Spectra of some light sources for photography - kinda long!
In particular, for indoor photography - portraits, products, art works.
Our digital cameras are amazing machines that can take high-fidelity images of all sorts of things. The color versions - basically what you get unless you’ve acquired a special digital monochrome camera - generally rely on a set of red, green, blue (RGB) filters overlaid on top of the individual pixel sensors in the camera focal plane, usually arranged in the so-called Bayer arrangement. You can readily find lots of information on how that all works, with a bit of googling. For some measurements on the color response for a few cameras, I suggest this work by Christian Mauer: Measurement of the spectral response of digital cameras | Semantic Scholar . Of particular interest for my write-up here, is that virtually all the cameras Mr. Mauer measured have a spectral response limited to 400 to 700 nanometers (nm; one nanometer is a billionth of a meter, and here refers to the wavelength of light). Hence, I have limited my results here to the same band - 400-700 nm.
Just to remind you of colors and wavelengths, here’s the 400-700 nm range
I have measured the spectrum of a number of light sources, using my Ocean Optics USB2000 spectrometer, acquired via eBay. My unit has a Sony ILX511 2048 pixel linear array CCD sensor, and is sensitive from about 180 nm to 880 nm, resulting in spectral resolution of about 0.35 nm. The wavelength values are calibrated by a fit of pixel number to the peak wavelengths of several spectral lines in a neon bulb spectrum.
The amplitude calibration (spectral power versus wavelength) is a bit more complicated. There are internal reflections within the system that cause ripples in the amplitude response, and the Sony sensor has variable sensitivity as a function of wavelength. It’s easy enough to correct for the CCD sensitivity - Sony kindly provides a curve of quantum efficiency. I have removed the ripples by assuming that an incandescent light bulb should produce a nearly black-body emission spectrum. (All hot objects produce emission (“light”) whose amplitude is a well-known function of temperature and wavelength.) Taking the ratio between a measured light bulb curve and an assumed black body curve, after correcting for the CCD sensitivity, produces a multiplicative correction factor for our wavelengths of interest. Fortunately, as noted below, I had a known spectrum from a commercial full-spectrum LED tube light to which I could compare my results.
All of my measured spectra shown below have been thus “calibrated.” My relative amplitudes are probably accurate to 5-10%. I’m reluctant to claim any better, but this is adequate to compare light sources and suggest how well they will light up our photography. The absoulte amplitude scales are completely arbitrary.
Our cameras usually image the reflected light from the objects in the field of view. (Not true, of course, if we are taking pictures of light sources (light bulb or candles, say) or astronomical images.) If we want to have good color fidelity in our images, the light source(s) we use should provide at least some light over the entire 400-700 nm range, so that objects of any color will have some light to reflect to the camera.
OK - let’s look at some light sources - some better than others!
The original light source - the Sun in the sky!
First, here’s a spectrum of the Sun itself, measure through a window at about 2 PM in the afternoon in late February here in Flagstaff and shining (through a neutral filter!) directly into the USB2000. The solar elevation is about 40 degrees (higher or lower elevation (more or less atmosphere) would affect the overall spectrum - the sun is MUCH redder at sunset!). The spectrum is remarkably flat, and clearly provides nice light across the 400-700 nm range of interest. (The various narrow dips in the solar spectrum come from various atoms near the surface of the sun, and are of much interest to astronomers!)
In these and subsequent spectra, the original uncorrected spectrum is shown by the fainter bottom line, while the corrected spectrum is the darker line. The absolute levels are arbitrary, and depend on how much light is entering the spectrometer and how long I integrate the spectrum.
However, if we are out and about in daylight, most of the light falling on us is coming from the entire sky, which on a nice sunny day (as was the case here, with just a few white puffy clouds) tends to be blue (i.e. “blue sky”). And, indeed it is:
Or, sometimes we have “cloudy bright,” as Kodak used to call it, when we get light scattered from high, thin clouds. The sky looks much whiter, and indeed it is: much less blue:
When we set the White Balance to “Daylight” or “Cloudy” in our cameras, we are telling the camera that it needs to adjust the relative levels of the amplitudes from the RGB pixels to balance out this light. So less gain for the blue pixels and relatively more for the red pixels for blue sky.
What happens when we go inside?
In the olden days, we had incandescent light bulbs and maybe a flash gun.
Here’s the spectrum of an incandescent photoflood lamp. The spectral shape looks just like we would expect for a blackbody at a temperature of around 3000 K. There is light at all wavelengths, but dominant in the red. In this case, when you set White Balance to incandescent, the camera knows it has to tone down its red response substantially, and pump up the blue values.
Incandescent lights run hot and use a lot of energy. In fact, more than 90% of the power that such bulbs use just heats up the filament. They are also rather fragile, and some don’t last all that long (the really bright flood lights designed for photo use had lifetimes as short as a few hours).
If you don’t want all that heat, you can go to fluorescent and/or LED bulbs. We’ll get to those very shortly.
Still in the “olden days” but also right up to now, you could use a flash gun to light up your subject. Here’s the spectrum of my Pentax AF-540FGZ
I also have spectra of Pentax AF200Sa and Vivitar 2600D flashes. They are very similar (the Vivitar extends a bit into the ultraviolet below 400 nm). These spectra have a nice broadband appearance, not too different from the blue sky (lacking some blue). Some cameras use a daylight White Balance for flashes, and that is probably not too far off.
