Looking for RGB or RGBW values of a sunrise

[Twoflower]

For a small project I need a table of RGB or RGBW values over time of the sky-color from the morning just before the sun rises until the sun is nicely over the horizon**. In the end I would like to play back this using a LED strip. Actually two strips: One RGB and one white. Thus the RGBW. But I think I can convert an RGB table to RGBW.

As I don't have a single pixel RGB sensor I need to find another way. My idea so far was to setup a cam looking into a white, opaque globe to get the sky colour as whole. Do a picture every minute or so. And reduce the rescale the picture to 1x1 pixel to further obscure the light and get the RGB value from the sky from that remaining pixel.

Any other ideas, or is there somewhere an existing table I can use? An existing table would be nice as here the winter fall is coming and the chances to capture this is not that high any more.

**I'm speaking of a rising sun as seen from Earth. And a nice one is preferred ;-)

[Twoflower]

Thanks evb149.

As I said, I'm looking for a nice one. So a location with low particle count (including water). OK, the longitude and altitude wasn't specified. So the altitude should be about sea level (I like the reds of the rising sun). About the latitude: Here I'm probably not that picky, lets say somewhere in the range of +/-50° would be acceptable. To my understanding that mainly changes the speed of the phases and I can adjust that. The sun position should be between 8° below horizon (blue hour) and 15° above horizon (no visible reddish sky any more). Looking at all the data, couldn't be a simulation a source of this table?

Using the white balance and lux value is some kind of different scale which should be possible to transform into the requested RGBW table. But one or two values are not enough to calculate a nice transition between the points. And so far I haven't found any table that provides more than 2 or 3 values unknown how they timely relate. That's the reason why I'm looking for a higher resolution on the time axis. Probably 20 entries are enough to extrapolate the steps in between. I wouldn't complain if I can get more steps.

About the dynamic range. That's also right, I'm limited to 12bit resolution per channel. And that's plenty for my application. Probably except that the lower end might be still to bright for my taste.

[Twoflower]

Thanks again evb149.

I scanned through the papers. And that provided the right keyword in one of them: "CIE daylight model". Even if the model needs to be purchased it can be used as Google search term. Lots of interesting information pops up. Including a Daylight Visualizer from a company producing windows (free download). Probably I need to play with it to see if I can abuse it and model a integrating sphere instead a room with windows.

[ajb]

As I don't have a single pixel RGB sensor I need to find another way. My idea so far was to setup a cam looking into a white, opaque globe to get the sky colour as whole. Do a picture every minute or so. And reduce the rescale the picture to 1x1 pixel to further obscure the light and get the RGB value from the sky from that remaining pixel.

I'm not sure integrating the whole sky will get you the effect you want, since the quality of light at sunrise is highly variant depending on viewing direction.  You may get more useful information from an integrating screen aimed towards the sunrise, or perhaps multiple captures with the screen aimed at different parts of the sky.  You'll also need to be careful that the white balance of the camera is fixed through the whole capture, otherwise the changing white balance will confound your normalization process. 

Another thing to consider is that if you are looking to illuminate a room or something with your RGBW LED strips you're going to have a very difficult time replicating the way a sunrise illuminates a scene.  This is down to how the illumination spectrum interacts with the reflectance spectrum of the objects under illumination.  An object illuminated by an RGBW "orange" is going to look quite different than when illuminated by a sunrise "orange" depending on its pigmentation.  The only real way to do better here is to use more wavelengths--IE, instead of red, green, and blue, you might want deep red, red, orange, yellow, green, and blue.  Or even more--some LED stage lighting fixtures use seven different wavelengths to achieve smooth color mixing and less erratic interactions with the pigments in the scenes they illuminate. 

So if you wanted to really closely replicate the illumination of a sunrise, you'd want to take a spectrograph of the sky over the course of a sunrise and then try to match that spectrum as best you could with a handful of different wavelengths (and possibly multiple color temperatures of white as well).  It still wont' be perfect, but it will be far better than RGBW.  Of course if you're just illuminating a white screen or something, then RGBW may be just fine.

