Do you want to use the LM4 to measure the flicker of a computer monitor?
1 Thank
Obviously, the device does not see even a flickering LED in front of the aperture. How can it measure monitor flicker reliably then?
besides the white aperture there lies a sensor which is responsible for collecting flicker data. but I am not very sure whether it can measure computer monitor, it actually sampling at a time interval of 11us, 25us, 146us, and according to Nyquist sampling theory, you can infer the maximal frequency is about at 1000000/22, 1000000/25 and 1000000/146. we have three gears, Light Master would first sample at around 25us, and then determine whether it needs to change lower or higher gear to be more accurate.
3 Thanks
the proposal #3 works, the accuracy is just very near to the so called golden sample. the difference is about 0.01 according to my test of a Light Master comparing to the so called golden sample. so the problem lies in how to port C++ code to JaveScript which can be used in both Android and iOS platform.
maybe considering cross compile C++ to arm8 or arm9 compatible dynamic linkage library(.so or .a file extension files, because both Android and iOS have some some Linux or Unix heritage, so I guess it may works.) I know nothing about JS and app development.
I think the metrics basically achieve its full potential, and in most scenarios it meets your requirements according to my test.
however, I can not guarantee it is as accurate as an IS, after all it is a low cost chip, only around 2~3 U.S. dollars. I have full confidence that LM4 outperforms LM3 definitely, will do some tests later. by the way, this update will only involve Ra, R1~R14, and CS maybe eml. lux/CCT/cx,cy/u,v/duv those metrics I will not update in this version, because it is of another colleague’s responsibility. by the way, I never tested lux/CCT/cx,cy/u,v/duv, how is their accuracy?
4 Thanks
or maybe consider chapgpt to transform c plus plus to JavaScript, you guys may be having to wait for some time, even I start to push for update there is a tedious process, you have to cooperate with other staff.
2 Thanks
Well I can confirm that V3 does measure monitors and screens of phones etc. But at max brightness, the lower you get the worse the results. So I will test with removing the diffuser, any guides how to take it apart?
Basically the features why the LM3 was mostly bought (x/y, duv and even CCT) are not in the new update? This is very unfortunate.
So no reliable color measurements are possible?
if you intend to measure flicker of monitors and screens of phones, no need for removing the white diffuser. flicker sensor is exposed to outside environment directly, you actually can see it beside the white round diffuser.
2 Thanks
so you mean currently lux/CCT/cx,cy/u,v/duv are not accurate too?
I didn’t expect this, those metrics are calculated by another colleagues, but I’ve already calculated those metrics too, just have not verified yet. then another problem arises, will he think that I am going to replace him and get annoyed towards me?
1 Thank
I don’t know because I don’t use an Opple myself (I went straight to spectrometers).
However, I imagine a correct calculation/measurement display to be very difficult if there are only a few data sets from the sensor are available and the spectrum can hardly be reconstructed. It is enough that the cyan hole is less pronounced or that the peak is at 430 nm instead of 450 nm to give higher error margin. In the end, the problem is simply the sensor, 10 (or 12) datasets are far too few to get accurate spectral reconstruction and therefore somewhat precise measurements…
I’d suggest you guys compare lux/CCT/cx,cy/u,v/duv measured by LM4 and spectrometer, if it is not correct, I’d ask somebody to compile or package standalone Android or iOS installation files with my version of algorithm for calculation of those metrics.
I think actually those metrics are generally easier to calculate, no need for spectra reconstruction(of course if you’ve got accurate measured spectra, it’s easily to do too). mapping from 8 VIS channels to XYZ is entropy decreasing process if I understand correctly, so basically it’s much easier. If you’ve got XYZ tristimulus then cx, cy/CCT and others are calculated directly, there are formulas out there. actually this is called XYZ calibration as specified in the official documentation. conversely to calculate cri or cs/eml is difficult because it needs spectra, and as7341 can provide 8 VIS channels, while IS has 81 or 401(5nm or 1nm interval, if the starting from 380 to 780 nanometer). from 8 to 81 or 401 is entropy increasing, definitely you will not always reconstruct precisely as IS measured, sometimes if the curve is complicated, the details can’t be recovered.
4 Thanks
for my office environment, my calculation of cx,cy/cct is(0.38367, 0.36876) and 3863.8, my colleague’s is (0.38474, 0.37356) and 3871.3
1 Thank
If you send me an LM4 I can make tests (and send you the data/comparison) with my huge LED collection, which has everything from 60 CRI to 98 CRI and 1800 K to 28000 K LEDs.
That would also be an occasion to finally take care of coating my integrating sphere with barium sulfate 
1 Thank
I don’t think Steve is allowed to do that. On my part I ordered the LM4 yesterday because I found a good deal and I am confident of Steve’s work. Looking forward to trying it out.
1 Thank
Ok, so you have to buy it again if ever a reworked version is available:
you are spreading misinformation:-)
this is just one of the proposal that I gave, and considering maybe thousands of LM4 are already purchased by users. definitely I will not take this way, I am just give you a full picture of alternatives. I can actually make the other colleague’s golden device immediately useless and mine as golden device if I don’t consider all the users who have already purchased Light Master IV.
3 Thanks
Ok, then ignore my post.
That’s what it sounded like. Thanks for the clarification.
1 Thank
already cross-compiled library for Android named “libcolorimetry.so”, but now app team has no time to integrate my library(perhaps in one month they are not available), besides the library, also need to write JNI thing which I haven’t done before.
there is also a windows version of the library, anybody want to try? you can modify a Android charging wire to achieve serial communication with the device, and I will give you an executable version of the library(Microsoft Windows executable file).
4 Thanks
As soon as my LM4 arrives I can try the Windows version, I’m sure other users are also interested. Please share the executable when it’s ready.
I assume a RS232 Serial-USB adapter is needed
1 Thank
Windows executable is ready now, I will post it to maybe google storage or dropbox place.
you can do a wire like below(use type c wire, I think just normal Android charging wire):
the library inteface is as below(for those who wants to customize UI):
parameter “channels” is LM4’s readings of channels value, and parameter “coefficients” are actually calibration coefficients and actually it’s constant, stored on non-volatile flash, those two parameters you can retrieve by sending commands like “Test getscale” and “Test getvalues” once you connect LM4 with Windows, use a serial tool on Windows. baud rate is 115200.
my setting is as below:
11 Thanks