1. Block Course 4 Martin Jesson
Lamp Performance and Colour Quality
Abstract
By understanding what defines quality in a light source, we can accurately compare
performance. This report outlines the modes for identifying quality and compares samples of
incandescent, compact fluorescent and LED light sources and critics there spectroradiometer
results.

2. Introduction
The lighting industry is exposed to vast amounts of luminarie and lamp types. As a designer
you must be able to critically analyse each to understand specific performance criteria to
determine a suitable light source for any given application. The properties of the light source
must be taken into account to understand differences in colour and how this affects quality.
Spectral Distribution and Colour Rendering.
Spectral Power Distribution Curves provide a visual profile of the colour characteristics of a
light source. Spectral Power Distribution is defined as W/m
−2
/nm
−1
the power per unit area
per unit wavelength of an illumination (radiant exitance). That is; the radiant power emitted by
the source per wavelength, usually referred to the visual band expressed in nanometers, 380
– 780nm.
The Spectral Power Distribution is important to consider as this will dictate the ability of the
light source to render colours of objects and surfaces. For a colour to be seen by the eye that
colour must be present in visual spectrum. See the below example of the spectral
distribution of an incandescent GLS lamp.
Note how the visual spectrum is complete from 380-780nm, this gives the incandescent lamp
a good ability to render colour accurately. This source is described as full spectrum and is
closely related to daylight (the perfect source) in its ability to show true colour.
In contrast, see the below example of a low pressure sodium lamp whose spectrum is very
narrow in the yellow region of the bandwidth.
This source has no ability to render any colour other than certain yellows accurately.

3. All light sources have different abilities to render colour and so the Colour Rendering Index
(CRI) was constructed to define these abilities. The CRI test’s how well a light source can
reproduce 14 test colours.
The CRI is calculated by measuring the difference between the lamp in question and a
reference light source like a black body radiator. The average differences measured are
subtracted from 100 to get the CRI. This means small differences will result in a higher score
and larger differences will result in a lower score. No differences = 100.
The most important test colour is the R9. This is because some form of red is present in most
colours, so to be able to accurately render a colour it is vital red is in its spectrum. Its so
important that its generally accepted that a light source must have a >0 R9 value in order to
have any colour rendering accuracy at all. On most lamp data sheets the R9 value will be
displayed as evidence of conformity.
Chromaticity of Light Source
The chromaticity of the light source is important because if you have more than one lamp in
close proximity the difference in colour will be noticeable and reflect on the surroundings.
Uniformity will not present which could create an undesirable lighting solution.
Chromaticity is described as a correlated colour temperature (CCT) of the light source. This
was formed, as not all light sources are full spectrum and therefore cannot have their colour
temperatures located on the black body locus. CCT does not define a specific chromaticity
however, rather a range of colours. See the breakdown of CCT values in the below diagram.
Colours 1 – 8 are Saturated
Colours 9 – 14 are Un-
Saturated
Unsaturated are the most
vivid of colours

4. These correlated colour temperatures are defined as the black quadrangles seen in the
diagram. Any colour within a quadrangle can be assigned that correlated colour temperature.
Note that although a lamp may be classed as 3500K, it’s colour will vary the further it plots
from the black body locus. You could have one colour above the black body locus and one
on the black body locus, both being classed as 3500K, but both vary different in chromaticity.
In a lighting installation this degree of change in chromaticity would be perceived by the
viewer if both sources within close proximity to one another. To gauge this potential of
change and quantify it, we use the Duv scale.
The Duv number will be either a positive or a negative which will place it above or below the
black body locus - a zero duv will place the CCT on it. The greater the number (whether
positive of negative), the further from the black body locus it will be and the larger the
potential colour shift. A positive number (above BBL) resulting in greenish-bluish tint to the
colour and a negative number resulting in a pinkish-purple tint to the colour.
The Threshold at which a colour difference can be perceived by a person is defined by
MacAdams Ellipses. Colorkinetics.com describes MacAdams ellipses;
“A MacAdam ellipse is drawn over the color space in such a way that the color at its
center point deviates by a certain amount from colors at any point along its edge. The scale of
a MacAdam ellipse is determined by the standard deviation of color matching (SDCM). A
color difference of 1 SDCM “step” is not visible; 2 to 4 steps is barely visible; and 5 or more
steps is readily noticeable.”
Definition of the degree of
difference in colour as outlined in
the ANSI standard and related to
CCT and Duv.

5. This report will show the relationship between lamp performance and colour quality by way of
analysis of three light sources, incandescent, compact fluorescent and LED, with the use of a
spectroradiometer.
In the diagram we can see the
MacAdam Ellipes. Note the
Centre points. The SDCM value
is the deviation from this point.
MacAdam experimented on a
range of colour temperatures.
Note: the Ellipses are drawn 10
times there actual size for
clarity.
This diagram shows the
quadrangles with the 7 step
MacAdam ellipses transposed.
They almost correlate directly
over the standard CCT we use
and with this we can used
SDCM values to define
whether a colour will have a
noticeable difference from the
black body colour standard.