[Twoflower]

ajb: Yes you're absolutely right. The automatic setting would spoil the process. On the other side one could probably use the while balance result the Cam would use from the EXIF data. But for sure that will be the first thing I will do: Fixing the exposure settings.

And I also plan to calibrate the final setup with the same approach. Because the environment is not neutral white, but it is static. The idea is that set some specific RGBW values, taking a picture and compare what I wanted to see and what I actually get. As the equipment is very compute limited that factor will then included into the lookup table in a offline process. Unfortunately I can't fit more than 2 LED stripes into the aluminium profile. So one RGB and one White has to be sufficient. then again I don't want to have a 100% replica of a 2*10^30 kg fusion reactor shining from 150mio km through 500km gaseous molecules in my home. It would be nice to come close without getting a sunburn and the heat. But so far I need some starting point with the possibility to improve the result in a iterative process.

Unfortunately I don't have a spectrograph, just a Canon EOS 5DMk3. That's the reason for using this one to collecting the data. Plus the converting from the continuous result from the spectrograph to discrete RGBW values. But that is probably doable.

[mrpackethead]

You'll not be able to reproduce it with RGBW..  You'll need to widen the pallet out, and include Amber, Cyan and Violet LEDS as well to do it accruately.

[Twoflower]

As said: I have only RGBW available. So I have to live with the area I can cover with that. Again: I don't need a full spectrum rendering. Whatever I can do with the possibilities that I can build.

In theory the driver circuit has two spare channels. But mechanically I can't add additional colours.

[cdev]

I like this idea..

However I think the use of a diffuser is going to sell you short because youll only get the average color also finding an color neutral dome sounds well nigh impossible (if there was such a thing it would either be insanely expensive, I suspect, or likely be a gel, not a dome)

The rising sun actually seems to me to at some points in time display almost all the colors of the rainbow which could be scanned in a vertical line.

But its very subtle and likely only a quite good sensor would capture it. Also, those beautiful subtle shades are right at the beginning, long before the suns disk appears.

The sun's light would fry your camera's sensor, so don't let that happen

Quote from: Twoflower on Yesterday at 13:53:12

For a small project I need a table of RGB or RGBW values over time of the sky-color from the morning just before the sun rises until the sun is nicely over the horizon**. In the end I would like to play back this using a LED strip. Actually two strips: One RGB and one white. Thus the RGBW. But I think I can convert an RGB table to RGBW.

As I don't have a single pixel RGB sensor I need to find another way. My idea so far was to setup a cam looking into a white, opaque globe to get the sky colour as whole. Do a picture every minute or so. And reduce the rescale the picture to 1x1 pixel to further obscure the light and get the RGB value from the sky from that remaining pixel.

Any other ideas, or is there somewhere an existing table I can use? An existing table would be nice as here the

winter fall is coming and the chances to capture this is not that high any more.

**I'm speaking of a rising sun as seen from Earth. And a nice one is preferred ;-)

[BrianHG]

You want one of these:
https://www.digikey.ca/product-detail/en/vishay-semiconductor-opto-division/VEML6040A3OG/VEML6040A3OGCT-ND/5168309
Also, you need to approximately point it upward or better in the direction away from you viewing angle to replicate the view from that direction.  Higher quality huge outdoor LED screens use them to correct their brightness & color throughout the day.  High quality cameras use them to figure out what the auto white balance should be...

[cdev]

Ages ago I used to use a program called Chromatica which deconstructed an images palette in a really quite useful(for a web developer or illustrator) way.

You'll lose the potential to do something much more effective if you attempt to merge all the beautiful individual colors in a sunrise or sunset into one.

[Twoflower]

Ok, now we get somewhere! I haven't expected less from the forumites 

@evb149: I've seen the Matlab-code in the GitHub repository. But I didn't spend too much time in there. But the nature article and the github description mentioned that there the RAW information are hidden. It's on my to-do list to dig further in there. That looks promising.

