So, I’ve been reading some posts here and decided to ask.
Should I get a LM4 now or wait?
I’ve been wanting one for some time now.
It doesn’t really matter if you get it now or later. The update is going to work on every LM4 ever sold, so don’t worry.
Steve said that with the next update LM4 is going to be superior than LM3 (which we know is good enough), but right now its readings aren’t reliable and currently LM3 is better than LM4. If you are willing to wait for the update and/or if you get a good deal on the price then you can buy it. Otherwise you can save your money and wait 
1 Thank
Good luck with getting improvements to LM4 
Honestly, your negativity isn’t really helping anyone. All you did was complain. If you want to help, since you said you have a full fledged spectrometer, you can buy an LM4 and do some tests for the community to validate Steve’s own results. Otherwise, I don’t see your comments as useful.
2 Thanks
Thank you Sam. That’s what I wanted to know, whether it was a hardware or software thing.
I’ll be on the lookout for any sales and finally buy one.
Cheers
1 Thank
I certainly understand your point. Active participation is always better than just complaining. But I think it’s a sensitive issue. I simply find it difficult to ‘work’ for a Chinese company “just like that” without getting anything in return or at least benefiting from it. In the end, it’s primarily Opple that benefits because they can sell the device more profitably thanks to the data and more precise results or the number of units sold increases because of that.
Especially considering that the device itself and the app are not open source (correct me if I am wrong), I think this is even more questionable. This would be really cool because it allows for much more collaboration and more features or more possibilities to improve the device from the start. And then I would happily buy some devices for testing because I know this would help to keep the whole project alive, like projects as Pinecil, which is Open source and held by a large community. The LM is nothing of this, unfortunately.
But this is not the right topic to discuss this I think.
2 Thanks
That makes sense, an open source device like this would be awesome, I agree. Unfortunately, the manufacturer is not going to share the source code and schematics of the device. It’s true that they are going to profit from this, but without Opple (and Steve) currently people like me would not have the chance of getting their hands on a tool that does what the Light Master can do. I do not need and I can’t afford a thousand-euros spectrometer just to meause and match CCT and CRI of the lights inside my house. My point is, if you are not willing to help, and I certainly understand why you might not want to, then I don’t think it’s fair to make disparaging comments here.
In the end, it’s a win-win situation. Maybe one day we can learn something from this and try to build our own FOSS light meter, and I will try to help for sure
I also think that many users like myself are learning a lot about colorimetry and spectral color sensors thanks to this thread and AMS application notes, I find this topic really interesting as an electronics engineer with no lighting experience.
2 Thanks
google drive link: https://drive.google.com/drive/folders/1V5bvOdlhajkRVi4jpVYMNp2K-imvQRe8?usp=drive_link
libraries only now, actually Python can easily invoke a dynamic linkage library, and Python actually supports struct data structure. maybe later will upload a command line executable file, but currently I am working on a Qt project, will incorporate my dll for calculation. maybe bluetooth and serial connection supported too.
@SBT41 Sam, thank you for your understanding.
6 Thanks
plus, I want to open source it. I tried, I talked with my supervisor/leader, he said it’s a company level thing, and considering we’ve signed some NDA documents with third parties, it’s really out of my capacity. I personally think the revenue is generated by selling the real object(LM4), no need to hide the details, but it’s own my personal view on this problem. by the way, ams osram also didn’t disclose their algorithm, the algorithm they specified in the documentation as below is outdated(ams osram Shanghai staff told me in front of me), and you can verify it by yourself, it didn’t generalize at all.
2 Thanks
if you follow the algorithm, you will generate result as below(I will draw a graph, actually it’s a matrix):
not like the graph as below which I think it demonstrates some sense of beauty:
2 Thanks
But why is AMS selling a sensor without telling anyone how to use it at its best if they already developed an algorithm? What is the reasoning behind this behavior? It’s not like they are selling a finished product that competes, they just sell the sensor. Just curious
I can certainly understand that, but the problem here is that in the worst case, all the work of the community can be wasted in a short space of time. Nobody knows how long the “Light master” project will be supported by Opple, nobody knows what the sales targets for this product are and nobody knows what strategy Opple is pursuing behind this product.
