|National Gallery: Spectral Power Distribution (SPD) Curves by Joseph Padfield is licensed under|
This page has been set-up to provide more detailed descriptions of the various terms, values and calculations performed to prepare and present the data within this project. Where appropriate links to Wikipedia articles have been provided.
What is the Spectral Power Distribution (SPD) of a light?
Limitations of only measuring Lux
Measuring the SPD of a light
Light source descriptors
Normalising SPD based on human perception
Relative proportional power distribution
Correlated colour temperature (CCT)
Colour rendering index (CRI)
Colour quality scale (CQS)
CIE 1976 u'-v' chromaticity distance (Duv)
Relative spectral sensitivity: s(λ)dm,rel
Applying s(λ)dm,rel to normalised SPD curves
Examining how different light source effect the colour of reflected light
Graphical display of standard functions and measured data used in calculations
"In color science and radiometry, a spectral power distribution (SPD) describes the power per unit area per unit wavelength of an illumination (radiant exitance), or more generally, the per-wavelength contribution to any radiometric quantity (radiant energy, radiant flux, radiant intensity, radiance, irradiance, radiant exitance, or radiosity)." Wikipedia
Example spectral power distribution (SPD) curves, standard tungsten, filtered daylight and a LED based source.
In additon to the data presented within this website a number of lower resolution SPD curves of current light source can be seen on the GE Lighting website.
Lux levels represent a measure of the brightness of a given light as perceived by the human eye, but it does not establish the colour quality of a light, or accurately record the intensity of the light that the human eye is less sensitive to.
When comparing the intensity or power output of a single type of light source, for example tungsten halogen, Lux levels allow for a good comparison of the overall relative light exposure a work of art has been exposed to. However, when many different types of light need to be compared, a similar Lux reading could potentially be produced from a wide range of markedly different SPD curves each representing a different overall relative light exposure.
Recent examinations have shown that differences in the SPD curves of lights at a constant Lux level can effect the rate of fading of some materials, though to a lesser extent than presence or exclusion of UV. So these materials can be seen to have a different sensitivity to light than that of the human eye. Further work may be required to determine how these differences in light sensitivity may alter the long term degradation of materials under different types of light source.
However, it is very easy to measure Lux levels and many museums and galleries already own the required equipment. It would be good if it could still be used as part of a more accurate light level definition. If perhaps new SPD curve specific Lux level limits could be meaningfully recommended for light sensitive work of art.
SPD curves are measured using a Spectroradiometer, which are designed to measure the spectral density of an illuminant. They measure the relative intensity of a light source at a number of defined steps across the visible region of the electromagnetic spectrum, generally 380-780nm. Within the National Gallery, London we currently use a Konica Minolta CS-1000a Spectroradiometer, to produce 1nm SPD curves. The SPD data can be exported in the form of a simple comma separated list of wavelengths and readings, which can be graphed and manipulated other appropriated pieces of software. This current version of this device is the CS-2000.
Konica Minolta CS-1000a Spectroradiometer.
The title/names for each of the light sources presented within this project are constructed in the following way:
|"Supplier or source" "Product code or name of source": "Class of source" - "Published CCT" - "Published CRI"|
|["Calculated CCT" - "Calculated CRI"]|
|Where the light source class are as follows:|
|LED => Light-emitting diode|
|OLED => Organic light-emitting diode|
|TF => Tungsten Filement|
|TH => Tungsten Halogen|
|FL => Fluorescent lamp|
|CFL => Compact fluorescent lamp|
|MH => Metal halide|
|HPS => High pressure sodium|
|DL => Daylight|
|TS => Theoretical Source|
As noted before a SPD is the relative spectral output of a given light, measured at a number of defined steps across the visible region of the electromagnetic spectrum. These readings can be expressed in a number of different power related units, which can be graphed to produce a characteristic trace for a given light source. However the actual absolute recorded values are directly proportional to how intense a light source is or even how close it is to the the Spectroradiometer taking the measurements. Therefore in order to compare the SPD curves of different light sources they need to scaled or normalised.