@BrianHG: The Vishay VEML6040 colour detector looks very nice and it's cheap too. And easy to use as the whole analogue circuit is included (I'm a digital guy; for me everything between 1 and 0 is broken ). Probably I'll order few as a colleague saw it and wand two as well. That would make the final calibration much easier. But I don't plan to do a closed loop. Just a one time calibration of the environment. The recording of  rising sun with this will be a piece of cake as well. I haven't expected that they are that cheap and easy to use. Except the soldering might be a small challenge. But I've managed to solder SMD devices first time right I'm not too much worried.

@cdev: Unfortunately I can't manage to find the Chromatica software. There's too much noise if you enter the name plus some other keywords.

@cdev: I thought about using individual LEDs to do something like the Philips Ambilight. But I decided against it as the stripes I plan to use have more power (RGB combined: 420lm/m, white: 2240 lm/m), higher density (RGB: 60/m, white: 140 LEDs/m), and the white a much better colour rendering (Ra > 80, not the best, but reasonable). So I decided to go down to a resolution of 1x1 pixel.

Again thank you so far for the input. Still if someone has some additional nice ideas, hints. I'm open for any input.

[BrianHG]

If you just wanted to replicate the white of the sun, not sky color, I have done a Planckian locus in the CIE 1960 UCS color temperature correction for my video scaler and I can give you the RGB correction factors for approx 9000K through 3000K, the sun at high noon being flat 100% RGB = 5000K and approximately being 3400K 1 hour before dusk and 1 hour after dawn.  The adjustment is close to the human eye's visual response once a reference color is converted to RGB.
IE, put an RGB color in, use the selected color temperature correction factors and get a new RGB color out.

[cdev]

You might be able to create a sunrise look with mixed dichroic light.

[cdev]

It was a Photoshop plugin for the first iteration of plugins under OS 9 mostly. The best period for plugins.

[rs20]

Ummm, why are people suggesting RGB light sensors chip? Almost everyone has a color digital camera, and the sensor of a digital camera happens to consists of millions of RGB light sensors, pre-connected to a trivial-to-use USB / SD-card interface.

Your problem reduces to recording a video of a sunrise with a decent-quality camera, and then also record a second movie of your RGBW strips being run through a sequence of different values. The second movie is your Rosetta stone; telling you which RGBW intensities result in a given recorded value in the camera. Just apply that transform (some basic interpolation required) to find the RGBW intensities required to produce the same response in the camera as the sunset, and your RGBW strip will now produce output that looks identical to the sunset (at least as far as your camera is concerned; a crap camera may let you down here). It goes without saying, as mentioned by others, that we way the light distributes through the room will not be identical, just the average hue of light (depending on how you crop/average the recorded images.)

If you just point an SMD RGBW sensor vaguely at the horizon, you have virtually no control whether it mixes in the blue sky above or the green grass below, which would mush away the orange of the sunrise.

I did something similar once to produce an image on a computer screen that looked exactly like an apple, using a camera taking a photo of the screen to correct for the transfer functions of both the monitor and camera. I had the real apple propped up in front of the screen and it worked surprisingly well -- as in, people would be walking up to my demo from afar saying "Why do you have two apples propped up in front of a blank computer screen-- OHHH WOW". (Context: then you'd look through a hyperspectral filter and they looked totally different, ergo "look how much information your eyes are missing" etc etc.) Just a consumer digital camera and MATLAB.

Just remember that sRGB is a non-linear representation; things will be seen to be behaving much more linearly if you convert back to a linear space (e.g. Nikon RAW files are in fact just 16-bit TIFFs with linear color space.)

[Twoflower]

About the 'Sun colour' and the 'sky colour'. That's a question I had also in my mind and feared to answer. To be honest I can't give an answer here. Probably a mixture from both. As my project is actually a try to get me awake in the winter time this is probably a question I should ask in a medic-forum. The only paper I'm aware of is that the blue light seems to be problematic if you want to sleep. So probably it's a good idea to have the blue hour in the show included. I'm sure there are papers out there but I haven't looked. But as the lookup table isn't set in stone OTP-ROM. I can run some tries to see how it will eventually turn out.

