We present the design and characterization of a high flux solar simulator (HFSS) based on metal halide lamps and built from commercially available components. The HFSS that we present was developed to support the evaluation of a solar thermochemical reactor prototype. The HFSS consists of an array of four independent lamp/reflector modules aimed at a common target location. Each module contains one 2500 We lamp and one electroformed ellipsoidal reflector having an interfocal distance of 813 mm. The modules are oriented with an angle relative to the target surface normal vector of 24.5 deg. Design simulations predicted that the peak flux of this HFSS would be 2980 kWth/m2, with a total power delivered to a 6-cm target of 3.3 kWth, for a transfer efficiency of 33.3%. Experimental characterization of the HFSS using optical flux mapping and calorimetry showed that the peak flux at the focal plane reached 2890±170 kWth/m2, while the total power delivered was 3.5±0.21 kWth for a transfer efficiency of 35.3%. The HFSS was built at a material cost of ∼$2700.00/module and a total hardware cost of ∼$11,000.00 for the four-lamp array. A seven-lamp version of this HFSS is predicted to deliver 5.6 kWth to a 6 cm diameter target at a peak flux of 4900 kWth/m2 at a hardware cost of ∼$19,000.00 ($3400.00/kWth delivered, $1100.00/kWe). ## Introduction Solar simulators are used in the research and development of new solar energy technologies, as well as in the study of novel materials for service in harsh environments [1]. Rather than building full-scale prototypes for on-sun testing, solar simulators are used to evaluate new technologies under controlled conditions at a smaller scale and for extended duration (e.g., 24/7 operation). There are two types of solar simulators commonly used in solar energy research and development. In the first case, a low-flux simulator is used to approximate normal sun (one sun) conditions in support of the evaluation of photovoltaic cells and solar heating systems. These systems attempt to apply a uniform flux to the test article and to closely match the solar spectrum (e.g., AM 1.5). Another type of simulator is used in the development of concentrating solar power systems. These are called high-flux solar simulators (HFSS) and are the focus of this work. They typically use one or more ellipsoidal reflectors to focus light from short-arc lamps to produce a beam of light that replicates focused solar energy. The beam is typically aimed at a target located at the focus of the reflector array, and hence the highest level of light concentration. It is important for the output from an HFSS to match the total power delivered and flux distribution expected from the full-scale solar concentrator, and somewhat less important to match the solar emission spectrum. These devices serve primarily to concentrate light, sometimes to over several thousand times the normal solar irradiance of 1000 W/m2 [2]. In operation, electrical power is provided to the lamp by an electronic ballast. The primary role of the ballast is to actively manage the power flow to the lamp during ignition and during steady operation; most ballasts allow for the modulation of output power down to about 60% of full capacity. Power is routed through the ignitor, and, during ignition, this device serves to provide a high voltage (∼5 kV) pulse to the lamp to initiate the arc. Once the arc is established the voltage drops to ∼100 V. Most metal halide lamps are designed to be attached to a housing using a lamp holder. These devices are designed to provide electrical isolation from other system components up to a voltage equal to the anticipated ignition voltage. Cooling of the lamp is often needed to maintain a safe “pinch” temperature at the interface between the lamp and the lamp holder. This may be provided with a fan, or more locally with compressed air. The radiant output from each lamp is reflected by the truncated ellipsoidal mirror and focused to the target plane that is typically positioned at a distance on the order of 1 m from the center of the lamp. Only a small fraction, 20–40%, of the electric power sent to the lamp by the ballast arrives at the target as focused, radiant power. The remaining 60–80% of the energy is either absorbed by components in the system or scattered away from the target. Ekman et al. provide a detailed discussion of the energy losses within a metal halide solar simulator [3]. There are many HFSSs of various sizes and configurations in use around the world [4,5]. In general, the majority of these systems utilize Xe arc lamps as the source of radiant power. Xe lamps are advantageous in that they generally have short arc lengths (<7 mm out to a power level of 5 kWe) enabling relatively high concentration at the target plane compared to metal halide lamps having relatively longer arcs (7–25 mm out to 5 kWe). Both Ekman et al. and Roba and Siegel provide a detailed discussion on the relationship between arc size and concentration ratio [3,6]. In recent years, several solar simulators using metal halide lamps have been built to take advantage of the relatively lower cost of these lamps and the somewhat better match to the solar spectrum [7,8]. While these systems are typically smaller in size, with respect to electric power rating, than their Xe counterparts, they can be designed to deliver several kilowatts of radiant power to the target plane and achieve peak flux values in excess of 3000 kWth/m2, a level sufficient to support research on high temperature thermochemical processes. In this work, we present the design and experimental characterization of a four-lamp, 10 kWe HFSS built with commercially available parts and using metal halide lamps. This four-module HFSS transfers 3.5 kWth to a 6 cm target located at a distance of 813 mm from the arc center. The measured peak flux of this