Design of a Multiple-Lamp
Large-Scale Solar Simulator
S. P. Kenny 1
Solar Energy Applications Laboratory.
Colorado State University,
Fort Collins, CO
J. H. Davidson
Mechanical Engineering Department,
University of Minnesota,
Minneapolis, MN 55455
A simple solution to the conflicting constraints of providing uniformity and collimation of irradiance in multiple-lamp solar simulators is proposed. As proof of
concept, irradiance measurements obtained in a simulator comprised of 28 I-kW
mercury-iodide gas discharge lamps and capable of irradiating a 1.22 m-by-2.44 m
collector plane are given. The design is based on preventing a portion of the light
from each bulb from reaching the collector plane. Light blockage is achieved by
placing a "shadow board" 1.02 m from and parallel to the plane of the lamps.
Lamps are arranged in an hexagonal pattern with 4 columns of 7 lamps at a lampto-lamp spacing and column-to-column spacing of 0.45 m. Lamp-to-collector plane
distance is 3.05 m. The design is determined from measurements of the spatial
distribution of radiant energy from a single lamp. Irradiancefrom an array of lamps
is then simulated. Measurements of irradiance in the full-scale simulator confirm
that uniformity and collimation conform to the American Society of Heating, Refrigerating and Air Conditioning Engineers' standard. Average irradiance is 1080
W/ m 2• Maximum irradiance is 1190 W! m 2 and minimum irradiance is 980 W!m 2 .
Every point on the plane of the collector receives JOO percent of radiant energy from
an area on the lamp array contained within a subtended angle of 20 deg.
Introduction
Using lamps to simulate the sun is not a new concept. A
crude solar simulator was used nearly 50 years ago to test solar
air collectors (Lof, 1992). Worldwide, more than 30 large-scale
simulators were in use in 1983 (Tanemura, 1983). Unfortunately, the four original large-scale simulators in the United
States have been decomissioned. Although it is possible to
construct a simulator capable of irradiating 4 m2 using a single
argon arc lamp, cost of the lamp is $250,000 (McClenahan,
1993). The objective of this work is to provide a lower cost
alternative using multiple light sources. The problem of using
multiple lamps is the difficulty of simultaneously satisfying
specifications of uniformity of irradiance and light collimation
as stated by ASHRAE (American Society of Heating, Refrigerating and Air Conditioning Engineers) in standard 93 (1986).
The simulator is intended to test flat-plate collectors and water
heating systems such as integral storage and thermosiphon
systems.
The ASHRAE standard sets minimum average irradiance at
790 W/ m2 • Maximum and minimum values of irradiance must
be within ten percent of the average. Average irradiance is
based on values measured in a uniform rectangular grid at
intervals :50.15 m. Spectral distribution of light over 0.3 JLm
:5 t- :5 2.6 µm must "reasonably" match that of natural sunlight
'Current address: Kenny Associates Inc ., 5840C McHines Place, Raleigh, NC
27604.
Contributed by the Solar Energy Division of THE AMERICAN SOCIETY OF
MECHANJCAL ENGINEERS for publication in the JOURNAL OF SOLAR ENERGY El'GINEERJNG. Manuscript received by the ASME Solar Energy Division , Dec. 6,
1993; final revision, June 30, 1994. Associate Technical Editor: J . J . O'Gallagher.
200 I Vol. 116, NOVEMBER 1994
defined by air mass 1.5 (American Society for Testing and
Measurements, ASTM, 1987). Average radiant energy should
not vary by more than ± 3 percent. Collimation must be sufficient to ensure that ~ 90 percent of the energy received at
any point on the collector test plane emanates from a region
of the solar simulator contained within a subtended angle of
20 deg. Infrared irradiance (3.5 µm :5 t- :5 50 JLID) should not
exceed that of a theoretical blackbody at ambient temperature
by more than 50 W/ m2 •
The objective in designing a low-cost large-scale simulator
is to minimize the number of lamps and place them in an area
not substantially larger than the test plane. Prior multiplelamp designs have used either hundreds of low-power tungstenhalogen lamps in combination with fresnel lens or higher powered gas discharge lamps (typically xenon or mercury-iodide)
placed near the perimeter of the irradiated area. Placing the
lamps away from the center portion of the irradiated plane
significantly reduces any collimation provided by the lamp
reflector but can facilitate obtaining uniform radiant energy.
The xenon discharge lamp has good spectral qualities but requires water cooling and complex controls (Krusi and Schmid,
1983). The most reasonable lamp choice is the mercury-iodide
gas discharge lamp (called Compact Source Iodide, CSI, by
the manufacturer). The CSI lamp is a sealed beam unit with
the bulb placed in front of a parabolic reflector. In the bulb,
iodide is combined with mercury gas to produce light which
has a spectral output very similar to AM-1.5. The iodide reduces evaporation of mercury and thus extends bulb life and
produces greater useful output (Keitz, 1971). The lamps were
originally developed for use in stadiums to improve color television filming (Krusi and Schmid, 1979). Lamps are easy to
Transactions of the ASM E
1200
........
N
E
'......
3:
w
u
z
<
<
a::
Ci
2:
1000
Collector
800
600
400
200
0
-1000
-500
0
500
Shadow Board
1000
x (mm)
1. h
a)
1200
,-,
........ 1000
N
\
I
E
'3:
......
w
u
z
<
<
a::
Ci
2:
800
\
y
z
\
600
400
\.,
200
0
-1000
x'
z'
\
-500
0
\
'•.
1000
500
1.lb
x (mm)
b)
Fig. 1 Centerline irradiance distributions from (a) one 1·kW CSI lamp
(b) two CSI lamps spaced 0.45 m apart. The dashed line is total output
from two lamps.
control, have a 1000 hour life, and require no special cooling.
They cost approximately $1000, including ballast, ignitor,
socket, lamp and variac. To keep total cost below $30,000, it
was desired to use less than 30 lamps.
Since the built-in reflector on the CSI lamp does not produce
uniform, collimated light, this paper presents a simulator design based on use of a "shadow board" which blocks a portion
of the light emitted from each lamp from the plane of the
collector. Figure l(a) shows a typical horizontal centerline irradiance distribution of a single lamp located so that light
reaches the collector test plane at normal incidence. The solid
lines in Fig. l(b) show irradiance distributions of two lamps
placed 0.45m apart. The dashed line represents total output
of both lamps. Hot spots are produced as the "tail" of the
distribution from lamp 1 adds to the irradiance peak of lamp
2, and vice versa. In a full-scale simulator, uniformity is even
harder to achieve since the point on the collector directly underneath a lamp is affected by as many as six other lamps.
Increasing the spacing between lamps decreases maximum irradiance, but creates cold spots.
It was hypothesized that by reducing the area on the test
plane affected by each lamp, uniformity could be improved.
Figure 2 is a sketch of the geometric effect of the shadow board
Fig. 2
Schematic of shadow board concept
concept. The detrimental effects of multiple lamps are minimized by blocking some light from each bulb so that at the
point of maximum flux from one lamp there is minimal added
irradiance from surrounding lamps. To restrict the irradiated
area to a circle of radius Y, measured from the centerline of
the lamp, an opaque plane with a hole of radius x' is located
distance z' from the lamp. Lamp-to-test plane distance (z),
can be varied depending on the desired irradiance at the collector. Dimensions x ' and z' are arbitrary as long as the ratio
of x' l z ' equals Y! z.
Methodology
Light uniformity and collimation depend on configuration
of the lamp array and shadow board. To ascertain the effects
of lamp-to-lamp and lamp-to-shadow board spacings, irradiance flux maps of a single lamp were measured and then
radiant energy output of 28 lamps was simulated. Experimental
parameters include: distance from the lamp to collector test
plane, lamp orientation, lamp power setting, and shadow board
geometry. Lamp-to-lamp spacing and geometric pattern were
considered in numerical simulations of the entire irradiated
plane.
