave.
*The Uni-T UT382 meter is the other way around, the ultra-cheap UT383 that I tested in april did not do bad at all considering the price, how does this 44 dollar model with USB data transfer do?
*The SmartSensor meter seemed the cheapest of the cheapest to me, does its performance surpise?
First a (for flashaholics 
) real world comparison: of a number of flashlights with different tints and CRI the throw (brightness of the hotspot) at 7 meter was measured with the different luxmeters, including my Mobilux luxmeter as a reference, and as a second reference the reading of the Gossen Mavolux Base (this spectrometer is also a class B luxmeter). Notice that to get stabile output (=less than 0.5% output drop over the course of these measurements) most flashlights were not at the highest mode , so the numbers have no absolute value, just relative between the luxmeters. The calibration of my Mobilux meter was corrected to a well-calibrated unit earlier this year (mine read 3,4% low), this correction is incorporated in these measurements. As a new feature I recorded the spectra of the measured hotspots :-) :
Luxeon V 4000K 70CRI (Thorfire VG10)
|
XM-L2 6500K 70CRI (Sunwayman D40A)
|
XM-L2 3000K 90CRI (Kaidomain K2)
|
dedomed XP-G2 S4 3D ( "Mitko Thrower" )
|
N. E21A 6500K R9080 (Olight M10 Maverick)
|
219B V1 4500K R9080 (Jaxman M2)
|
Epileds 400nm UV-led (Ultrafire C10)
|
|
| MobiLux Class A | 61,7 | 289 | 42,3 | 1108 | 131,1 | 122,5 | 2,17 |
| Extech LT45 | 64,9 +5% | 309,1 +7.0% | 43,0 +1,7% | 1167 +5,3% | 137,4 +4,8% | 126,6 +3,3% | 26,12 |
| Tasi TA8132 | 59 -4,4% | 300 +3,8% | 41 -3,1% | 1052 -5,1% | 150 +14,4% | 131 +6,9% | 9 |
| Uni-T UT382 | 62,8 +1,8% | 364 +26,0% | 40,7 -3,8% | 1144 +3,2% | 180,1 +37,4% | 144,2 +17,7% | 22 |
| SmartSensor AS803 | 39 -36,8% | 266 -8,0% | 24 -43,3% | 902 -18,6% | 127 -3,1% | 104 -15,1% | 119 |
| Gossen Mavolux Base | 56,8 -7,9% | 281 -2,8% | 37,9 -10,4% | 1075 -3,0% | 126 -3,9% | 114,6 -6,4% | 15 |
The results are clear:
*despite the substantial purchase price the Extech LT45 luxmeter is good value for your money, you can express the difference with the Mobilux meter over the numbers obtained for these measured light sources as a 4.5% calibration difference and 1,35% variation in spectral sensitivity. This is closer than how the classB luxmeter read-out of the spectrometer compares to the Mobilux classA (with 5.7% calibration difference and 2,5% spectral variation).
*the surprising performer is the Tasi meter, for 18 dollar I got a luxmeter with on average 2.1% calibration difference and 6.3% spectral variation, compared to the Mobilux meter. I calculated the corresponding numbers for the LX1330B in my april test: 3.2% calibration difference, 8.3% spectral variation. So I take this Tasi meter any day over the LX1330B. But beware!, you may be in for a disappointment, in post #2 of this thread I describe another Tasi meter from the same series that looks the same but is built much much cheaper and although I have not measured it, it probably has much worse performance, you will not know what you buy!
*the surprising looser is the 44 dollar Uni-T UT382 luxmeter, measurements are all over the place, this is a very bad luxmeter.
*the SmartSensor is cheap and bad.
To get a grasp on why the luxmeters do not just show a calibration difference but also read so different for different light sources, just as in the april 2018 test for these new luxmeters I recorded the spectral response to a 2700K tungsten light source. This should give an idea of the differences in spectral response of the various luxmeters and may explain why they read so different for light sources with different spectra. For a description of the method, using an antique but strong-going Zeiss prism-monochromator, see the link above to the sept.2014 test.