simulator is 2890 kWth/m2. The simulator is designed to be reconfigurable, including up to a total of seven modules that would deliver an estimated 5.5 kWth at a peak flux of 4900 kWth/m2. The cost of this HFSS is$3300/kWth delivered to the target. State of the art systems based on Xe arc lamps can cost up to $60,000/kWth delivered [9]. ## Simulator Design and Characterization ### Design Constraints. The HFSS design that we present was developed to support the experimental evaluation of a ceria-based thermochemical hydrogen-production reactor sited at Sandia National Laboratories (SNL) [10]. The reactor requires radiant heat input to support the partial reduction of cerium oxide in a reduced pressure (1–10 Pa) atmosphere and at a temperature greater than 1400 °C [11]. A schematic of the test assembly is shown in Fig. 1(a). All heat input from the HFSS enters the thermal receiver/reactor through a quartz dome shown in detail in Fig. 1(b). Fig. 1 Fig. 1 Close modal In operation, the metal oxide reactant is heated in the reduction chamber to a temperature of ∼1500 °C to facilitate thermal reduction. Achieving the necessary conditions with the receiver requires that the simulator produce an average flux level across the 6 cm aperture of ∼1050 kWth/m2, delivering a total thermal power of at least 3 kWth to the aperture. The total electrical input power required by the simulator array was estimated to be between 9 kWe and 12 kWe based on an expected transfer efficiency, the ratio of thermal power delivered to electrical power input, of between 25% and 33%. This thermal receiver operates under vacuum and as such utilizes a quartz dome to seal the aperture. The dome is mounted to the receiver using a high temperature gasket that must not be allowed to exceed a temperature of 200 °C. This constraint requires that all light from the simulator intercept the receiver aperture within a cone having an acceptance angle of 90 deg. This defines the allowable rim angle of the simulator array and effectively limits the number of modules in the array to a maximum of five based on the diameter of each module and best-case packing configuration. ### Single Module Design and Characterization #### Optical System Configuration. The design constraints of the thermochemical reactor dictate that our simulator array consist of no more than five lamp-mirror modules delivering a total of 3 kWth to a 6 cm diameter aperture. Our overarching design objective was to produce a HFSS from commercially available components in order to minimize cost and improve the accessibility of HFSSs in general. A survey of commercially available lamp options resulted in the selection of the Philips MSR Gold 2500W/2 FastFit metal halide lamp as our light source (Philips Lighting, (Amsterdam, The Netherlands) [12]). This lamp has a relatively short arc length for its power class and, assuming a transfer efficiency between 25% and 35%, should deliver between 625 Wth and 875 Wth for a single module, or between 3125 Wth and 4375 Wth for a five lamp array. This lamp also has a single-ended, bayonet-style connection and requires a PGJX-50 lamp holder, which is rated to provide electrical isolation up to the lamps ignition voltage of 5 kV (Bender and Wirth, Kierspe, Germany). Mirrors for HFSSs can be produced in a number of ways. Many systems in use today are based on custom-built mirrors made using a process called metal spinning in which tool is used to press a spinning metal sheet against a mandrel, forming the shape of the mirror. We opted to use an electroformed mirror in an effort to achieve a higher-quality reflective surface than possible with metal spinning. Several mirror options from Optiforms, Inc. (Temecula, CA) [13], were considered and are summarized in Fig. 2 (Optiforms). Fig. 2 Fig. 2 Close modal Each of these mirror options is composed of a structural nickel substrate (1.3 mm thick) topped with reflective silver and an aluminum oxide overcoat (50 nm combined). The metal halide spectrum-weighted reflectivity and diffuse fraction, presented by Roba and Siegel [6], is 90.0% and 0.4%, respectively. The RMS surface roughness was reported by the manufacturer to range between 20 nm and 30 nm. Of the options shown in Fig. 2, the E-813 has the least amount of truncation and will intercept more of the radiant energy from a lamp than will the other options. This was illustrated quantitatively by calculating the view factor between a lamp and the mirror surface (VFLM) as $VFLM=1−(VFL1+VFL2)$ where VFL1 and VFL2 are the view factors between the lamp and the circular surfaces corresponding to the truncation planes. The results of this calculation are given in Fig. 2 and show that the E-813 has a distinct advantage over the E-1023 mirror, by only a slight advantage relative to E-1585. These two options will intercept roughly the same amount of light, but the peak flux for the E-813 reflector will be greater owing to its smaller interfocal distance and correspondingly lesser amount of beam spread between the mirror surface and target. We note that the mirror options shown in Fig. 2 do not represent the full extent of the design space; custom electroformed optics can be produced that would enable a greater interfocal distance while still intercepting a large fraction of light from the lamp. We chose a stock reflector geometry to minimize cost and reduce lead time. The structure of the lamp-mirror module is largely built from extruded aluminum framing from 80/20, Inc (Columbia City, IN) [14]. This hardware is easy to work with and avoids time consuming and costly machining and welding operations. Two module components, the attachment between the lamp holder and the framing, and the aluminum plate at the top of the module, were custom produced from laser cut aluminum sheet. The mirror was connected to the structure using an aluminum ring provided by Optiforms that is bonded to the outer surface of the mirror with silicone rubber. A picture of a single module is shown in Fig. 3(a), and an exploded assembly view with all major components labeled is shown in Fig. 3(b). Fig. 3 Fig. 3 Close modal A cooling fan was included above the mirror to manage the temperature of the mirror and the lamp holder. A thin aluminum sheet was wrapped around the entire structure to prevent access to energized components. The module can operate in any orientation, although some lamps have limitations on the allowed burning position that can restrict the allowable simulator orientation. #### Module Characterization. The performance of the single module was evaluated through a combination of full beam calorimetry to measure radiant power delivery, and optical flux mapping with a lambertian target to characterize the flux distribution. The results of these tests are presented by Roba and Siegel [6], and summarized in Fig. 4 for clarity. Fig. 4 Fig. 4 Close modal Each module can deliver about 900 W of power to the target plane at a peak flux of 860 kW/m2. This corresponds to a transfer efficiency of 37%. It should be noted that the performance of an individual module can be very sensitive to the position of the lamp center relative to the mirror focus. The flux map shown in Fig. 4 was the result of a parametric study in which the position of the lamp was varied in 0.5 mm increments (using shims) a total of 2 mm in front of and behind the mirror focus location specified by the manufacturer. ### High Flux Solar Simulators Array Design and Characterization #### Optical Design. The single module test data indicate that the target of 3 kWth of power delivery can likely be satisfied by an array of four 2500 W lamps even when the lamps are not oriented normal to the target plane. We considered two possible array configurations shown schematically in Fig. 5 and including both a four lamp array and a five lamp array. Fig. 5 Fig. 5 Close modal Each module in the four lamp array is oriented such that the angle between its optical axis and the vector normal to the center of the reactor aperture, $θMA$, is 24.5 deg. For the five lamp array, the angle is 30 deg. The rim angle, $θRIM$, is the angle between the aperture normal and the edge of any of the mirrors. This represents roughly half of the angular width of the beam of light produced by the simulator array and must be less than 90 deg, the acceptance angle of the reactor shown in Fig. 1. The five lamp array does not meet this criterion, and spillage on the aperture hardware is expected. We note that a rim angle of 45 deg is commonly used for parabolic dish systems and for solar furnaces as this configuration results in the greatest theoretical concentration ratio [15]. Optical ray tracing models were developed to directly assess likely power delivery to the aperture and the resulting flux distribution. For the analysis, we used the commercial ray tracing package, Trace Pro (Lambda Research Corporation (Littleton, MA) [16]), to evaluate both the four lamp and five lamp configurations. The results of this analysis are shown in Fig. 6 and have been truncated to fit on a 6 cm diameter target. Fig. 6 Fig. 6 Close modal The simulations showed that the both the four and five lamp arrays would be sufficient to meet the power delivery and flux targets of the thermochemical reactor prototype. Of the two configurations, the four lamp array spills less energy on the hardware surrounding the aperture owing to its slightly smaller rim angle. #### Fabrication. The four lamp simulator array was constructed from identical lamp-reflector modules attached to a common structural frame, as shown in Fig. 7. Fig. 7 Fig. 7 Close modal Each module must be positioned with respect to azimuth and elevation to allow the beam from each to be overlapped at the target plane, thus maximizing intercept by the aperture. Figure 7 shows each module hanging from a single support beam. Each of these beams is, in turn, attached by a bolted connection at a central hub. This connection allows for motion in the azimuthal direction. Position in elevation is constrained by a hinged connection between each module and the overhead beam along with a second connection via a turnbuckle. The turnbuckle is used to change the elevation angle of a given module within a specific range. Alignment of the HFSS with the target was facilitated using laser diodes mounted to the lamp holder and projecting along the module optical axis. These lasers allowed for a simple visual approximation of beam intercept with the target. Arc lamps require electronic control systems, commonly referred to as ballasts, to moderate the current flow to the lamp during startup and operation. There are many ballast manufacturers producing hardware sufficient to drive any of the lamps that we considered. The price for a ballast varies from$500 to $4000 per lamp for lamps requiring up to 3 kWe of input power. The range in price is largely due to the features included with the ballast, which can be nothing more than the solid state electronic core at the lowest price point, or a complete rack-mounted unit with integral cooling, control, and display and alarm functions. The ballasts used for these modules were dual output 6 kWe ballasts manufactured by Powergems Limited (model EB2X250PR) (PowerGems (Manchester, UK) [17]), provided by Sandia from a previous project. These cost about$6000. Alternatively, we have also used a single output, 2500 We ballast from CCI Power Supplies, LLC (model MHS2500M7ZC) (CCI Power Supplies LLC (Pardeeville, WI) [18]). This option consists of the electronic core and costs about $700 for a single 2500 We output but requires the addition of a suitable housing, cooling system, and control/display/alarm systems. These features can be fabricated for an additional$100 per ballast.