Measurement Procedure. Global radiant energy was measured with a Licor 200SA calibrated against an Eppley PSP.
Nomenclature--------------------------------G
a
p
x
irradiance, W / m2
average irradiance, W /m 2
power setting of lamp, W
column-to-column spacing of
lamp array, m
Journal of Solar Energy Engineering
x,y
x'
y
coordinates along test board,
m
radius of shadow board hole,
m
lamp-to-lamp spacing in each
column of array, m
Y'
z
z'
offset between alternate columns in lamp array, m
lamp-to-test plane spacing, m
lamp-to-shadow board spacing, m
wavelength, m
NOVEMBER 1994, Vol. 116 / 201
f'x~
T
y
0
...LQ
0
0
0
0
0
0
0
0
0
0
0
0
2000
0
0
0
0
0
0
0
J_
of
0
0
0
0
0
0
e
::1.
AM 1.5
!n.,
1 ~":.
N
E
~
w
z
<
c
<
a:
!!::
CSI 1kWLAMP
i-;,.
1500
I
:
~
:
1000
f
0
0
....
:i .
I t '!·.~
i
::
f
0.0
..
fU'"~
t~
·~
Ii:
i
500
::
ti
0 .5
1.0
WAVELENGTH
1.5
2.0
(µ m)
2.5
•..
Fig. 3 Variables X, Y, and Y' are dimensions varied in the simulation
of 28 lamps
Fig. 4 lrradiance as a function of wavelength for the CSI lamp and air
mass 1.5
To obtain a stable mean value, the signal from the pyranometer
was sampled with a 12 bit A/ D board at 200Hz for five seconds.
Response time of the silicon photodiode is 10 ms. Measurement
accuracy given by Eppley is ±(1.5 percent of reading ±6 W I
m2). Assuming a calibration accuracy of ±3.5 percent, total
uncertainty is == ±4 percent, including the error of conversion
between the two instruments. Resolution of the data acquisition system is 0. 73W / m2 • An Eppley Precision Infrared Radiometer was used to measure infrared radiation (IR). The
measurement procedure developed by Albrecht and Cox (1976),
which includes a correction for heating of the silicon dome,
was used.
though the spectral output curve of the CSI lamp falls below
that of AM-1.5 , neither the xenon nor tungsten lamp completely satisfies the ASHRAE requirement for solar spectrum
(Krusi and Schmid, 1983). According to the manufacturer,
spectral characteristics of the CSI lamps change below 70 percent of rated power.
Simulation. The objective of the numerical simulations was
to investigate many possible lamp configurations to determine
the lamp array pattern which best meets requirements for uniformity and collimation. The output of 28 lamps was predicted
from the output of a single lamp with a spreadsheet. The
spreadsheet simply adds the radiant energy from the lamps
based on the measured spatial distribution of intensity of radiation from a single lamp. Data were obtained at 0.056 m
intervals over a 1.06 m-by-1.06 m area. Average irradiance of
the simulated output of 28 lamps over a 1.22 m by 2.44 m test
area was calculated and areas of irradiance greater than ± 10
percent of the average were used to specify the degree of uniformity. Figure 3 shows the geometric parameters of the array.
Approximately 100 different combinations of X, Y, and Y'
were tried in each simulation. The values of X and Y ranged
from 0.34 m to 0.56 m and the value of Y' ranged from 0 m
to 0.28 m.
Results
Lamp-to-test plane distances from 2.75 m to 3.35 m were
considered. The minimum distance was chosen to keep the
peak irradiance at approximately 1000 W / m2 • The maximum
distance is limited by ceiling height. Lamp orientation refers
to the position of the lamp face and was varied by rotating
the lamp ± 15 deg from the manufacturer's suggested position.
Power setting was varied from 60 percent to 110 percent of
the lamp's nominal level of 1 kW . Different shadow board
hole diameters were tested in an effort to improve uniformity
of irradiance.
Spectral Output. Irradiance as a function of wavelength
for the CSI lamp and for AM-1.5 is plotted in Fig. 4. Data
for the CSI lamp were obtained from the manufacturer. Al202 /Vol. 116, NOVEMBER 1994
Irradiance. To determine lamp performance as a function
of power setting, the pyranometer was placed 3.35 m from the
lamp and power level of the lamp was decreased from 1.1 kW
to 0.7 kW (the lamps should be turned on near full power).
Below 0. 7 kW , gases within the bulb condense and spectral
output is altered . Above 1.1 kW there is risk of lamp failure.
Irradiance versus power level is
G= -0.74+2.33p-0.81p 2 [kW! m 2] ,
(1)
where p is in kW. All subsequent irradiance measurements
were obtained at 1 kW with the collector test plane located in
the vertical plane.
Another variable which affects the output of the lamp is
bulb alignment. To ensure proper lamp performance, the tail
of the bulb must be pointing directly toward the lowest point
on the rim of the reflector. The maximum output of the lamp
is reduced by up to 14 percent when the lamp is rotated as
little as 15 deg.
Figure 5 is a plot of normalized irradiance measured at the
geometric center of the lamp and at z = 3.35 m. Irradiance is
normalized by average irradiance over the five-hour testing
period. Stability of the lamp is within ± 5 percent. This deviation from the standard ± 3 percent should have little effect
on average quantities, but could affect instantaneous measurements.
Three lamps were randomly chosen to assess lamp-to-lamp
variations. The maximum variation at the point of maximum
flux is five percent. In the simulations, all lamps are assumed
identical.
Figures 6 and 7 show a vertical asymmetry in the output of
the lamps. The most likely cause of this asymmetry is interference by the bulb tail with light reflected from the built-in
parabolic reflector. (The asymmetry is in the vertical direction
since the lamps are tested with a vertical collector test plane.
Gravity has no effect.)
To determine if the CSI lamps can be placed so that the
ASHRAE specifications of light uniformity and collimation
are satisfied, nearly 100 lamp-to-lamp distances and geometric
patterns were simulated for lamp-to-collector plane test distances of 2.89 m, 3.05 m, and 3.35 m. Neither a rectangular
Transactions of the ASME
Table 1 Shadow board configurations tried for lamp-to-lamp
spacing of 0.45 m and lamp-to-collector plane distance of 3.05
18'
w
1.05
N
:::i
<
:!:
Lamp-to-Board Distance
0.15
0.14
0.13
0.11
0.10
0.09
0.08
0.06
0.05
0.95
0.86
0.77
0.69
0.60
0.52
0.43
0.34
x ' (m)
0
z
<
i5
<
a:
5
Q
w
m
Hole Radius
1.00
0.95
a:
0
z
z'(m)
1.02
1200
0.90
9-WXNI BOARD
+...............................~~...-.--~........--~............,....,......................+
60
0
120
TIME
240
180
300
N'
(min)
Fig. 5 Normalized lrradiance over time for 1 lamp. Power level= 1 kW.
z=3.35 m.
e
~
i5
<
a:
400
5
...
-+.+-+4-f-~1-4-~-l-l~~~~l--I
800
600
z
<
g
1000
w
0
HOLE RADIUS (m)
-
0.15
-
0.13
0.10
-
0.08
-
0.05
200
0
· 1000
-500
0
x
500
1000
(mm)
Fig. 8 Horizontal centerline lrradlance distributions for different shadow
board configurations
-0.S
0.56
x (m)
Fig. 6 lrradiance map of 1 CSI lamp at z= 3.05 m. Tick marks= 0.056
m. Contours=100 W/m 2•
--
1200 +-......................................................................._..................................._--+-
N'
1000
.E
~
800
w
600
0
y-axis
z
<
i5
<
a:
5
400
200
DISTANCE FROM CENTERLINE
(mm)
Fig. 7 Vertical (y) and horizontal (JC) centerline profiles of radiant energy
at z=3.05 m.
nor an hexagonal pattern creates adequate uniformity of irradiance. The best configurations are obtained with a lampto-collector spacing of 3.05 m. The best rectangular grid has
four columns of seven lamps with lamp-to-lamp and columnto-column spacing of 0.45 m. Average irradiance is 1190 W I
Journal of Solar Energy Engineering
m2 • Maximum irradiance is 17 percent above the average, and
minimum irradiance is 14 percent below the average. Thirteen
percent of the area receives more than 110 percent of the
average irradiance, and three percent of the area receives less
than 90 percent of the average irradiance.
