[SFF Network] RGB - An Epiphany

[ricochet]

So would you guys rather we move the RGB technical details to the RGB button thread or just leave it here?

Good idea, I vote to move it over so we don't lose this cool info that definitely pertains to @iFreilicht's project.

 

[HeroXLazer]

So would you guys rather we move the RGB technical details to the RGB button thread or just leave it here?

Move it, everyone who want to see can see there.

 

[|||]

The relevant posts have been copied into the RGB Vandal Button thread.

 

[nix]

I've never had an RGB "enhanced" product. It's news to me that these are software controlled. As a Linux user that makes it all the less likely I'll bother getting into it.

 

[HeroXLazer]

The relevant posts have been copied into the RGB Vandal Button thread.

Oh yeah, I forgot you were a moderator and could do that. I may get there some day.

 

[ricochet]

I've never had an RGB "enhanced" product. It's news to me that these are software controlled. As a Linux user that makes it all the less likely I'll bother getting into it.

I believe this is going to be open source therefore Linux-friendly ;-) as per: https://smallformfactor.net/forum/threads/rgb-vandal-button.1055/page-4#post-21698

 

[Ceros_X]

I'm not expecting professional levels of delta-error, but even just knowing the primaries of your LEDs and their voltage/current/PWM (depending on driving mode) response would be enough to get a rough map of the available colourspace. Once you have even a good estimate of the colourspace you can now translate between colourspaces which would greatly ease in matching multiple systems to the same colour.

The short (ish) version is:
The range of possible colours an RGB light can produce is determined by the 'primaries' (the colours of the R, G and B sources). You can think of that range as a 'volume', roughly cuboid, whose corners consist of: pure R, pure G, pure B, white (R G and B at max, AKA your 'white point') and black (R G and B at minimum). Put the black point at the 'bottom', the white point at the 'top', and that shape is the 'colourspace'.
If you imagine the most pure red possible, the most pure blue possible, and the most pure green possible, then you can make a space of all possible colours. This cannot be achieved in reality, because no physical object could produce these theoretically pure colours. We're mapping colour as perceived by the human eye, which doesn't directly correspond to a spectrograph of a given light source.
If you have different coloured primaries, then this cuboid will change shape and move about within that 'full' colourspace.
TV, cinema, desktop monitors, and HDR all have one or more specifically defined colourspaces they are expected to conform to. sRGB is the common one for desktop monitors though sometimes you'll find AdobeRGB, rec.709 for HDTV, DCI P3 for cinema, rec.2020 for UHD, rec.2100 for HDR, etc.
The problem is, the way colour data is stored is not 'here is the exact colour I want to represent with this pixel', but instead 'here is a mix of R, G and B for this pixel'. Notice how we just handed over three numbers without bothering to define colour primaries, or absolute brightness, or whitepoint? This is why it's so hard to get an image on any given screen to look the same as on another: a colourspace is just assumed, and if you assume the wrong one the image is not displayed as intended.
Some displays even deliberately use a 'wrong' colourspace to have punchy saturated colours, or are just unable to actually match any specific colourspace due to poor choice of backlight and/or colour filters and.or phosphors etc.
The generally accepted standard for describing the 'space of all possible colours' is 'Lab colour space. If you know how your system fits into Lab space, then you can use that as a go-between to know your it maps to every other colour space.
If you know what colourspace a colour is stored as (for PC use, assuming sRGB is almost always correct), and know what the colourspace your light source can achieve, you can map one onto another. The actual mapping can either be just a simple intersection (i.e. limited to choosing colours both spaces have in common), or you can clip out any colours that go outside your achievable colour volume, or even 'squeeze' in colours that fall near and outside the edge of the volume. The last option seems initially attractive because it means no sharp transitions if you;re trying to do a fade, but overall it means trying to actually match any given colour is nearly impossible. Think of it like trying to solve the problem of being to tall to fit your entire view in a mirror by bending the mirror and viewing a funhouse reflection: technically solves the problem of fitting everything in, but gives a useless result.

Now, a full Lab representation of an RGB lighting system would be massive overkill. For the forseeable future (any use of non-sRGB colour is still a clusterfuck on PC, let alone HDR), even a rough translation between sRGB and what your LEDs can achieve would but it well beyond the capabilities of any RGB lighting system.