I am definitely not a fan of supporting a project where it is not clear how long it will be kept alive. And since everything is closed source, you can’t even take the remnants of the project and make something of your own out of it.
And for me, that’s a major problem here. In contrast to a lamp, where a community member gives a few tips or does small tests, this would be a time-consuming and lengthy project that takes a lot of time. I want to make sure that the knowledge is not lost if a company decides to simply discontinue the product and no longer offer support. (This does not refer to @stevechang but to the upper management who decide whether the device is still sold or not). This happens a lot with smart home components: manufacturer no longer wants to sell or has missed out on sales? Simply discontinue support, remove the app from the App stores and - in worst case - directly turn the device into electronic waste.
Btw, I’ve just had a look at how expensive an LM4 is.
It is basically impossible for me to get an LM4 in Germany from a reputable source. Aliexpress doesn’t even show me a device (probably because these sellers don’t offer any shipping to Germany), and Amazon has an offer that I don’t even know if I have to pick it up at customs or not since it would be sent from USA it seems.
In this respect, I would have to rely on the help of someone here to get such a device anyway. 
2 Thanks
I can sell you mine, Germany here.
Bought it in the US last year, but really don’t use it anymore.
1 Thank
So we have to create a slightly changed algorithm based on reallife measurements of many different LEDs. This seems to be the main thing here. And yes, I think all (or mostly all) of these issues can be solved.
I checked the general design of raw data on certain channels (F1-F8 VIS, data from my measurements, I did not put the FWHM into account for now):
These are the spectra which would look like if they would come straight out of the sensor. Basically the sensor is not bad. I think most LEDs can be measured quite good, like 519A sm573 R9080 or also the low CRI counterparts.
I see a problem with some LEDs tho. Some LEDs have the blue peak not at 450 nm but on 430 or so. (Like optisolis emitters by Nichia. The NF2W757GT-F1 has 420 nm peak and a hole at around 440 nm which would give huge errors. Even the F1 channel at 415 nm cannot get the correct value.)
This can led to measurement errors since reconstruction of the original spectrum is basically impossible due to too few channels. Also there are lighting sources like CFL which have totally different spectrum. Even if we say ‘only LEDs should be measured with LM4’ this would be a big challenge to get all the properties of these spectras right.
1 Thank
@koef3 thank you for your checking. I had noticed the F1 channel problem at the end of last year. even F2 channel has this issue too, the peak of F2 channel is at 445nm.
by the way, in my opinion there is slight difference between the simulated spectrum and the reconstructed spectrum, let me take F1 channel as an example, the output of the sensor is approximately the inner product of the real light spectrum and responsivity, so even F1 channel reflects some characteristics of other parts of the whole spectrum. I think it can’t be reflected in the line graph you draw, because it is discrete. in a word there is minor difference, my understanding may not be correct however. I have a CFL lamp in my office, I will verify it next week.
I will continue to work on this project at least for the year 2024.
I agree with your point that it’s time consuming. if I were in germany, I would give you my own LM4 for testing. unfortunately I am not.
4 Thanks
It was nice to meet you yesterday! 
2 Thanks
Same here
3 Thanks
I wonder, does each LM4 device have different coefficients stored, based on calibration? Or maybe all of them contain the same values?
Because in theory it should be easy to modify these values after retrieving them from memory.
So after you finalize what coefficients are the best, the team behind the Android/iOS apps can apply modifiers to these values and use them for all calculations and charts drawing.
This way already sold LM4 devices won’t become obsolete. A software patch should be enough.
BTW, I tried to request access to this file, but it looks like you don’t share it with everyone? Should I PM you if I’m interested?