One of the simplest methods is to scale each curve so that either they all have the same maximum value or that they all have the same value at a given wavelength, often at 555nm, which coincides with the peak sensitivity of the human eye.
Lux levels represent a measure of the intensity of a given light as perceived by the human eye. As it is easy to measure the Lux level of a specific light source, scaling SPD curves so they represent the same level of intensity, as perceived by the human eye, allows for a meaningful comparison.
This form of normalisation was carried out as follows:
Example SPD curves, scaled between 0-1000, multiplied by the photopic sensitivity function and normalised using the appropriate photopic sensitivity scalar.
In recent years the use of LED based light sources within museums and galleries have become more common, primarily due to their relatively high efficiency and long life. However, the presence of relatively strong "blue" peaks in their SPD curves have caused some concern among conservation scientists. It has been suggested that this relatively strong peak in the higher energy end of the visible part of the electromagnetic spectrum might increase the rate of light induced fading of sensitive works of art. In fact many of the replacements for standard tungsten halogen lamps contain a substantially higher proportion of "blue" or high energy light. These relative differences can generally be seen by comparing their normalised SPD curves, but it is often clearer to compare the relative intensities of discrete sections of the electromagnetic spectrum rather than the full curve.
In this project the visible range, 380 - 779nm, has been divided into 8 separate 50nm steps; 380 - 429nm, 430 - 479nm, etc. The sum of the power recorded in these separate section could then be displayed in the form of a percentage of the total power recorded over the entire range. To make it easier to compare these sums of power across various different light sources the percentages have actually be calculated relative to the total power recorded over the entire range for a theoretical uniform light source.
In the example below the normalised standard tungsten light source can be seen to have more than twice the total power content measured for one test LED source and the example daylight equivalent (solux) and fluorescent light sources have more than twice the high energy content of the LED source. Each section of the visible spectrum is likely to effect different materials in different ways. As more wavelength dependant fading studies are carried out, the relative importance of these or similar proportional power distributions could help facilitate the choice of appropriate light sources to illuminate particular works of art.
Example relative proportional power distribution, the graph can be seen more clearly here.
"The color temperature of a light source is the temperature of an ideal black-body radiator that radiates light of comparable hue to that light source. The temperature is conventionally stated in units of absolute temperature, kelvin [K]. Color temperature is related to Planck's law and to Wien's displacement law." Wikipedia
The equation required to calculate an approximation of the correlated colour temperature of a given light source is introduced here. The calculation carried out within this particular web site are taken from Version 9.0.1 of the NIST CQS Excel spreadsheet developed by Yoshi Ohno & Wendy Davis, NIST 20/03/2013
The Colour rendering index of a given light source is a measure of how well it can reproduce the colour of a specific set of test samples relative to an ideal light source with the same colour temperature. Light sources with higher CRI values (95-98) are general recommended for colour critical applications. It should be noted that a high CRI value does not guarantee that a light is capable of reproducing all colours "accurately", only that it will perform as well as a matching ideal light source.
In recent years the CRI value has come under increasing criticism, particularly when it is used to describe light sources with spiky SPD curves, such as fluorescent lamps or white LEDs. A new metric, the Colour quality scale (CQS) is currently under development.
The Colour quality scale is new metric being developed to provide a better description of how reliably a given light source can reproduce colour. The CQS values calculated and displayed within this system are calculated using the equation and data included within Version 9.0.1 of the NIST CQS Excel spreadsheet developed by Yoshi Ohno & Wendy Davis, NIST 20/03/2013 and quoted from that spreadsheet:
"This program is shared with you for scientific research exchange purposes only. Please do not use this program directly for your company's products or activities. When you do so, please verify the correctness of the program on your responsibility. NIST is not responsible for any loss to your company caused by any errors in this program.
