General > General Technical Chat
Looking for RGB or RGBW values of a sunrise
<< < (6/7) > >>
Twoflower:
Sorry if my simple description from the first post was misleading. A as good as I can manage with the RGBW sources is good enough for me.

@cdev: No worries about looking direct into the LEDs. It's behind my bed and cowered with a opaque diffusor.

Wrt different sunrises at different locations. That's actually one additional plus for the sensor: It's easy movable and I can ask someone who travels at a nice place to do the recording.

Yes I could use my camera in HDR or do some additional calculation to compensate changes of the exposure time (I would keep the diaphragm fixed). But I feel that this is rather cumbersome setup; especially as my tripod is a heavy load (Manfrotto) and I don't own a car.

@rs20: About the usage of the white channel: Thanks for the heads up. It would be very obvious to get completely overblown rendering of the record. That needs to be taken care of. I noticed that extreme wide range of the W channel too. I think ill still record it and keep the data but not use it.

I think a mismatch of the Filters and the LED colour might be bad. Just imagine the red LED is shifted compared to the sensor that only 20% passes the filter the RGBW sensor: The reading will be much to low and you would think you need to beef up the R channel more than necessary.

I see the sensor like a small project. For both the Cam and the Sensor I need the hardware setup (opaque dome or reflector, ...) the calculation from the sensor/cam needs to be done. Here the sensor gains a bit as I have already the RAW data in a nice format. And I have several controllers flying around which are currently waiting to be used.

@BrianHG: Probably I'l use one sensor which is used for both usage models. As I don't plan to compensate changes on the room I do a single calibration run to identify correction curves that are applied to the RAW sensor data on a PC. the processor might have enough CPU power to do this online. But imagine something red falls on the sensor and you will wake up under a green-blue sky :-DD

The pinhole lens sound like a good idea! Worth a try.

@reen flash: (OK, that's not a user rather a topic) I've watched so many sunsets over the sea and never was lucky to see one. I was aware of this effect and hoped so many times.

To my understanding a place with a view over the sea will do. And the weather conditions must be perfect: The sky must be clear and the right temperature conditions over several air layers must help to bend the light a bit around the earth surface.

I'm not sure if a plane is helpful. So far I only saw pictures from it on sea level.

Just for your entertainment I've added two pictures of the driver (6+1 12bit PWM channel, the MOSFETs are bottom side mounted, controller not mounted) and the LED stripes loosely placed in the profile with the opaque cover.

Edit: Added Pictures
BrianHG:

--- Quote from: Twoflower on September 21, 2017, 05:51:23 pm ---@BrianHG: Probably I'l use one sensor which is used for both usage models. As I don't plan to compensate changes on the room I do a single calibration run to identify correction curves that are applied to the RAW sensor data on a PC. the processor might have enough CPU power to do this online. But imagine something red falls on the sensor and you will wake up under a green-blue sky :-DD

The pinhole lens sound like a good idea! Worth a try.

--- End quote ---

As for the pinhole lens, you can safely point this instrument directly at the sun and it wont fry the sensor + the distance between the pinhole and the RGBW sensor controls the field of view the sensor sees of the sky.  Like a zoom control, but, no focusing needed.  IE, the further away from the RGBW sensor, the more narrow section of the sky falls on the sensor.  The closer to the sensor, more of the sky's view-able area will be centered on the sensor.

rs20:

--- Quote from: Twoflower on September 21, 2017, 05:51:23 pm ---I think a mismatch of the Filters and the LED colour might be bad. Just imagine the red LED is shifted compared to the sensor that only 20% passes the filter the RGBW sensor: The reading will be much to low and you would think you need to beef up the R channel more than necessary.

--- End quote ---

The already-stated assumption is that the Red filter in the RGB sensor has the same response as the corresponding cones in the eyes. So, if the red LED is "shifted" such that only 20% passes the red filter, then that red LED only produces a 20% (proportionally speaking) response in the red cone, and it's exactly the right thing to do to beef up the R channel accordingly.

Meanwhile, matching the filter to the emission spectrum of the LED leaves your overall system completely blind to the fact that this LED only weakly stimulates your eye's red cones, producing far too little red light accordingly. Matching the filter to the LED instead of matching the filter to the eye is exactly the wrong thing to do.

The only issue that can be caused by a poor choice of LEDs is if the combination doesn't have a wide enough gamut. But since sunrises are not particularly saturated in colour (all the wavelengths are there in some quantity), I seriously doubt you'll run into gamut issues. (You'll know you've run into a gamut issue if your interpolation software turns into extrapolation software and requests a negative drive for one of your LEDs.)
Twoflower:
@BrianHG: Right that distance might work as zoom. But I think as the sensor is very small this will end up in a huge 'crop factor'. Probably useful distances are in the mm range. But worth a try.

@rs20: Yes you're probably right as the sensor is designed to match the human eye. I missed that fact and just thought the total luminosity. So if the the LED has a strange centre wavelength the eye will be affected similar to the sensor. This should cancel it out. I checked the LED vs. Sensor wavelength. They're not that far away that this actually matter for my needs (620 vs 650nm, 520 vs 550nm and 465 vs 450nm).
rocketscienceguy:
I'm also looking for an RGB sequence of a sunrise, or, preferably, a daylight cycle.  It does not need to be as precise as this original thread author intended, just an approximation for a cycle I can use as a translucent image overlay to broadly simulate day/night cycle in a game.  I've read through all the postings on this thread, and it looks helpful to someone who wants to produce their own accurate and precise measurements for a particular time and place.  No generic sequences were actually given, and I'm hoping that with my relaxed requirements someone would share such a sequence knowing it would be helpful. 

Minus a chart, I'll start experimenting with sequences.  My first thought is to start from black, and then plot a path mostly along the edge of the RGB color cube: black(255,255,255)->magenta(255, 0, 255)->red(255, 0, 0)->yellow(255,255,0)-[diagonal from yellow to cyan, bypassing green]->cyan(0,255,255).  I'd step through all the states between each of these vertices in a straight line, and, upon reaching the endpoint, I'd reverse the sequence.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod