Accurate performance measurement of silicon solar cells

Abstract

Performance measurement is an important part of the solar cell manufacturing process. Two classes of measurement can be considered: accurate calibration – for the creation of reference cells and the setting of records; and routine measurement – for cell sorting and process improvement. This work describes techniques that address both issues – an accurate calibration technique using natural sunlight, and a routine measurement technique using a xenon flashlamp. Both techniques are low-cost, yet in combination they achieve very good accuracy. The light source is very important when calibrating solar cells. Commonly used light sources – good quality solar simulators – are expensive and frequently inaccurate. This work shows that testing of solar cells under natural sunlight is simpler, cheaper, more accurate, and more reliable than all but the most careful simulator measurements. Periodic outdoor calibrations under natural sunlight can therefore eliminate the need for an expensive solar simulator and greatly reduce the need for calibration at standards labs. Solar spectra generated with the model SMARTS2 show that the direct solar spectrum, under clear sky and low air-mass conditions, is an excellent match to the AM1.5G standard spectrum – dramatically better than simulators costing US$20,000 or more. Millions of simulations of a broad range of silicon cells (efficiencies 6-25%) under the modelled direct spectra show that measurement errors of less than 5% are achievable. This is comparable to the reproducibility of results achieved by national standards laboratories. Climate data shows that the required atmospheric conditions occur commonly in summer for all but polar latitudes. Experimental verification of the modelling is encouraging but not yet conclusive. The outcome of the research is a testing ‘recipe’ that uses low-cost equipment and gives an estimate of measurement uncertainty. For routine measurement of large numbers of solar cells, an indoor solar simulator is essential. While simulators can introduce large measurement errors, most of these errors can be eliminated by using a reference cell that is well matched to the test cells. The use of a simpler, lower-cost simulator is possible since reference cells can easily be calibrated under natural sunlight. The second half of this work describes the design of a low cost flash-lamp solar simulator. Most conventional flash testing systems maintain constant light intensity while rapidly sweeping out the I-V curve of the cell. This method has two disadvantages – it requires a flash that is specially engineered to produce constant light output, and the rapid change in cell bias-voltage causes transient errors. The new approach developed in this work essentially does the reverse – it maintains a constant bias voltage on the cell while allowing the light level to vary. This allows the use of a commercial xenon flash, and reduces sensitivity to transient errors. In addition, it extracts a family of I-V curves at different light intensities, whereas the conventional approach only extracts a single I-V curve. The new design has been implemented, using commercial-off-theshelf equipment costing less than US$10,000, and it works well. One of these systems is in use at the Centre for Sustainable Energy Systems, ANU, where it has tested tens of thousands of concentrator cells. A second system has been sold to BP Solar, UK.