The best hexagonal grid also has four columns of seven
lamps. The 2nd and 4th columns are offset 0.225 to form an
hexagonal pattern. Column spacing and lamp-to-lamp spacing
remain at 0.45 m. This configuration yields a predicted average
irradiance of 1200 W / m2 • Maximum irradiance is 16 percent
above the average, and minimum irradiance is 16 percent below
the average. Ten percent of the area received more than 110
percent of the average irradiance, and four percent of the area
receives less than 90 percent of the average irradiance.
Shadow Board. Since the specification for light uniformity
and collimation cannot be met by using only CSI lamps, the
shadow board is used with the hexagonal lamp configuration
producing the best level of uniformity with lamps only. The
hexagonal pattern was selected in preference to the rectangular
grid because modification of the final design for off-normal
testing is easier. Evaluation of alternate shadow board designs
was first conducted with a single lamp. Table I lists the shadow
board configurations evaluated.
Figure 8 shows irradiance distributions along the x-axis for
selected configurations listed in Table l. Other configurations
are omitted for clarity. The lamp face has a radius of 0.10 m.
If the hole radius is :5 the lamp radius, the peak irradiance is
lower than if the hole radius is > the lamp radius. A shadow
board with hole radius of x ' = 0. 15 m is picked for the final
design to ensure that the peak irradiance is not reduced.
Figure 9 shows an irradiance distribution along the x-axis
for a lamp with and without the shadow board. Lamp-toNOVEMBER 1994, Vol. 116 / 203
Fig. 9 Centerline irradiance distribution for a lamp with and witho
the shadow board
Fig. 11
Completed solar simulator
nine percent lower than the average irradiance. The irradiance
pattern of this configuration , shown in Fig. 10, exceeds the
ASHRAE requirements for light collimation.
0
1.10
1.22
x(m)
I.OS
1.00
0.95
0.90
Fig. 10 Simulated irradiance of 28 lamps in an hexagonal array with
the shadow board
collector plane distance (z) equals 3.05 m and the shadow board
has dimensions x' =0.15 m and z' = 1.02 m. At x= ±0.45 m,
the differences in irradiance between the bare lamp and the
lamp-shadow board combination are on the order of 25 W I
m2• With this reduction in irradiance, the maximum flux directly under any lamp can be reduced by nearly 150 W/ m2 ,
based on all lamps being affected by six neighboring lamps.
Decreasing a hot spot by 100 W / m2 , or ten percent of an
average 1000 W / m2 , is enough to solve the non uniformity
problem.
The shadow board must be designed for specific lamp-tolamp spacing and test distance. Therefore, exhaustive numerical simulations are not required to optimize the irradiance
output. However, simulated irradiance of 28 lamps is still based
on output of one lamp. The shadow board , with hole radius
x' =0.15 m, is placed 1.02 m from the lamps. The numerical
simulation for the 1.22 m by 2.44 m test area predicts an average
irradiance of 1050 W/ m2 with a maximum of irradiance six
percent greater than the average value and a minimum value
204 /Vol. 116, NOVEMBER 1994
Final Design
The completed simulator facility consists of a space frame,
a light mounting frame and shadow board, and a collector test
stand with all necessary plumbing and controls required for
testing. Complete details of the design are given by Kenny
(1993). A photograph of the complete simulator is shown in
Fig. 11. The simulator irradiates a 1.22 m-by-2.44 m test area
with an average irradiance of 1080 W / m2 • The lamp-to-collector plane distance is 3.05 m, and the shadow board is located
1.02 m from the 28 1-kW CSI lamps. The shadow board hole
radius is 0.15 m. The array is composed of four columns of
seven lamps each spaced 0.45 m apart. The distance between
columns is also 0.45 m. The 1st and 3rd columns are offset
0.225 m below the 2nd and 4th columns. The shadow board
is constructed of plywood. To prevent overheating of the board,
the wood is covered with reflective aluminum on the side facing
the lamps. The other side of the board is painted flat gray.
