It appears LM4 are “batch calibrated”, so If I understand correctly Opple chose and calibrated a “golden device” which represents the typical characteristics of an entire batch of sensors with similar sensitivity.
I don’t think they are “device-calibrated” one by one.
Edit: not true
Yeah, that’s exactly the way Steve chose, older units will be fine
Thanks for joining the gang, Cod3r!
well, we actually calibrate device by device. we first calibrate one so called “golden device” against professional spectrometers like cl-500a or cl-200a, and then we developed a fixture, then we calibrate each one of the mass-produced LMs using this fixture.
the fixture is a black box containing 3 lights with different CCTs, we developed a desktop application connecting with the fixture through WiFi, and put the to be calibrated LM4 inside the black box, the 3 lights will light up sequentially the LM4 would transmit the sensor reading to the PC(reading 3 time corresponding to 3 lights, each reading would contain all the channels value), our desktop application on the PC would calculate a group of coefficients and then send back to the LM4, LM4 would write those coefficients on the non-volatile section of the flash chip.
so basically, each device is calibrated in the factory. you align one of the LM4 from a batch of LM4s to the cl-500a and called it the Golden Device, at the factory each one of the mass produced LM4 is aligned against the Golden Device.
so basically Light Master IV is way more complicated than some people may think. It’s time consuming. and I encountered a lot of production issues too.
each LM4 would have different coefficients,
the coefficients could not be updated through a software patch, either embedded software or iOS/Android patch.
Interesting, I stand corrected then. I was wondering, is there a chance that you can get your hands on your colleague’s Golden Device? It would be a shame to waste the factory calibration work
not easy, first you must get the so called Golden Device and then takes it to another city to measure some data using sophisticated professional device called monochromator and then do some calculation. so in total 3 steps, we contact ams osram through agent, ams osram & agent they just ignore our request frequently, probably because the chips we opple ordered is quite limited, I don’t know the exact number. perhaps if you order 1 million USD chips, they would pay attention to your requests, otherwise nobody treat you as a customer.
btw, opple doesn’t have that device.
Ok, so if I understand correctly, calibrating the GD against a professional spectrometer is not enough and it also requires measurements to be taken with the help of a monochromator. Sounds complicated, seems like it would be impossible to pursue the first method. Guess we’ll just settle with the third method 
well, for some metrics like cx,cy/CCT I think the calibration against a professional spectrometer is enough. this part is done by another colleague, I also did this part with the help of monochromator. but currently for this part, my colleague’s algorithm is incorporated in the iOS/Android app not mine. and I don’t know the precision for those metrics. I haven’t evaluated mine Vs. my colleague’s. for some metrics like CRI, I’d highly suggest using monochromator for calibration, because the calibration method in official documentation is outdated.
Ah, it’s a pity then. So I guess it won’t be possible to improve the accuracy of all current LM4 devices through new app versions or firmware updates.
Opple will need to implement the results of your recent work as a new device or a new hardware revision.
But thank you for clarifying.
I think LM4 can still be helpful to people even in its current state. At least for comparative measurements, to answer questions like “is this light source better than the other or not”.
That’s why I’ve ordered this device myself. I would prefer LM3, but it’s not available to me for a reasonable price.
Unfortunately I missed the opportunity in 2022 and 2023. Only started to be interested in this topic a month ago.
So I modified a few versions of the Opple apps for Android - Smart Lighting and Home.
What’s implemented:
− Removed the need to turn on the Internet and register an account
− Light Master page is opened on startup
− Removed all unnecessary permissions
− Removed some analytics and push services
− Removed Opple watermark from the background
Download mods for OPPLE Home:
− 3.3.1 v1 
recommended app and version, see below and here
− 3.5.1 v1
Download mods for OPPLE Smart Lighting:
− 3.3.2 v1
− 3.5.1 v1
− 3.7.1 v1
These APKs can be installed together with the official apps and next to each other. There’s no need to uninstall the originals first. A version for LM3 can be found here.
