Right. The marketing was/is inaccurate. I think developing the camera functionality turned out to be harder than Purism expected, but fortunately it’s getting there now, thanks to the efforts of @dos and @dcz (and maybe others I’m unaware of?).
But yeah, the marketing should have been revised a while ago, in my opinion.
I love 2400K for late night lighting, but unless I am shooting pictures in a candle lit environment (or lighting with other 2400K LEDs), it should be a little higher. I wonder how hard it would be to desolder 1 of the 2 COBs and replace it with something 5000K-ish. The current might have to match the existing COB. The LM36922 allows each LED to have different voltages, as it regulates by current. In the datasheet, register 0x10 lets you enable or disable the LEDs individually. I tried reading this with the i2cget tool, but I would have to recompile the kernel to disable the driver to get direct access. The Linux kernel driver hard code’s both LEDs to be on or off at the same time. So, with the right hardware skills and software skills, this feature can be hacked in.
I still like the 2400K for a flashlight late at night, although the CRI might be a little off. However, using it for illuminating stuff in dark corners during the daytime hours is less than ideal.
My quick and initial “preparation” wasn’t quite successful:
white card balanced 
) light color at exactly 5000K/CRI95 (I’m quite sure):Not quite sure what you’re showing there though - gain at 32 is extremely low, probably only useful in bright sunlight.
At first place I wanted to take screenshot of millipixels user interface (in version 0.20.0-1pureos1) anyway (to this thread related version) and realized that my permanently installed light bulb of 9W/2700K with 60 degrees focus reflects or bounces back too much light from flowers (to much contrast) and even lowest gain of 32 wasn’t helping much (as this light source was very near in distance) there.
After replacing first LED light bulb for the second 15W/3000K LED bulb in regular 270 degree angle picture taking photo (at same above my table lamp distance) looked quite differently (to the Librem 5 main camera). And yes, this clarification was needed as my initial indoor light source (first lamp) needed to be purposely replaced (not to move this table with those flowers somewhere else). Thanks for asking!
I’d rather measure the output before posting any numbers, I measured the output from two separate Librem 5s, both were in the region of 2400K, below is the summary results from one of the captures…
Agreed, the CRI results don’t make great reading, as a simple torch LED it isn’t really much of an issue, as a camera flash source however, it’s not going to be winning any prizes. In today’s world, I wouldn’t hang my hat on CRI, there is a number of problems with CRI in general and although not completely meaningless, it doesn’t provide a complete picture, it has been largely abandoned in favour of TLCI (the Librem 5 torch LED measures 37 for TLCI). For white light sources SSI against the relevant CIE Standard Illuminant is also becoming a common reference metric.
I find it strange to see so much talk of colour temperature with no mention of [white] balance. For a quick and dirty demonstration, below is an image of the meter used in the measurements, the image is taken with a Librem 5, zero ambient light, so it’s being lit purely by the torch LED at 100% intensity, the [white] balance is set at the default “d50”…
Doubtful, all of the major manufacturers have put out technical papers detailing the difficulties of both manufacturing LEDs to tight tolerances and accurately measuring the characteristics, in short, no manufacturer would commit to exacts.
That is some great information.
I have not mentioned white balance because my main concern is not having all 2400K light sources. Thanks for pointing out that if we are going to have a 2400K flash, we need a 2400K white balance setting.
Personally, I would like to push for a dual tone flash that can work from candle light to sunlight. Not that we would use a flash with sunlight, but some artificial sources mimic it. It seems like the hardware supports selecting one, the other, or both LEDs. We just need to patch the kernel driver and make Megapixels use it. However, it seems like the hardware does not support dimming the LEDs individually, although there might be a way to set the maximum current individually, but I have not seen that feature.
Nice, yet now that we know (very important post #84 related) that R9 is at around 0%, replacement of current COBs would be mandatory (those are of standardized size anyway) for the next devices (yes, please, as rather no investment than buying poor product, if and when making some photos related) even before any workaround in sight to be taken.
It’s not that uncommon. Flash can serve as an additional, not primary, source of light. I don’t know if anyone uses this technique with phones though. I normally don’t use flash in my favorite kind of photography, so I wouldn’t know.
EDIT: meaning that there are advantages to flash matching the temperature of common light sources.
I can see it useful for making shadows less obvious, but I am not sure if 2 COBs on a phone can make much difference outdoors. If it would be helpful, it would have to be indoor lighting that uses 5000K, which might be common in retail, but not in households. I am sure that there are a few situations where it can be helpful. I would guess that you would need a more powerful flash. At that point, most people are not using their phone camera. But I could see a market for a USB flash to use with a smartphone for times where you do not want to carry a dedicated camera around, but a compact, limited-use purpose-built flash would fit.
These were taken with just the torch light with the new white balance preset from the merge request above:
This reminds me - our photos are skewed towards warmer colors because our color calibration matrix is done really badly (by none other but myself).
Does anyone of you folks maybe have access to a color calibration chart and could take a photo of it (with a L5 of course) illuminated by a D65 source? 
(there are other reasons too, like using XYZ scaling instead of something like Bradford for chromatic adaptation, but first things first)
I can ask the colleague who works in the lighting lab.
But I haven’t seen them use a light source yet, so far they have always measured the color space and brightness of LEDs when I was there.
(and my phone must work properly again at the moment I have a problem with battery and charge).
I took some calibration photos but didn’t get around to successfully creating a matrix yet.
I can create a matrix, I just need a well-exposed and in-focus photo that’s illuminated by measurable D65.
Best I can offer is semi-controlled cloudless summer sunlight.
I try to get something there, the colleague did not know if they have the source, but he asks.
Thanks! One tip - when taking the picture, make it so the tester is in the middle of the photo and takes only around 50% of the area (this way we don’t get color distortion from lens shading affecting the fields on the tester)