... CQS is still at research level, and can be used only for your research purposes. It should not be used for your product specification. Note that this version has slightly different CCT factors and scaling factor than those in the SPIE paper. If you publish results from this program, please refer to the papers below:"
Duv has been defined as:
Closest distance from the Planckian locus on the (u', 2/3 v') diagram, with + sign for above and - sign for below the Planckian locus. (ANSI C78.377-2008)
The CIE 1976 u'- v' coordinates of ideal black body radiators follow a curve though the colour space called the Planckian locus, however the coordinates of real light sources, with matching CCTs, will be distributed above and below, positioned along a line perpendicular to the Planckian locus, depending on the quality of the light. The Duv value defines how far the coordinates of a given light are from the ideal position, defined by it's CCT.
Calculated Duv values with a magnitude higher than 0.006 are not preferred as a white light and ideally values should have a magnitude of less than 0.001. Further information can be found at here.
Taken from Version 9.0.1 of the NIST CQS Excel spreadsheet developed by Yoshi Ohno & Wendy Davis, NIST 20/03/2013.
The information in this section is mainly based on the work presented in the CIE Technical Collection 1990. The data and functions presented here only represents the average relative spectral sensitivities for museum objects, calculated within the referenced publication. Further object specific relative spectral sensitivity functions could be calculated to demonstrate the specific sensitivities of an individual object.
When it comes to the light induced degradation of museum objects all light is not equal. It is generally accepted that, on average, UV, and to a lesser extent blue light, (with it's lower wavelength and higher energy) can induce more degradation relative to red light (with it's higher wavelength and lower energy) but the exact relationship between wavelength and potential damage, an objects relative spectral sensitivity, is not as widely understood.
By measuring the wavelength dependant degradation of a number of appropriate samples it has been shown that although the exact relative spectral sensitivity for different materials does vary, in general the relationship between wavelength and potential damage can be expressed with the following exponential function:
s(λ)dm,rel = a * exp (-b * λ)
By examining a graph of this exponential function it can be seen that on average, relative to UV light at 300nm, objects are 0.38 times less sensitive to blue light at 380nm and 0.005 times less sensitive to red light at 740nms
The general s(λ)dm,rel function, this graph can be seen in more clearly here.
Applying the general s(λ)dm,rel function was carried out as follows:
Example normalised SPD curves, the s(λ)dm,rel function and the same example SPD curves after the s(λ)dm,rel function has been applied.
These graphs can be seen more clearly here, here and here
When a light is shone on a surface the colour that we perceive is a combination of the SPD curve of the light, the reflectance spectra of the surface and the colour sensitivity of the human eye.
A page has been constructed to display and compare the CIE Lab value calculated for a each of the 24 patches of a Macbeth ColorChecker under two different light sources and calculate the average CIE ΔE 2000 difference between them. Additional pairs of light sources can be compared by updating the light source selection.
It should be noted that it is not possible to perform a straight conversion from SPD to a coloured patch for some of the patches, under some of the light sources, as their colours cannot be represented within the limited screen RGB colour space without additional mathematical transformation. This represents one of the reasons why it can be difficult to accurately image and reproduce certain colours.
Please also note that at this moment this is a simple representation of converting spectral information into a standard colour space to begin to examine colour gamut issues and colour difference calculations. Further work with colour appearance models, such as CIECAM02, will be required before more accurate theoretical representations of how these colours will appears to the human eye, under different light sources, can be presented.
|Graphical display of standard functions and measured data||External Link|
|Theoretical black body curves||Black body radiators|
|CIE 1995, Munsell test colour samples||Test color samples|
|CQS test colour samples||Information Page|
|Reflectence Spectra for the Macbeth ColorChecker||Macbeth ColorChecker|
|CIE 1931 colour matching functions||The CIE standard observer|
|Characteristic vectors of illuminant D; component SPDs S0 (blue), S1 (green), S2 (red)||Illuminant series D|
|Relative spectral sensitivity||CIE 089-1991|