Using OPPLE Home app and Light Master Pro 4, my new ZebraLight SC700d HI shows Ra = 84.0.
Using Light Master Pro 3, Ra = 92.3.
I’ve owned this Light Master Pro 4 device since May 2023, and I am confident in saying it is going in my garbage bin. I’ll never be able to trust it.
Thanks for all of the information on this topic, Steve, but I’m officially out. Take care.
sorry, LM3 seems right according to the specs.
After playing around with the LM4 for a bit, I noticed a few things I would like to see in future updates.
Export reports needs a rework.
In general, please give the option to export everything the LM4 measured as .csv (split up in multiple files maybe? Not sure) - with raw data files I can work. I can create a script that will automate data evaluation and plotting into charts, even in Javascript charts for my homepage if I should decide to do reviews some day. Right now I’d have to screenshot all data an stehen manually enter it in a spreadsheet, and I have no way to work with raw flicker data.
What works well:
I understand that your device is mainly aimed at your customers to see if their lights perform as they should - none of them will ever need the advanced export features I requested. But for power users and us flashlight people those would be a great help, and make the LM4 a lot more powerful.
basically, whether adding new functionality/feature is decided by product manager. app team is in another city, and they rarely pay attention to this forum. I am doing a lot of extra stuff, and they may think I am burdening them with new features request.
sorry to say this reality, I think everything related to people relationship is complicated.
It all depends on exactly what your company wants. There is the engineer’s approach - I invented this, and you may or may not buy it. And there is the approach focused on the client and his wishes. In case of the first option your product will not stay on the market for a long time, unfortunately. IMHO. There is at least one more forum with discussions of LM3 and 4 and there conclusions are the same ±. (((
last time, I prod the PMs(one for "欧享家”, one for “欧普智能系统”), the just ignored my suggestion, making me kind of awkward. I think if we want to do a great product, the PMs should visit this site, listening to our users and making friends with users.
there is a critical coefficient, and the LM3 left without a smooth transition long before LM4 started,so nobody knows how to calculate this coefficient for a new chip, if there was no frequent reorg before, or there was a good transition, LM4 ought to behave better maybe.
feel really sad.
Steve, could you reach out to the original SME(s) and extend a short term / part time contract for knowledge recovery of this coefficient? Of course, you will have to sell this to the people who control the money. You might have to put together some business proposal to cost justify it.
actually I chatted with the engineer through wechat who was responsible for calculating this coefficient for LM3(wechat is Chinese version of whatsapp), the engineer expressed that he wanted to come back to opple. and he took this coefficient for bargaining chip for returning, but in the end he didn’t return.
someday later I asked HR, she told me that this company has face to save because years ago when the engineer decided to resign from this company, the big boss tried to persuade him to stay but in vain. so now you know the dilemma I described before, that guy is a true expert in light spectra. We do have other guys who know light spectra well, but they just happen to know nothing about that coefficient.
our current method of calculating is actually based on AI(machine learning), that coefficient is critical for bridging the machine learning model and the real sensor readings, in the future we’d use spectra reconstruction method, we can skip the calculation of this coefficient.
I think AI method is great, because code is compact & the speed for refreshing the UI is very quick, but just face this coefficient hurdle.
for LM3 the coefficients are just plain real numbers, but I really don’t know how they are calculated, I tried a few other methods, but as you guys know, the CRI/R9 sometimes goes wrong for some spectra. this really makes me can’t sleep last year for sometime, I know programming and machine learning, but sorry I don’t know spectra very much. I think this is not my fault, but I still want to solve this problem, so you know how stressful I was. I was crazy last year.
Wow that’s crazy! I don’t know how things work over there, but in the US I would sign an agreement with the company that hired me, typically stating, something like, the intellectual property I create belongs to the company. If this engineer did the same thing in the US, most companies would litigate … and they would be in big trouble.
I’m not an expert on colorimetry or ML, but I would think that the physics are well known, and a mathematical model would be better than an ML one … I’m worried the ML model would be brittle, not generalize well, and be unexplainable.
I feel really bad about this. You didn’t have to engage this community but you did … you did far more than many other engineers would have. And you care about your product and have worked very hard on it. You are a good engineer and your company is lucky to have you.
but he already left the company for a long time, I don’t know the litigation thing.
ML method is very ingenious in its way, actually ML method doesn’t reconstruct the spectrum, but it does use spectrum information.
thank you for your understanding
I don’t just mean the inaccuracies of the LM4. There are unpleasant aspects of using LM3 as well, which cause irritation and actually make you give up using it, buying something more expensive. Competitors will sooner or later make devices in the same price range, but without such unpleasant features.
Sometimes we have to make lighting measurements in places (e.g. rooms below ground level, etc.) where there is no GPS signal (LM3) or internet access (LM4). And then we get an absolutely useless device… Why do we need it, even for free, if it is useless?
for LM3 there is an app named “light master pro”, I installed it on my own phone. but it is discontinued
I’m using it. Without access to geodata it doesn’t start. And if I turn off the GPS after starting the app (or when the signal is lost), it sometimes disconnects.
I fundamentally do not want to buy LM4.
A couple of us are trying to extract data from the LM3/LM4. Your requirements are fairly simple, if I’m ever successful I will add them to my beam logging app (work in progress, Android) - it may not be possible to read flicker and spectrum at the same time with this sensor.
flicker data is measured by another sensor, equivalent to bpx65.
later, I will try to open the protocol, so you can retrieve the flicker data.