As other people mentioned, unfortunately you can’t get rid of the GPS permission if you need to scan for available Bluetooth devices. See the official documentation here: 蓝牙权限 | Connectivity | Android Developers
I’d like to say that I don’t want to harm the company in any way and I don’t seek any monetary benefits. In fact, I hope that more people are going to start buying Light Master devices now that these improved apps are available.
And I also have big respect for people like Steve who find time to come here and share their thoughts and insights with the public.
That being said, it’s obvious that user experience of Light Master owners suffered greatly when new apps for LM4 were introduced. Since they are universal for all Opple products, they ask for a lot of permissions, including the Internet connection, and even require to register an account. Startup time and overall performance is quite slow too.
That’s why I decided to try and bring back the behavior that could be seen in Light Master Pro app. You click the icon, the measurements page is opened. That’s it.
I took a few different versions as a base because people mentioned how they can produce different results. See here and here.
My LM4 is still not delivered, but I asked other people to test out these APKs and they confirmed everything works fine.
If you have time, please also test them out and report if you see the same or different results in different versions.
The only drawback for now is: reports saving doesn’t work, since they are implemented as web pages stored on Opple servers, and I disabled any communications with such addresses. But I’ll try to find a solution in version 2.
I also created a Telegram chat to further discuss reverse engineering of Opple Light Master 3 and 4. You can talk in both English or Russian there.
Wait, please note that Steve proposed 3 different paths for improving the accuracy:
Steve chose method 3 and seems it’s going to work as intended
your understanding is exactly what I expressed.
actually the coefficients are calculated as the average of 3 different lights(under 2700K, 4000K, 5700K nichia Led), remember the fixture I posted yesterday? the three Leds would sequentially turn on and off inside the fixture.
I think one of the characteristics of the chips’s inherent property is non-linearity, for example under those three CCT Leds, each would derive a group of coefficients(for each VIS channel, so in total 8), and they differ slightly actually, but not very much. so if my own GD is aligned with the real GD with a group of coefficients, at least under 2700K/4000K/5700K, it will behave nearly exactly as the real GD. I need to check other CCTs due to the non-linearity, but I think basically it won’t differ much.
you are great, you know how to reverse engineering, I have no idea of how android works.
why can’t app on Android and iOS can’t easily invoke my cross-compiled dynamic linkage library(.so or .a format, written in C/C++), sometimes we’d use old versions of Android which doesn’t support iic communication, I compiled .so files for older versions of Android, still app can’t use/invoke the library directly, my colleague said JNI interface needed. don’t know how it works.
on PC platform ,seems every language can invoke the functions residing in a C/C++ library.
I’m afraid I don’t know much about Android too, since I’m not an Android developer 
I’ve just tinkered with some APKs in the past, so I applied my limited knowledge here as well.
But hopefully your colleagues will help you with the compilation. If the library is cross-platform, it should work on Android and iOS devices too.
Also I’m sorry to bother you again, but is it possible to allow access to the library you’ve shared above? Opple Light Master 4 discussion thread (new 2023 model) - #955 by stevechang
I’m not sure if I’ll even be able to use it (I don’t have a FT232RL device yet), but maybe I’ll try it out in future.
My email starts with “radrad” if you decide to approve the request.
cross compiled by myself. currently iOS library not available, still need Mac computer to compile it.
https://drive.google.com/drive/folders/1V5bvOdlhajkRVi4jpVYMNp2K-imvQRe8?usp=drive_link, you can not access it?
I am working on a Windows executable, actually serial connection already near done. want to add Bluetooth connection.
I can confirm, access to the file needs you accept the request
i copied link as in above graph, I don’t know why you can’t access?
plus, I didn’t received your request, where is the request? 
I will get a LM4 in the next 2 weeks.
Hopefully the modified app works fine. No interest in creating an account for Opple things 
@stevechang If you need help with spectrometer and many different LEDs…
In the same submenu where you copied the link, press on the “Share” button and switch the access setting to “Anyone can view”