Using the the RGBW sensor/Camera/Simulation I can probably tweak the opening of the viewing angle to change the ratio between full sky (180°) and more sun-whigted configuration (e.g. 20°).

@BrianHG: That sounds interesting. Maybe I'll come back later to your offer. The RGBW sensor looks very tempting and adds to my current project another one. Plus an easy way to run a room calibration.

@cdev: That dichroic filters would require a full spectrum light-source. I'm not sure is a white LED will working well enough. And would eat a lot of the light as the not used light will be reflected the other direction. The plan is to flood a roof-wall (45° angle). I'm not sure if such filters exist for a line shaped light source and probably would require motors/servos to change the colour. I've seen lamps having such filters, they look very interesting.

So probably the plug in is not available. Might have been interesting of the results.

@rs20: I started my idea with the cam approach. But it has some downsides as well.

One of the problems of a cam might be the dynamic range of a prosumer/consumer cam (like the EOS D5 Mk3 that is floating around here). Starting a the blue hour running until the sun is reasonable above the horizon is probably too much. So you need to take care there too. Not impossible but something you need to take care of.

The colour sensor would be fitted with a opaque dome. I never came to the idea to use the bare sensor. This could also reduce the environmental impact. Just limit the viewing angle to the upper part of the sky.

the demo you mentioned sounds nice. I would like to have seen it. Yes the human eye-brain combo is fascinating. The dynamic range of the eye on one side and how easy the brain can be fooled on the other side.

Thanks for the heads up with the non linearity of the standard colour system. One more reason to use the RGB sensor. You can keep the whole flow relatively controlled. The only pitfall here I see is the room-calibration. In case the LED colours don't match reasonable with the filters of the sensor. That would probably need a fall back to the white sensor channel and run the calibration for each colour channel  independent. But that probably counts also if I would use a camera for this calibration run, but with no fall-back solution.

Again thank you very much for this productive discussion.

[cdev]

Blue LED light all by itself should be diffused by bouncing it off of something - Blue LEDs shining at you above a certain level thats not as high as we would wish may damage the eyes.

I found lots of stuff about this on pubmed.

Interesting, other colors may have other effects.. Red light, as I am sure you know, is postulated to be better for people before they go to sleep.. I even found some tantalizing side excursions that I bookmarked but have not had the time to explore more fully..

Green light may be good for your health  

Its my understanding that the pineal gland is where our body's circadian rhythm is regulated.

[cdev]

I thought you were undertaking some kind of artistic project or installation.

There is a lot of variation in sunrises and sunsets both seasonally and geographically.

If you can get a decent camera, whose color rendition you trust, you could maybe mount it in a fixed position and use it to take standardized photographs at fixed intervals. Maybe once each minute at exactly the minute. Store them in a database and then you could call it up, you could also create custom functions so you could say, get a animation which represented the same time before or after the sunrise and or sunset throughout the year. You should also bracket extensively so you can do HDR. You likely could keep a single camera very busy all the time.

That would tell you more about the colors and trends than anything else I could think of, perhaps.

You might want to take a look at the biological pathways involved in the circadian rhythm system.

I have two friends who have studied the process of wakefulness and sleep - extensively. Haven't talked to one in decades and the other is now deceased..

But its very interesting stuff.

Issues like SAD may be rooted in some amino acid imbalance.

Quote from: Twoflower on Today at 11:33:24

About the 'Sun colour' and the 'sky colour'. That's a question I had also in my mind and feared to answer. To be honest I can't give an answer here. Probably a mixture from both. As my project is actually a try to get me awake in the winter time this is probably a question I should ask in a medic-forum.

[rs20]

The RGBW sensor looks very tempting...

Another point; I'm not sure if people are suggesting taking the 4 channels out of an RGBW and using them to drive the RGBW strips? I.e., take R reading from sensor and drive red LEDs accordingly, take W reading from sensor and drive white LEDs accordingly... Because if that's what is being suggested, that's a terrible idea. Consider what happens if you ask this system to reproduce a pure red laser. Both the R and W channels of the RGBW sensor will respond, and therefore both the Red and White LEDs in the strip will light up. But this is obviously the wrong thing to do; the best impersonation an RGBW strip can do of a red laser is to turn the red LEDs on only. Turning the white LEDs on as well would just wash out the colour.

