A current clamp connected to a digital oscilloscope was used to monitor the temporal variation of the electrical current into the DC inverter inputs. Fig. 1 shows the normalised DC current flowing into one of the inverter inputs during two different switching regimes, reflecting relative power output from the system. In Fig.1(a) the current trace reflects a two-step switching process, whereby the operating point of the system is switched between MPP, at which 100% of the available power is extracted and OC, at which the DC current into the inverter and the extracted power are zero. Note that under OC condition the cells embedded in the modules emit maximum luminescence intensity, whereas at MPP about 95% of the photogenerated carriers are extracted, reducing the luminescence intensity to close to zero. In Fig.1(b) the operating point is periodically switched in three distinct steps, including an interim step at which 40% of the MPP current (I_mpp ) is extracted from the module, resulting in approximately 40% of the maximum power being extracted. While not the focus of this paper, the ability to switch between different operating points enables calculation of image differences that allow differentiating series resistance variations from recombination defects, as demonstrated in the context of daylight PL imaging by Bhoopathy et al. .

The switching of the power output from 100% to zero takes place in < 150 ms, but due to a long dwell time of about 1.5 s the overall cycle time is about 3 s. This comparatively long cycle time was chosen intentionally in our proof-of-concept studies reported here as a conservative setting. We anticipate being able to perform switching at considerably higher frequencies and with shorter switching times in the near future, noting though that for residential rooftop applications, as discussed here, the above switching rate is sufficient to enable rapid high resolution DPL inspection of all modules of a system within a few minutes.

Individual camera images were acquired with exposure times between 15 ms and 20 ms. Image registration was applied to improve DPL image quality, minimizing DPL image imperfections caused by very significant camera shake. It is noted that camera shake does not notably affect the quality of individual camera images, given the above short exposure times. However, it results in lateral offsets between images, producing artefacts when difference images are calculated. Image registration prior to image subtraction eliminates this unwanted effect.

Fig.2 shows two PL images acquired using inverter-based switching on the residential rooftop system at Wylie’s Baths, as described above. From the fixed camera position the 50mm focal length lens that was used for these measurements provides close-up DPL images showing half a module to one full module, depending on the specific module location on the roof (i.e. distance to the camera). Each DPL image was calculated from 10 image pairs.