First I show some validation data of the method, just because I want to show that my home-made contraptions actually do work.
Every time I do a series of tungsten lamp spectral response tests on luxmeters, I have to build up the measuring rig completely (me and my family live small, my allowed hobby space is 2.5 square meters of the appartment) align the light source with the monochromator again, and power the tungsten lamp with the exact same current (6A). This alignment is a bit different every time, and so is the read-out of my Mobilux meter that I use as reference. To get an idea of the robustness of the method, here's a comparison of the tungsten lamp responses of the Mobilux meter of the april2018 session and the sept2018 session, with the highest read-out values normalised to 100:
That looks consistent enough to me, what I see mostly is a shift of about 2nm, differences with tungsten lamp responses of other luxmeters are way higher than this :-)
A valid question is why do I use the Mobilux luxmeter as my reference, as if it shows the absolute truth. It is an instrument after all and will have its errors and tolerances. First and mostly: because it is the best portable luxmeter out there, and it is recently calibrated against another Mobilux luxmeter that is being officially calibrated each year. I have no option to do any better so it will just have to do. But to give an idea how good the spectral response is, I confidentially received a test rapport of this type of meter showing the errors of the spectral response compared to the reference V-lambda curve, and without showing the actual (confidential) data, here is the tungsten lamp response of the Mobilux meter in the visible light range compared to the tungsten lamp response if the Mobilux had a perfect V-lambda response:
This is assuming that the Czibula&Grundmann test data are trustworthy, which I have no doubts of. You can see that they managed to make a detector/optical filter combination that comes extremely close to the V-lambda response.
So far the validation, here are the tungsten lamp responses of the 4 newly tested luxmeters, together with the Mobilux luxmeter:
I find it extremely difficult to pinpoint, by analysing these curves, why the luxmeters read the light sources in the table above exactly like they do. But the general picture that arises from these curves is more clear:
*the Uni-T UT382 luxmeter and the Smartsensor AS803 luxmeter both deviate immensely from the reference luxmeter for some light sources in the table, and for different reasons as you can see in the graph: the SmartSensor's spectral sensitivity shows a large shift, on average 20nm, to the right compared to the Mobilux, and the Uni-T's spectral sensitivity is just completely weird and nothing like a V-lambda response.
*the extech LT45 luxmeter has a tungsten lamp response that matches the Mobilux response pretty well, shift and tilt the curve just a bit and you are there, and the Tasi TA8132 meter does not bad either, it overreads blue and red somewhat.
One more graph, a couple of interesting curves from the april and september sessions combined in one graph:
The LX1330B has the worst curve, followed by the very cheap Uni-T UT383 (that in itself does way better than the weird newly tested Uni-T UT382 which I left out in this graph), the two Tasi-meters seem to perform similar, their curves deviate differently from the Mobilux curve but the amount of difference seems comparable.
Conclusion:
*I trust the Extech LT45 meter to be pretty good, the tungsten lamp response curve makes me think that also other white light sources would be measured pretty accurately.
*The Uni-T UT382 completely fails to perform, do not buy. I may open it up at some point to see if it has a shifted optical filter so that light can enter the sensor bypassing the filter. in that case the quality control has failed.
*The SmartSensor AS803 is cheap and performs bad, even worse than the LX1330B. Even for the price it is a bad buy.
*The tested Tasi TA8132 meter performs surprisingly good for what you pay. It would be very good value for money if only Tasi was not in the process of cheaping out their TA813x luxmeter series and you may be buying a low-grade self-clone from Tasi. (see post #2 for the explanation)
Edit: more luxmeter dissection and tests in post 2, 18, 22, 30, 49, and look through this thread for Docc's and other's thorough crusade into where to find the "good type" cheap Tasi meter 
Thank you!