This is why it's preferable to use some sort of basic feedback or calibration/characterisation; and once you're doing that the W channel of the sensor is of very little value.

One of the problems of a cam might be the dynamic range of a prosumer/consumer cam (like the EOS D5 Mk3 that is floating around here). Starting a the blue hour running until the sun is reasonable above the horizon is probably too much. So you need to take care there too. Not impossible but something you need to take care of.

Fixed ISO + Aperture, but leave the camera to choose a shutter speed. Check the EXIF data to compensate. Also, whether you actually want to be matching the brightness of a real sunset is questionable; I highly doubt that your LED strips can compete with direct sunlight. So you're going to want to deliberately compress your dynamic range anyway.

Thanks for the heads up with the non linearity of the standard colour system. One more reason to use the RGB sensor. You can keep the whole flow relatively controlled.

Meh, I'm suggesting just calibrating the whole RGBW strip + camera filters + RGB transform as a single black box -- that's the beauty of this approach, you don't need to tease these things apart at all.

The only pitfall here I see is the room-calibration. In case the LED colours don't match reasonable with the filters of the sensor.

No, you explicitly don't need to worry about this! That's what the calibration is for. As long as you trust that your eyes have the same response as the digital camera; everything else is sorted.

But hey, whatever floats your boat. I guess it's plausible to claim that the response of the RGB filters of the chip recommended earlier (just ignore the W honestly, it extends way too far into the NIR) are closer to that of the human eye than the filters on a typical digital camera. And there's nothing inherently wrong with using the sensor, it's just that it feels to me like a lot more work.

[cdev]

Don't LEDs also work in reverse, so to speak, with their maximum sensitivity being at the color they emit light at?

[rs20]

Don't LEDs also work in reverse, so to speak, with their maximum sensitivity being at the color they emit light at?

No, they will respond to colors at a higher energy/frequency than what they emit at -- i.e., photons with enough energy to promote electrons across the bandgap or whatever (if the LED epoxy is tinted red, though, that will obviously contribute). In any case, even if the sensitivity curve matched the emission curve perfectly, that's explicitly not what the OP wants -- the OP wants sensitivity curves that match the human eye, as creating an experience that matches a sunset (as seen by humans) is the aim of the game here.

[BrianHG]

Use 2 RGB sensors.  1 for the sky and the other for you LED strip.  Match the color characteristics.  The RGB sensor has the built in analog with temperature correction to guarantee a fixed reference reading from device to device.  Also, a cheap pinhole lens may be used to capture the sky with a narrower field of view.

[cdev]

Please let us know if you capture the legendary green flash. (a brief moment of green light that occurs at a specific moment during the sunrise and sunset as the light from the sun is emerging or vanishing.)

[BrianHG]

Please let us know if you capture the legendary green flash. (a brief moment of green light that occurs at a specific moment during the sunrise and sunset as the light from the sun is emerging or vanishing.)

Wicked.... I've never seen that before....

[cdev]

Indeed,

Okay, now you need to see it.. how?

 you need to have a clear view towards the west or east. Very flat places are the best- or in an airplane. I think the reflection off the water magnifies it so you're most likely to see it at a beach alongside a large body of water.

Ridgetop hiking, at sunset, I think you may have a better chance although it seems to me the inversion/blob kind of green happens less from high up.

(to be honest with you I don't know the science, just trying to go from memory, sunsets from ridges while hiking=spectacular/colorful..)

If you don't live anywhere near a beach, but fly from time to time, perhaps your best chance of seeing one soon is from the air?

Quote from: BrianHG on Today at 03:29:20>Quote from: cdev on Yesterday at 22:52:12

Please let us know if you capture the legendary

green flash. (a brief moment of green light that occurs at a specific moment during the sunrise and sunset as the light from the sun is emerging or vanishing.)
Wicked.... I've never seen that before....