Displays & Optical Technologies, Inc.

DESIGN AND PERFORMANCE OF A TRIPLE SOURCE AIR MASS ZERO SOLAR SIMULATOR

Phillip Jenkins
Ohio Aerospace Institute, Brook Park, OH 44142
David Scheiman
Ohio Aerospace Institute, Brook Park, OH 44142
David Snyder
NASA Glenn Research Center, Cleveland, OH 44135

Simulating the sun in a laboratory for the purpose of measuring solar cells has long been a challenge for engineers and scientists. Multi-junction cells demand higher fidelity of a solar simulator than do single junction cells, due to a need for close spectral matching as well as AM0 intensity. A GaInP/GaAs/Ge solar cell, for example, requires spectral matching in three distinct spectral bands. A commercial single-source high‑pressure xenon arc solar simulator such as the Spectrolab X‑25 at NASA Glenn Research Center can match the top two junctions of a GaInP/GaAs/Ge cell to within 1.3% mismatch, with the GaAs cell receiving slightly more current than required. The Ge bottom cell, however, is mismatched +8.8%. Multi‑source simulators are designed to match the current for all junctions but typically have small illuminated areas, less uniformity, and less beam collimation compared to an X‑25 simulator. It was our intent when designing a multi‑source simulator to preserve as many aspects of the X‑25 while adding multi‑source capability.

Figure 1 caption: The quantum efficiency of a GaInP/GaAs/Ge solar cell plotted with the AM0 spectrum.
Figure 2 caption: Spectrum of the Spectrolab X‑25 solar simulator compared to AM0.

The X‑25 faithfully reproduces the AM0 spectrum from approximately 350 nm to 750 nm, but beyond that region the spectrum quickly diverges from AM0. This multi‑source simulator takes advantage of this property and uses the X‑25 to provide the bulk of the UV and visible light. A non‑absorbing dichroic mirror (designed and fabricated by Displays and Optical Technologies, Round Rock, Texas) is used as a beam splitter to reflect the blue portion of the spectrum and transmit the infrared. The transmission and reflection curves of the filter show a good cutoff transition at 700 nm, but the reflection bandpass is limited to approximately one octave and begins to transmit excessively at wavelengths shorter than 450 nm. This behavior is typical for high‑efficiency thin‑film dielectric stack filters. Although not ideal, the filtered beam provides sufficient UV and visible light for solar cells and excellent IR transmission for adding in the balance of the AM0 spectrum.

With the cell test plane and the X‑25 simulator mounted 90° to each other, the dichroic mirror is mounted at a 45° angle to reflect the blue portion of the X‑25 beam onto the cell and allow for additional light sources from above the cell plane.

Figure 3 caption: Transmission and reflection of dichroic mirror.
Figure 4 caption: Optical path for combining the IR source with the X‑25 beam.

The initial design strategy for the IR source was to pack as many lamps into as small an area as possible and mount the lamps as far from the test plane as practical. This yields the best collimation angle and beam uniformity. The lamps used are tungsten filament bulbs mounted in a gold reflector. They operate at 12 V, 75 W, with an average life of 4,000 hours (Gilway Technical Lamp, Woburn, MA). The reflector has a nominal beam angle of 14°. The first design used a rectangular array of 40 lamps, but this required two degrees of pointing control per lamp to achieve acceptable uniformity. Without this control, the beam had a significant center hot spot.

A second design capitalized on the natural radial symmetry of the beam pattern. Here, 36 lamps are oriented in a radial pattern, each with one degree of pointing freedom. Each lamp is mounted on an aluminum block with four spring‑loaded screws, allowing approximately ±5° of tilt—enough to sweep the beam from center to edge of the test plane. In practice, all lamps are splayed outward to reduce the center hot spot.

Figure 5 caption: Each IR lamp is mounted on an aluminum block with 4 spring‑loaded adjustment screws.
Figure 6 caption: Spectrum of the tungsten lamp compared to AM0.

The spectrum of an unfiltered tungsten lamp does not adequately simulate the IR portion of AM0. Below 1000 nm, the tungsten lamp’s power density drops compared to AM0—critical for the GaAs junction of a GaInP/GaAs/Ge cell. If the tungsten intensity is raised to match the GaAs current, the Ge cell receives 40% excess current. A filter was designed to shape the tungsten spectrum to match AM0.

Figure 7 caption: Design of the filter to shape the IR lamp spectrum to match AM0.

A third light source was created by filtering a subset of the 36 tungsten lamps. A non‑absorbing filter transmits only 700–900 nm, optimized for adjusting the GaAs middle‑cell current independently of the Ge or GaInP cells. Filters are placed in front of 6–12 lamps depending on spectral mismatch. These filtered lamps are powered separately and controlled independently.

For the 12 filtered lamps: – 4 parallel strings of 3 lamps each – 36 V DC at 25 A

For the remaining 24 lamps: – 4 parallel strings of 6 lamps each – 72 V DC at 25 A

Using all three light sources, the short‑circuit current of each junction of a GaInP/GaAs/Ge cell can be matched to reference values to better than 0.5% in under 30 minutes.

Figure 8 caption: Photograph showing the 12 filtered lamps.
Figure 9 caption: Spectrum of the simulator when all three junctions are properly illuminated, compared to AM0.
Figure 10 caption: Uniformity of the long IR light source over a 20×20 cm area.
Figure 11 caption: Uniformity of the X‑25 simulator over a 20×20 cm area.

The collimation half‑angle is determined by the radius of the IR lamp cluster and the distance above the test plane: 7.6°. Uniformity is most affected by the number of lamps that can be articulated. All lamps contribute to the 700–900 nm region (effectively 36 lamps). Adjustments to the filtered lamps are small (<5%) and do not significantly affect uniformity. In the 900–1700 nm region, 24 lamps contribute, achieving ±1% uniformity over a large area.

Intensity maps were made using an X‑Y plotter with a photodiode. Lamp stability has not been fully characterized, but after a 30‑minute warm‑up, no perceptible drift occurs over four hours. Lamps are flicker‑free due to DC power.

A schematic of the system is shown in Figure 12. For many applications, the single‑source X‑25 is sufficient. The system can be reconfigured by replacing the dichroic mirror with a simple aluminized first‑surface mirror, preserving X‑25 capability.

Hardware cost (excluding X‑25): 25,000 USD (2002). Operating the simulator presents challenges. Ideally, separate calibrated reference cells exist for each sub‑junction. If not, spectral correction or calculated spectral intensity must be used.

We successfully converted a Spectrolab X‑25 into a multi‑source simulator capable of precisely matching the required currents in a triple‑junction GaInP/GaAs/Ge solar cell. The system is inexpensive, preserves X‑25 functionality, has excellent uniformity, good stability, and a collimation angle better than ±8°. Future improvements include adding another xenon light filtered for blue enhancement.

Figure 12 caption: Schematic representation of the multi‑source simulator.

References

Joseph H. Apfel and Alfred J. Thelen, “A Solar Simulator,” Proceedings of the Solar Working Group Conference (2nd PVSC), Vol. II, 7‑1 to 7‑21, PIC‑SOL 209/2.1, Washington D.C., Feb. 27–28, 1962.

Solar Cell Standards Group, Energy Conversion Task Group, West Coast Solid State Devices, “Photovoltaic Solar Simulator Specification,” AIEE, June 7, 1961; revised Sept. 5, 1961.

D. B. Bickler, “The Hoffman Solar Simulator,” Proceedings of the Solar Working Group Conference (2nd PVSC), Vol. II, 6‑1 to 6‑21, Washington D.C., Feb. 27–28, 1962.