Can reflected light be used to create an accurate spectrometer?

[Bob3141592]

I have need of a spectrometer to assist me in my artwork, so I plan to build one. Some of the paints I use are opaque, and many are transparent. I want to capture the spectrum as they would look painted over white canvas, or painted over other colors in several layers, so I will be reflecting light from the sample into the spectrometer, not looking directly at the sample through a light source.

I have a Raspberry Pi with a touchscreen, the camera for the Pi, and a diffraction grating at 1,000 lines per inch. There are instructions for construction of the spectrometer on the web, so I'm not asking about that. It's the illumination that concerns me. Ideally natural daylight is the best source, but artificial lighting may be necessary. Can I just put white LEDs onto a white paper, and normalize the intensity to mimic daylight in software? Then when I illuminate the sample, it's just a multiplication for the return light at that frequency to get what would be the reflectance in daylight.

Can I get a better approximation of daylight by mixing white LEDs with halogen bulbs, or would that be unnecessary since I have to normalize the source anyway?

It's the illumination aspect of the project that I wonder about. Is the project even possible using reflected light? Will it be bright enough after passing through a slit? I'm going to experiment and build it anyway, but the advice of anyone with experience would be much appreciated.

 

[rumborak]

Strangely enough, I just earlier today bought a diffraction grating for the exact same purpose. My plan is slightly different, I want to 3D-print a small adapter you can put on a smartphone camera.

 

[sophiecentaur]

Can I just put white LEDs onto a white paper, and normalize the intensity to mimic daylight in software?

LEDs vary and the spectrum is fairly continuous but is is not 'flat'. Have you looked at google images of LED spectra? They could well be good enough for the subjective effect - I suggest that what you are actually making is a spectroscope and not a spectrometer if you are not actually measuring the spectral power. I think you will need a pretty bright source for this because, once the white light has been spread out over a large area, it will be be pretty faint. (Newton - and others used bright sunlight and did his observing in a darkened room.) Did you consider a Halogen lamp source?
You may find it a very interesting exercise to mimic spectral colours with pigments. A split screen, with the projected spectrum separated with a 'wall' from a white sheet will be necessary, I think. You do not want any spillage from one half onto the other.

I bought a £30 cheapy spectroscope a while ago (a 13mm cylinder, 45mm long). It is very handy for 'fun' investigations. I was amazed at how good the LED lighting was, compared with CFLs which really suck. You might consider buying one (eBay and shop around for Gemstone analysis tools).

 

[anorlunda]

Paint stores in the USA now have machines that very accurately measure color. i don't think they are very sophisticated. Perhaps the OP can research how they work.

Edit: perhaps the OP can take samples of the subtle differences he is trying to detect to a friendly paint store and test how accurately the machines measure them.

 

[Andy Resnick]

It's the illumination aspect of the project that I wonder about. Is the project even possible using reflected light? Will it be bright enough after passing through a slit? I'm going to experiment and build it anyway, but the advice of anyone with experience would be much appreciated.

This problem sounds a lot like the problem of 'white balancing' in digital cameras. As you (should) know, there is a standard illumination condition for artwork: a northern-facing window.

That's not particularly quantitative.

So, to provide a quantitative way to distinguish between the near-infinite types of lighting (halogen, fluorescent, LED, arc,...) there is a metric called the color rendering index (CRI): https://en.wikipedia.org/wiki/Color_rendering_index. AFAIK, the 'standard' lighting condition corresponds to a CRI of 100 at a color temperature of 3200 K (https://en.wikipedia.org/wiki/Color_temperature#Correlated_color_temperature).

Artificial lighting has different values of the CRI and CCT, people who set up museum lighting for a living pay attention to this detail to properly light paintings.

Good luck!

 

[Khashishi]

You will get different results depending one what light source you use. But if you normalize the source then it's probably fine as long as you use any fairly continuous source. The halogen bulb is probably a good bet for this purpose. I wouldn't use LEDs.

 

[Bob3141592]

Thanks to all for your comments. I'm going to build this thing and if it works at all see how well it works. Time to find out how many halogen bulbs it takes to ignite a swatch of canvas!