As for how to measure this, my first idea would be to grab the cheapest tristimulus colorimeter* (or spectrophotometer* if one turns up) off of ebay, which will be sold as a monitor calibrator. Pair this with DisplayCal. Because we're only trying to get a relationship with sRGB, we can ignore most of Displaycal's capabilities, and just pretend the LED is a single-pixel monitor.
The way DisplayCal works is to draw a box of a certain colour on your screen. You stick the calibrator over that box, and DisplayCal reads the measurements to tell how far off your display is from the colour it should be showing. The 'hacky' way to make this work to calibrate an LED would be to stick the calibrator on the LED, and 'sniff' that box for the onscreen colour, then set the LED to that colour for the sensor to read. DisplayCal will iterate through a bunch of different colours, then spit out a LUT (Look Up Table) that tells you how to get a desired colour from the actual range the LED can show. Depending on how the LED driver is set up, you could even correct how it drives the LED and recalibrate to get a more refined LUT.

*Tristimulus colorimeter = looks for how much red, green and blue light it can see using filters. Only gives truly correct results for a single colourspace and if the display primaries match those of that colourspace. Great for calibrating an sRGB display to sRGB, won't give quite the right answers in other cases
*Spectrophotometer = compact spectrograph. measures the spectrum of a lightsource. More broadly applicable than a tristrimulus colorimeter, but less accurate.

I am a smart guy but I feel dumb after reading things like this. Great discussion!

 

[Phuncz]

That could be quite a challenge though. You can't really expect every PC to have internal bluetooth connectivity.

Nope but it can be easily added though.

+

But this isn't something I'd desire out of your project (although would be very cool) but from all the manufacturers that already have an extensive hardware ecosystem. Though I'm not sure wireless communication inside a PC is even feasible.

 

[iFreilicht]

As a Linux user that makes it all the less likely I'll bother getting into it.

Just another reason why we need an open standard for controlling these.

But this isn't something I'd desire out of your project (although would be very cool) but from all the manufacturers that already have an extensive hardware ecosystem. Though I'm not sure wireless communication inside a PC is even feasible.

Man I don't know. Something just doesn't seem right about connecting internal stuff wirelessly, but on the other hand that's the sort of thing space agencies are orienting themselves toward to save space. And it's not like RGB stuff would require a lot of bandwidth.

What could be holding manufacturers back from this, even if it was a proprietary system, is cost.

 

[jtd871]

It's even money that my new motherboard will have LED capabiity (still on the fence about which one to get), but I likely won't be using it anytime soon. Now, custom vinyl on the outside: that might be in the cards.

 

[zovc]

My reservation about internal Bluetooth is that I have yet to see it have any sensible amount of range. I have Bluetooth on my 'new' Asrock Z97E-ITX/AC and it can't even talk to a PS4 controller ~3 feet from the case. Having a recommended Bluetooth adapter in a front USB port on my desktop had a similar degree of range.

Of course, just for making literally internal components talk it probably would have enough range (and tolerance for latency), but it's just been an impressively poor experience with Bluetooth.

And yet, proprietary devices that use the format seem to work pretty darn well. Like my Logitech 'wireless' devices or my wireless headset which I'm pretty sure uses Bluetooth.

 

[CC Ricers]

I am liking the side effect of components becoming more monochrome in appearance, since color accents are now delegated to lighting. This is good for people like me that like the monochrome look. If I buy a motherboard or RAM that has programmable LEDs at least I hope there would be an option to turn them off.

Otherwise it's cool to have them around in case you want to change accent colors without the hassle of replacing parts. RGB LEDs have that added convenience.

 

[zovc]

Yeah, I've always appreciated color accents, but have never committed to or bothered with them. Being able to pick one after the fact (or even change it) is a huge aesthetic convenience.

 

[iFreilicht]

It's even money that my new motherboard will have LED capabiity (still on the fence about which one to get), but I likely won't be using it anytime soon. Now, custom vinyl on the outside: that might be in the cards.

I was talking about an included Bluetooth radio for wireless RGB control as @Phuncz propsed it.

 

[GreatestUnKnown]

This. It's a subtle addition, but it's not ricer-level lighting

Lexus=Tuner Car✔
RGB LEDz=HP or FPS Gainz✔


Verge Article with pics

Had SportsCenter playing in the BG today but they said "RGB LEDz" and I just had to share....

 

[zovc]

 

[iFreilicht]

I mean that car is pretty fucking lit.

 

[Phuncz]

I'd love that for the front of a car so it could display "you're driving in the wrong lane": https://www.reddit.com/r/belgium/comments/64jq1j/the_brussels_times_belgian_drivers_are_the/
#frustration