6. Method
Apparatus
Spectroradiometer and associated software.
Sample lamps, 100W Incandescent, 20W CFL, 11W LED
25°C controlled ambient room temperature.
Procedure
The Sphere must be temperature stabilised before any testing can be completed. The sphere
must be at 25°C – the control temperature for photometric testing.
The test lamps must also be temperature stabilised and run for a nominal 20mins before
being tested. This is particularly important for the CFL and LED lamps. The CFL has a warm
up time until optimal lumen output is achieved and the LED lumen output is directly related to
its temperature, so to get comprehensive results it must be operated as if in domestic use.
Once the sphere has a stable temperature, its voltage is calibrated to that of the lamp rating.
The incandescent being run at 240V and the CFL and LED being run at 230V.
One lamp is placed in the sphere at a time. Its performance is monitored by viewing its lumen
output and once stable the analysis program is started.
The spectroradiometer will scan and record the data in a report which is printed and analysed.
Results
Comparison of Lamp Performance
GLS CFL LED
Luminous Flux Lm 1499 1327.5 673.41
Luminous Efficacy Lm/W 15.26 72.85 59.562
CCT K 2743 2844 3825
Ra 99.9 79.6 97
R9 100 0.6 94
Colour Shift duv +0.00009duv +0.00317duv -0.0013duv
Colour Difference SDCM 1.2 5.6 5.3
Data Sheets
Appendix A: Incandescent: Phillips 100W
Appendix B: Compact Fluroescent: Phillips 20W
Appendix C: LED: Cree 11W Module

7. Discussion
The sample lamps were run for about 20mins before they were put into the sphere. This was
to warm the lamps up and stabilise their output. Heat is directly proportional to light output, so
as a lamp is warming up its output characteristics will change. This is not so important for the
incandescent because the incandescent lamp reaches its peak heat in a relatively short
period of time (a period of seconds). The CFL on the other hand, will slowly ramp up to full
output over a period of minutes as the gas heats up and excitation fully occurs. Its important
to note that the CFL has (and is designed to have) a peak operating temperature of 25°C.
The LED also needs minutes rather than seconds to warm up, and its lumen output is very
sensitive to temperature. To get a honest report from the spectroradiometer, the LED must
be warmed up so that a true performance report can be taken.
The photometric reference standard for testing lamps is 25°C and this was maintained in the
sphere by use of a heat exchanger. This was feed via air conditioning units located outside of
the laboratory. The laboratory itself was also in a controlled temperature environment being
maintained at 25°C.
The Efficacy of the lamps was as expected. The CFL had 72.85lm/W, which is very good for a
t2 lamp. Important to note was the CFL has its best lumen output at 25°C which gives it an
advantage on achieving high efficacy over the other sources. The LED was second most
efficacious with 59.56lm/W, which was as expected due to the ambient temperature. During
the experiment the LED was noticeably hot. It was only rated at 11W, so it would be
interesting to know how much of that 11W was actually being transferred into visual spectra.
Also of interest in relation the efficacy, would be to compare a 20W LED module to the 20W
CLF. The added heat caused by the potentially 90% more heat being generated may take
the efficacy down even further. The LED only put out 673.4lm, half of the CFL output
(1327.5lm).
Overall the colour rendering properties of the lamps were good. The unexpected performer
was the LED which had an Ra of 97 and an R9 of 94. This is completely in contrast to
common notion of LED performance. On this information this LED would be suitable for
applications usually reserved for incandescent or halogen. The CFL on the other hand is
under achieving. It does not have a compliant R9 (-6) and its Ra doesn’t reach 80 (although
quite close at 79.6) This CFL is not really suited to applications needing good colour
rendition. Its most likely going to be used in a domestic situation for energy saving reasons,
so with this in mind it performs well as its efficacy is well above incandescent. The
Incandescent performs as expected with perfect R9 and Ra, 100 and 99.9 respectively. The
incandescent, whilst still having great colour rendition properties has been threatened by the
LED in this regard.
To replicate and produce vivid colours, high Ra and R9 values are required. R9 is
particularly important as most colours, but especially vivid colours contain large amounts of
red. Incandescent would render vivid colours most effectively, followed closely by the LED,
this being due to the high Ra properties (99.9 and 97). Additionally both sources have high R9
values (100 and 94). The CFL lacks impact on this front with Ra only 79.6 and R9 -6. This R9
value really id quite poor especially considering this lamps CCT is 2844K. It’s likely to be used
in a domestic situation and its ability to convey comfort and relaxed atmosphere will be
restricted due to these rendering properties. Its spectral distribution shows a large spike at
approx 530nm which is in the green spectra, not ideal for domestic environments.
If we were to rank the sources by SDCM values, then the order from top to bottom would
change once again. The incandescent leads as expected, its SDCM being very close to the
black body with a value of 1.2. If installing this GSL lamp you would not notice any colour
change/shift between the lamps, you could confidently place many together without the threat
on non-uniformity. The LED and CFL on the other hand had SDCM values of 5.3 and 5.6.
Anything above 5 is considered high risk and shouldn’t be place in groups where uniformity is
paramount. By looking at the Duv values we can assume that the colour shift is going be
toward the green for the CFL and pink for the LED. This SDCM outcome for the LED is very
unusual as this source had high colour rendering statistics. The ability to control chromaticity

8. of the LED is far more reliable and what appeared to be a high performing module, isn’t really
one that you would to install, as it likely you’ll never have uniformity in your installation. Yes
its rendering properties are good, but is probable that its appearance is going to let it down.
Refferences
Illuminating Engineering Society of North America (1993) Lighting Handbook, 8th
Edition,
Warren G Julian (2011, February). Lighting: Basic Concepts, 6th
Edition,
Colour Consistency for Colour and White LED’s. (2010). Retrieved from
http://www.colorkinetics.com/support/whitepapers/Technology_Overview_Optibin.pdf

9. Appendix A

10. Appendix B

11. Appendix C