Control of each simulator module is done through the ballast using a low voltage DC signal, and it is desirable to include a safety interlock that shuts the system down in the event of a component failure, such as loss of cooling that could lead to lamp failure. We configured the ballasts to operate with a current switch on the power connection to the cooling fans. This normally closed switch opens when power to the fan is interrupted, thus signaling the ballast to cut power to the lamp. A schematic of our control system is shown in Fig. 8.

Fig. 8
Fig. 8
Close modal

#### Array Characterization.

The simulator array was characterized experimentally in the same manner as single modules, using a combination of full beam calorimetry and optical flux mapping in beam down configuration shown in Fig. 9. The results of the characterization effort are summarized in Fig. 10.

Fig. 9
Fig. 9
Close modal
Fig. 10
Fig. 10
Close modal

The experimental peak flux of the HFSS is 2890 kWth/m2, and the power delivered to a 6 cm diameter aperture is 3531 Wth. Both of these parameters, as well as the overall shape of the flux map at the aperture, are in good general agreement with the values predicted by the optical model. The cross section of the flux map shown in Fig. 10(b) includes upper and lower limits for the experimental data corresponding to the uncertainty associated with their measurement.

#### Bill of Materials and Cost Summary.

The bill of materials for the four lamp HFSS array is presented in Table 1 along with a cost summary. The ballast cost is given as a range to reflect the fact that multiple options exist for this hardware at different price points. In addition, some of the reported costs should be considered high estimates given that they can decrease with the number of units purchased. Finally, we note that there are minimum purchase quantities tied to certain components such as lamp holders and ignitors, depending on vendors.

Table 1

Costs associated with the hardware required for a single 2500 We simulator module

Part(s)Total cost
Structural components and hardware$233.50 Electrical components and hardware$309.09
Misc. hardware$55.99 Ignitor$45.00
Ballast$711.00 Reflector and mount$616.00
Fast-fit lamp$371.57 Lamp socket$30.00
Current switch$20.00 Cooling fan$76.20
Total$2468.35 Part(s)Total cost Structural components and hardware$233.50
Electrical components and hardware$309.09 Misc. hardware$55.99
Ignitor$45.00 Ballast$711.00
Reflector and mount$616.00 Fast-fit lamp$371.57
Lamp socket$30.00 Current switch$20.00
Cooling fan$76.20 Total$2468.35

Each simulator module required roughly 5 h to assemble once the mounting plates were machined. An expanded version of Table 1 including individual part numbers and vendor information is available online along with the design drawings for the single module and four-module HFSS [19].

