Hello everyone,
So I am currently working with some people of my school on a CanSat, a satellite made with an arduino board put inside a can, we then are supposed to measure a couple of values and we should have a "mission". So, our mission is to make a visible solar spectrum based on the solar radiation the satellite receives, and we thought that should be relatively easy after we searched for a couple of sensors.
So this is our idea:
Get a good number of sensors, each detecting the intensity of the radiation it receives with a particular wavelength range, this way if we used, say, 15 sensors, each measuring its range, we should be able to make a decent spectrum, but here is the problem: we can't seem to find sensors that detect particular wavelength radiations. We have found this one TSL2561, but from what I have searched it detects all visible radiations and all infrared radiations, only measuring the intensity of visible radiation and the intensity of infrared radiation but not of each wavelength inside the visible range.
Do you have any idea on how we can do this/ know any sensor that measure particular visible wavelength ranges?
Thank you! 
This app note may have information of interest.
For filters you could try: color dichroic filters
(may be expensive from Edmund's, but 'dichroic filter' is a google google search term)
There are these devices: TCS3200, TCS3210 which have Red,Green,Blue photodiodes. You can find them mounted on boards for Arduino. The spectrum overlap is a bit messy tho (see datasheet).
Yours,
TonyWilk
You could also go back to Newton and his prisma.
Put the prisma on a servo and let it split a beam of light into the "rainbow". Put a light sensor behind a split and by adjusting the angle of the prism the color sent to the lightsensor is selected. The geometry and the amount of steps of the servo determine how precise color can be selected.
In an augmented version you could use a small linear motor or maybe even piezo's to move the split in minute steps to fine tune the color selected.
Furthermore you can have different sensors behind multiple splits to have optimal sensitivity for UV, blue, green yellow, red and IR.
(Your sat should of course play an MP3 of Pink FLoyd album, Dark side of the moon)
I've been thinking on how to do that in relationship to measuring PAR for plants.
One possible solution I can think of is find a way to filter the light through a grating so you make the spectrum visible (a prism would have the same effect), and shine that on a CCD like used in cameras. Then ideally every row of the CCD would correspond to a very narrow band with no overlap.
Not sure how well an Arduino would be suited to read such a CCD image, or how to properly construct the optics of the thing.
The SparkFun AS7262 visible spectral sensor has six photodiode/filter combinations that cover the entire spectrum. There is a similar module for NIR wavelengths.
wvmarle:
One possible solution I can think of is find a way to filter the light through a grating so you make the spectrum visible (a prism would have the same effect), and shine that on a CCD like used in cameras. Then ideally every row of the CCD would correspond to a very narrow band with no overlap.
I like the idea very much, would use an RPI for image processing though, better equiped for that.
The alignment of the ccd row with one wavelength can be solved by processing only one line.
So effectively one pixel per wavelength...
robtillaart:
You could also go back to Newton and his prisma.
Nice idea there using a prism !
And, of course, bonus points for playing the mp3 
Yours,
TonyWilk
Try one of these
robtillaart:
I like the idea very much, would use an RPI for image processing though, better equiped for that.
The alignment of the ccd row with one wavelength can be solved by processing only one line.
So effectively one pixel per wavelength...
Teensy or ESP32 would likely do as well. Maybe even an ESP8266. Resolution doesn't have to be crazy high: 400 lines would give you a 1-nm resolution for the visible spectrum. Make that 1024 lines to allow for overlap and some averaging. All you're really interested in is the brightness OR the colour (the first for an actual measurement, the second for calibration), and you don't have to read the full CCD in one go anyway, 0.1 second "exposure time" is fine.
OK now it's probably obvious that I have no clue how a real CCD works and is read, but that's how my approach would be to limit the needed processing power. It'd be interesting to develop this into a commercial sensor - it makes me wonder why no-one did this yet, I haven't been able to find PAR sensors that can work in any light. It's always based on sunlight, or specific light sources such as fluorescent lights with known spectrum. Nothing for LEDs, nor have I been able to find any commercially available light spectrum analysers.
CtrlAltElite:
Try one of these
Impressive!
Hi again,
Eventually found my old stock of these: 2" Square Card diffraction gratings
Used lots of these for lectures on 'Chemistry of Pyrotechnics' - gave one to every student.
They aren't too expensive (the price has gone up quite a bit since I bought 'em in 2002 tho')
Yours,
TonyWilk