Now, let’s bring our inside lights more up to date. Fluorescent tube lights have been around for quite a while, while Compact Fluorescent Lights (CFLs) went into wide spread use in the early 2000's. These lights run cooler than incandescents (although they, too, can get pretty warm) and are much more efficient in terms of light out for watts in. However, they have a pretty crappy spectrum for photographic lighting. Most of the light comes out at a few wavelengths. If you are photographing objects with certain colors (450-525 nm (blue-green) or 625-700 nm (red)), they may show up very dark in your image - there is no light there to be reflected to your camera.
Last, but far from least, we can use Light-emitting-diode (LED) bulbs. These are now the reigning light source. They are even more efficient than CFLs and generally have a better spectrum than CFLs, although some are better than others - read on!
LEDs are available just about anywhere (our grocery store carries them), and in my opinion should be the choice for household lighting. Our annual household KWH usage dropped from around 7000 KWH/yr to about 4000 KWH/yr after we converted all our lighting to LEDs. That 3000 KWH, at 10-15 cents per KWH, saves hundreds of dollars per year, more than making up for the somewhat higher cost of the LEDs.
Most lighting products these days, at least in the US, are labeled with their light output (lumens) and some sort of indication of their color temperature, perhaps an actual claimed kelvin value, or perhaps including a “warm” (sometimes “soft white”) or “cool” (sometimes “daylight”) classification, as well as perhaps a color rendering index (CRI), indicating how close its light spectrum matches that of daylight. Most run-of-the-mill bulbs have CRIs of 80 to 90.
An actual LED emitter, the little diode itself, puts out a relatively monochromatic (single color) spectrum - blue, red, green, and just about anything in between that you might want, depending on the elemental composition of the diode. To cover the optical 400-700 nm range, a bulb needs either a bunch of various LEDs inside, or more generally, a nominally blue LEDs has various phosphorescent materials added that emit at longer wavelengths.
So, here’s the spectrum of a “warm” LED flood light bulb - the kind you might use in a recessed ceiling fixture. The claimed equivalent kelvin rating for this bulb on its packaging is 2700 K. That’s a bit cooler than I assumed for my incandescent calibration bulb. This light does indeed provide a fair amount of red light, and except for the blue bump around 450 nm and the fall of in the red to longer wavelengths (above 600 nm), this spectrum is not to-o-o different from that of the incandescent bulb. This bulb would be nice in a living room, producing a nice warm overall tone, and would not be too bad for photography, either. For best photography results, you would want to set a custom white balance, although an incandescent WB might not be too bad - some red objects might not show up quite right.
Here is the spectrum of a “Daylight” (5000 K) LED light bulb. This bulb has three fold-out “wings” and is on the ceiling of my garage. The spectrum has much less red light than the “warm” bulb.
And here is the spectrum of a “cool” LED flood light bulb, with some quite unexpected spectral line features! Minus those lines, the spectrum is quite similar to that of the preceding bulb. I am guessing that some of the phosphors in the Philips bulbs emit spectral lines, as well as broad band fluorescence.
These lines seem to be a “feature” of at least some Philips flood lights. Here is a spectrum of a Philips bulb which I found on line in some of their literature, with a claimed color temperature of 4000 K. As close as I can tell, the lines are exactly the same.
These LED spectra are certainly better for photography than CFLs, but they are not quite as good as one might hope. After a fair amount of searching on line for high CRI bulbs, I found the Waveform lighting web site (Next Generation LED Lighting | Waveform Lighting), which has interesting presentations on all sorts of topics having to do with bulbs and color rendering. Unlike most vendors, they also have spectra for many of their bulbs.
I wound up buying 4 LED tubes lights from Waveform to use as lighting for a photography project at the local university art museum. These lights have a claimed CRI of 95 and were about $30 each. I made sure I got the “flicker-free” version of these bulbs. Here is the spectrum of one of these bulbs:
The spectrum is much more uniform. It still has the blue peak, but is considerably flatter throughout the rest of the spectrum than any of the LEDs shown previously, or the many dozen other LEDs I have measured. If you are serious about the lighting for your photography, I would suggest you check out Waveform’s site and products. (I have no relation to this company other than buying their product.)
And for comparison, here is the spectrum of the bulb as presented in the Waveform literature. I was rather amazed, after all my calibration shenanigans, at the agreement between their professional results and mine.
Finally, you might be contemplating trying to copy some old color prints, or scan some old color slides, to get them “digitized.” For best results, you should somehow get some daylight into your system. Next best, use some LEDs, perhaps the Waveform units for best results.
But wait, you say - I have a flat bed scanner. Can I use that? NO! Well, at least not the Epson or Canon scanners that I have. Here is the spectrum of the light source in my Epson Perfection V39 scanner (with a few bad pixels). It looks to me that it basically has three broad(ish) band blue, green, and red LEDs providing the light (many of them, of course, to cover the width of a scan line). My truly vintage Canon N1240U scanner has essentially the same spectrum. Judging from its spectrum, my Canon 8400F has a fluorescent light source, with a crappy spectrum like that shown above. None of these are going to give you high fidelity results. (I would be glad to send either of the Canon scanners to somebody for the price of shipping.)
Bottom line - daylight in its various forms is going to give you the best lighting for photography with good color fidelity. Most LED bulbs should be more or less OK, and spending a bit more for CRI 95+ bulbs will give you the best indoor performance.
I would be happy to measure any light sources sent to me.
|