#### HFSS Array Configurations.

Other HFSS configurations were investigated via optical modeling to define the performance space for arrays having three to seven modules each built from an E-813-1 reflector and a Philips MSR Gold 2500W/2 FastFit lamp. In these simulations, the geometric configuration of the lamps in the array was set to minimize the rim angle and the resulting flux maps were calculated. The per-module cost of materials from Table 1 was then applied to the number of modules for a given configuration to produce a total material cost for each configuration. The array support structure hardware cost was estimated to be $250 per module based on our specific beam-down configuration. Finally, the equivalent cost of thermal energy delivered to the target was calculated. The results of this analysis are presented in Table 2 [1214,1618,20]. Table 2 Predicted performance of various HFSS configurations Number of lampsPeak flux (kW/m2)Power (W)Average flux (kW/m2)Cost of materialsThermal cost ($/kWth)
323662535897$8155$3217
4298433281178$10,873$3267
5358040601437$13,591$3348
6424748391712$16,310$3371
7485755791974$19,028$3411
Number of lampsPeak flux (kW/m2)Power (W)Average flux (kW/m2)Cost of materialsThermal cost ($/kWth) 323662535897$8155$3217 4298433281178$10,873$3267 5358040601437$13,591$3348 6424748391712$16,310$3371 7485755791974$19,028$3411 The results in Table 2 show that the lowest cost HFSS on a$/kWth delivered basis is achieved by a three lamp array. The seven lamp array delivers the most power, at 5.6 kWth at the highest peak flux of 4900 kWth/m2. The cost of the array per kWth increases as the array size increases due to the cosine loss reducing the effective flux delivered by each subsequent lamp to the target plane. That said, the difference in price between the least expensive and most expensive configuration is only 6%.

## Conclusions

We have shown that a relatively inexpensive high flux solar simulator based on metal halide lamps can be built from commercially available components at a cost of about $3300/kWth delivered to the target plane. A four lamp array of this type delivers 3.5 kWth at a peak flux of 2900 kWth/m2. Depending on the constraints of the specific application it is possible to further increase the size of the array to accommodate experiments with higher power and flux demands. We presented several different HFSS configurations ranging in size to a seven-lamp array that delivers 5.5 kWth at a peak flux of 4900 kWth/m2 at an estimated hardware cost of$3400/kWth.

## Acknowledgment

The authors gratefully acknowledge funding from the Hydrogen Production and Delivery program element of the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy's Fuel Cell Technologies Office (FCTO). The FCTO supports research and development of thermochemical hydrogen production technologies.

## Funding Data

• Fuel Cell Technologies Program (DE-FOA-0000826).