Radiant energy measured in the simulator is plotted in Fig.
12. Average irradiance over the test plane is 1080 W/ m2 • The
maximum and minimum values differ from the average by
+ 10 percent and - 9 percent, respectively . The nonuniformity
of the distribution in Fig. 12 could be reduced by finer adjustment of lamp power levels. However , since the variation
in radiant energy is not much greater than the time variations,
adjustments were not made. Improvements in spatial distribution can also be made by more accurate aiming of lamps.
The lamps should be aimed so that the geometric center of the
bulb is located along a line perpendicular to the point of maximum flux from that lamp on the collector test plane. DeviaTransactions of the ASME
The simulator has been used for extensive testing of an
integral storage collector (Mason, 1993). Future challenges are
modification of the shadow board and lamp arrangement to
allow off-normal testing.
Ackowledgments
The financial support of Mr. Robert Hassett of the U.S.
Department of Energy, the Office of Energy Conservation of
the State of Colorado and Colorado State University are appreciated.
References
1.10
1.22
x(m)
0
1.05
1.00
Albrecht, B., and Cox, S. K., 1976, Pyrgeometer Data Reduction and Cali·
bration Procedures, Atmospheric Science Paper No. 25 l, Department of At·
mospheric Science, Colorado State University, Fort Collins, CO.
American Society for Testing and Materials 1987, Standard Tables for Terrestrial Solar Spectral Irradiance at Air Mass 1.5 for a Jr Tilted Surface,
Designation E 892-87, Philadelphia, PA.
American Society of Heating, Refrigerating and Air-Conditioning Engineers
(ASHRAE), 1986, Methods of Testing to Determine the Thermal Performance
of Solar Collectors, Standard 93-1986, Atlanta, GA .
Keitz, H . A. E., 1971, Light Calculations and Measurements, 2nd ed.,
MacMillan, London.
Kenny, S. P., 1993. "Design of an Indoor Solar Simulator," Masters Thesis,
Department of Mechanical Engineering, Colorado State University, Fort Collins,
0.95
0.90
Fig. 12 lrradlance In the completed simulator, normalized by average
lrradlance of 1080 Wlm2
tions as little as 2 cm can affect uniformity. Average irradiance
can be reduced by lowering the power level of each lamp.
Infrared radiation measurements were taken with 28 lamps
operational. Ambient temperature was measured in the plane
of the collector with a thermocouple in an aspirated radiation
shield. The difference between the measured IR and that of a
theoretical blackbody at ambient temperature does not exceed
50 W /m2 over four hours .
Journal of Solar Energy Engineering
co.
Krusi, P., and Schmid, R., 1979, " Design ofan Inexpensive Solar Simulator,"
Proceedings of the 1979 International Solar Energy Society Conference, Atlanta,
pp. 417-421.
Krusi, P ., and Schmid, R., 1983. "The CS! 1000 W Lamp as a Source for
Solar Radiation Simulation," Solar Energy, Vol. 30, No. 5, pp. 455-462.
Lof, G. 0 . G., 1992, Colorado State University, personal communication.
Mason, A. A., 1993, "Stratification and Performance Prediction in an Evacuated-Tube, Integral Storage Solar Collector," Masters Thesis, Department of
Mechanical Engineering, Colorado State University, Fort Collins, CO.
McClenaban, D., 1993, Canada National Solar Test Facility, personal communication.
Tanemura, S. , 1983, Solar Simulators for Collector Testing Specifications
and Operating Characteristics, International Energy Agency, Task III, Subtask
C, Solar Research Laboratory, Government Industrial Research Institute, Nagoya, Japan .
NOVEMBER 1994, Vol. 116 / 205