In using opaque paints, one color is laid next to another, and each color has it's own identity. In glazing, a very thin layer of one transparent color is laid on top of another, modifying it's color. Layering deeply is possible, but I usually only go four or five layers over the bare canvas. My best painting using this technique is 'The Emergence of Music' and can be seen here: http://www.cairone.com/GalleryPages/MusicalThemes.html. Glazing gives depth and intensity, kind of like candy color car paints. The effects can be subtle and dramatic, and I want to understand what is really going on.

Yes, I want quantitative results. Perhaps it's overkill, but that's what the Raspberry Pi is for. Plus, I want to undertake the project to learn about the Pi. The idea of making test samples to take to a paint store is a good one. I'm interested in how big I can get the difference between two greens of the same RGB reading. And I didn't know about the color rendering index. Good info!

 

[Tom.G]

The most important requirement to be able to normalize the source, is that the source's spectrum be continuous. Many LED lamps do a good job of this, at least the ones that go in a floor or desk lamp as a replacement for incandescents. Fluorescent lamps have three to seven discrete emission lines with little or no output between them. (You can't normalize zero. ) This leads to the metameric effect or metamerism, where a color looks completely different depending on what light source is used.

Since some (many?) colorants (dyes, inks, etc) have some rather fine-grained spectral details, your light source should be a decent match to a blackbody spectrum. If the source has a hole in its spectrum and the colorant has a reflectance peak at the same wavelength, it ain't going to look right.

http://www.envisionlabs.com/shop/viewitem.php?productid=0
Has a handheld spectroscope for around US$11. I picked on up as a toy a while back. It works. Any camera you use with it will have to focus down to 8 to 12 Inches.

Just two very minor concerns are:

1) The plastic housing needed a little tape to close the small light-leaking gaps
2) Not really a problem, but... The optics put the image around 10 Inches from the viewport. At my age, my eyes only go down to 30 Inches, so I use reading glasses with it.

But it works!

there is a standard illumination condition for artwork: a northern-facing window.

That "northern-facing window" is nominally 6500K (or D65), the spectrum of a blackbody radiator at 6500°K.

 

[Andy Resnick]

That "northern-facing window" is nominally 6500K (or D65), the spectrum of a blackbody radiator at 6500°K.

I honestly don't know too many specifics: does it also refer to overcast skies (as opposed to direct sunlight)?

 

[Bob3141592]

A north facing window never admits direct sunlight (in this northern hemisphere), but looks out over clear blue skies (or cloudy). It's light is consistent throughout much of the day and is fabled for its purity.

My windows face east, and don't see direct sunlight much past 10 am.

 

[sophiecentaur]

Time to find out how many halogen bulbs it takes to ignite a swatch of canvas!

You could always investigate using an IR Cut filter (as used in cine projectors with fragile plastic film) they can handle kW of heat.

 

[sophiecentaur]

I'm interested in how big I can get the difference between two greens of the same RGB reading

If you are trying to fit quantitative results with RGB values then you would also need to know about the particular analysis curves that are used with your sensor. It's a bit of a nightmare, actually. I wouldn't know where to start, to produce results that would be squeaky clean. What sort of meter / sensor are you planning to use for this?
Colour matching of pigments is a million times worse than the 'simple' problem of matching a TV picture to a source colour under a given illuminant. I'd bet you will find a whole range of pigment mixes that produce a fair mutual match to the eye but which should produce different RGB values.
Someone has already wished you "good luck" but it could be interesting.

 

[Tom.G]

https://www.google.com/search?q=spectrometer+site:ti.com
Has many hits for spectrometers.

Even a training page!
https://training.ti.com/developing-your-own-spectrometer

https://www.google.com/search?q=spectrometer+usb++(fob+|+dongle)
Shows a few possibilities but probably expensive.

 

[ohwilleke]

FWIW, there is a market out there for used scientific equipment that is in good working order but was discarded by educational institutions and science labs when the upgraded their equipment that often finds buyers among industrial companies that don't need state of the art scientific precision. I know of a industrial lab that purchased its equipment that way. It is much, much cheaper than the new stuff. 1%-10% of the original prices in most cases. But, that may still be too expensive for this application.

 

[rumborak]

I had some time to 3D print a prototype of my low-end spectrometer:

https://imgur.com/gallery/IHWOe

It's not working that well yet, I already have a V2. It's mostly fun, I can already tell the quality of output is nowhere near of what would be actually useful.