## References

1.
Ho
,
C. K.
,
Wilson
,
S.
,
Sandford
,
M.
, III
,
Anderson
,
R.
R.,
Kolb
,
W. J.
,
Ray
,
D. A.
,
Fajardo
,
D.
,
Christian
,
J. M.
, and
Yellowhair
,
J.
,
2016
, “
High-Flux Solar Simulator With Automated Sample Handling and Exposure System (ASHES) (Presentation)
,” Sandia National Laboratories (SNL-NM), Albuquerque, NM, Report No.
SAND2016-6135C
.https://www.osti.gov/servlets/purl/1368436
2.
Petrasch
,
J.
,
Coray
,
P.
,
Meier
,
A.
,
Brack
,
M.
,
Häberling
,
P.
,
Wuillemin
,
D.
, and
Steinfeld
,
A.
,
2007
, “
A Novel 50 kW 11,000 Suns High-Flux Solar Simulator Based on an Array of Xenon Arc Lamps
,”
ASME J. Sol. Energy Eng.
,
129
(
4
), pp.
405
411
.
3.
Ekman
,
B. M.
,
Brooks
,
G.
, and
Rhamdhani
,
M. A.
,
2015
, “
Development of High Flux Solar Simulators for Solar Thermal Research
,”
Sol. Energy Mater. Sol. Cells
,
141
, pp.
436
446
.
4.
Gallo
,
A.
,
Marzo
,
A.
,
Fuentealba
,
E.
, and
Alonso
,
E.
,
2017
, “
High Flux Solar Simulators for Concentrated Solar Thermal Research: A Review
,”
Renewable Sustainable Energy Rev.
,
77
, pp.
1385
1402
.
5.
Sarwar
,
J.
,
Georgakis
,
G.
,
LaChance
,
R.
, and
Ozalp
,
N.
,
2014
, “
Description and Characterization of an Adjustable Flux Solar Simulator for Solar Thermal, Thermochemical and Photovoltaic Applications
,”
Sol. Energy
,
100
, pp.
179
194
.
6.
Roba
,
J. P.
, and
Siegel
,
N. P.
,
2017
, “
The Design of Metal Halide-Based High Flux Solar Simulators: Optical Model Development and Empirical Validation
,”
Sol. Energy
,
157
(
Suppl. C
), pp.
818
826
.
7.
Codd
,
D. S.
,
Carlson
,
A.
,
Rees
,
J.
, and
Slocum
,
A. H.
,
2010
, “
A Low Cost High Flux Solar Simulator
,”
Sol. Energy
,
84
(
12
), pp.
2202
2212
.
8.
Boubault
,
A.
,
Yellowhair
,
J.
, and
Ho
,
C. K.
,
2017
, “
Design and Characterization of a 7.2 kW Solar Simulator
,”
ASME J. Sol. Energy Eng.
,
139
(
3
), p.
031012
.
9.
Wang
,
W.
,
Aichmayer
,
L.
,
Laumert
,
B.
, and
Fransson
,
T.
,
2014
, “
Design and Validation of a Low-Cost High-Flux Solar Simulator Using Fresnel Lens Concentrators
,”
Energy Procedia
,
49
, pp.
2221
2230
.
10.
Ermanoski
,
I.
,
Siegel
,
N. P.
, and
Stechel
,
E. B.
,
2013
, “
A New Reactor Concept for Efficient Solar-Thermochemical Fuel Production
,”
ASME J. Sol. Energy Eng.
,
135
(
3
), p.
031002
.
11.
Singh
,
A.
,
Lapp
,
J.
,
Grobbel
,
J.
,
Brendelberger
,
S.
,
Reinhold
,
J. P.
,
Olivera
,
L.
,
Ermanoski
,
I.
, Siegel, N. P., McDaniel, A., Roeb, M., and Sattler, C.,
2017
, “
Design of a Pilot Scale Directly Irradiated, High Temperature, and Low Pressure Moving Particle Cavity Chamber for Metal Oxide Reduction
,”
Sol. Energy
,
157
(
Suppl. C
), pp.
365
376
.
12.
Philips Lighting
, 2017, “
MSR Gold™ 2500, /2 FastFit 1CT/4 MSR Gold™ FastFit Special Lamps
,” Philips Lighting, Amsterdam, The Netherlands, accessed Dec. 5, 2017, http://www.lighting.philips.com/main/prof/conventional-lamps-and-tubes/special-lamps/entertainment/touring-stage/msr-goldtm-fastfit/928106405115_EU/product
13.
Optiforms
, 2017, “
Elliptical Reflectors
,” Optiforms, Temecula, CA, accessed Dec. 5, 2017, https://www.optiforms.com/electroforming/elliptical-reflectors/
14.
80/20, 2017, “
80/20 T-Slotted Aluminum Framing
,” 80/20, Inc., Columbia City, IN, accessed Dec. 5, 2017, https://www.8020.net/
15.
Segal
,
A.
,
2010
, “
Optics of Solar Concentrators
,” Weizmann Institute of Science, Rehovot, Israel, accessed Mar. 30, 2018, http://sfera.sollab.eu/downloads/Schools/Optics_of_Solar_Concentrators_Segal.pdf
16.
Lambda Research Corporation
, 2017, “
TracePro
,” Lambda Research Corporation, Littleton, MA, accessed Dec. 5, 2017, https://www.lambdares.com/tracepro/
17.
PowerGems
, 2017, “
2x2500W 19″ Electronic Ballast
,” PowerGems, Manchester, UK, accessed Dec. 5, 2017, https://powergems.com/product/eb2x250pr-2x2500w-19-electronic-ballast/
18.
CCI Power Supplies LLC
, 2017, “
Range of Products
,” CCI Power Supplies LLC, Pardeeville, WI, accessed Dec. 5, 2017, https://www.ccipowersupplies.com/pdfs/CCI-Product-Range.pdf
19.
Siegel
,
N.
, 2018, “
High Flux Solar Simulator Design and Test