A European proficiency test on thin‐film tandem photovoltaic devices

A round‐robin proficiency test (RR PT) on thin‐film multi‐junction (MJ) photovoltaic (PV) cells was run between 13 laboratories within the European project CHEETAH. Five encapsulated PV cells were circulated to participants for being tested at Standard Test Conditions (STC). Three cells were a‐Si/μc‐Si tandem PV devices, each of which had a different short‐circuit current ratio between the top junction and the bottom one; the remaining two cells were single‐junction PV devices made with material representative of the individual junctions in the MJ cells. The RR PT's main purpose was to assess the capability of the participating laboratories, in terms of employed facilities and procedures, to test MJ PV devices. Therefore, participants were requested to perform STC measurements of all cells according to their own procedure, which might not include external quantum efficiency measurements. The European Solar Test Installation (ESTI) of the Joint Research Centre (JRC) provided the reference calibrations against which the participants' results are compared. ESTI made also a verification of the cells performance at STC at the end of the RR PT, in order to allow a comparison between the initial stable state at which the cells were calibrated (just before circulation) and the one they had reached at the end of the RR PT. The overall results of the RR PT are here presented and discussed together with some aspects of MJ PV testing that emerged as not adequately applied or largely missing. Their full implementation is expected to improve the consistency of future results.


| INTRODUCTION
Characterisation of photovoltaic (PV) devices has expanded since some years also to study their performance under working conditions that represent the ones PV modules meet in real installations more accurately than Standard Test Conditions (STC). A milestone in this process has been achieved with the series of international standards on the energy rating of PV modules, 1 which was completed in August 2018 with the publication of the last two parts of the series. 2,3 Still, testing PV devices at STC remains important in order to set a reference point (i) to compare different PV technologies under the same reference testing conditions, (ii) to evaluate different modules of the same technology (e.g., crystalline Si) as produced by various manufacturers and (iii) as prerequisite to assess the variation in PV module and PV technology performance with change of operating conditions. 2 Testing and calibration of single-junction (SJ) PV devices usually follows the broadly applied procedures of the standard IEC 60904- 1,4 with support of other standards by the International Electrotechnical Commission (IEC). [5][6][7][8] The latter are necessary in order to adjust the measurement conditions to STC or to correct the measurement results to values that have to be reported at STC. In particular, the measurement of the spectral responsivity (SR) according to IEC 60904-8 7 and the calculation of the spectral mismatch (SMM) according to IEC 60904-7 6 are crucial steps for reliable testing and calibration of any PV technology. This is valid in general for all measurement procedures, regardless of whether the correction for the SMM is applied a posteriori analytically or, on the contrary, made a priori by adjusting the solar simulator's irradiance before the currentvoltage (I-V) measurement. The latter is indeed also a correction for SMM, achieved by using a reference device to set the effective irradiance as per IEC 60904-7. 6 This procedure in turn involves the SR of the reference device and of the device under test (DUT), the solar simulator's spectral and total irradiances and the reference spectrum 9 in the same way they are required for the a posteriori correction.
In the case of monolithic multi-junction (MJ) PV devices, the measurement procedure is more complex than for SJ, due to the intrinsically complex nature of MJ PV devices. 10  Although quite a recent introduction in the international standardisation of PV, testing of non-concentrating MJ PV devices is subject of two IEC standards that were published together in May 2017.
The IEC 60904-1-1 11 deals with the measurement of the I-V characteristics of a MJ PV device. The IEC 60904-8-1 12 sets the requirements for their SR measurement, which involves on the one hand the use of specific bias light to activate the junction(s) not under test significantly more than the junction to be tested and, on the other hand, the use of a bias voltage to bring and keep the junction under test to short-circuit current conditions. While the measurement procedures for most SJ PV devices were already well established and systematically applied in many laboratories, with a large variety of expertise levels and available facilities, MJ PV testing was still not fully integrated in the procedures of all laboratories at the time of the organisation of the measurement comparison reported here. This was also partly due to the lack of an internationally agreed standardised procedure to test them, although prenormative research had already produced scientific publications on the topic (see, e.g., other studies [13][14][15] ). In addition, as good practice for measurement and testing laboratories, and even required for calibration laboratories accredited to ISO/IEC 17025, 16 in the last 30 years several interlaboratory comparisons and round-robins have already been organised for SJ PV testing at STC. [17][18][19][20][21][22][23][24][25] The World Photovoltaic Scale (WPVS) itself, which nowadays has become a sort of alias to refer to a solar cell with some standardised package, was in fact established in 1999 as measurement scale for PV starting from a dedicated world-wide intercomparison on calibration of PV cells. [26][27][28] Similar measurement comparisons are less numerous and less geographically wide in the case of MJ PV devices. 21,[29][30][31] Also, a large part of those available is mainly related to concentrated PV (with reference to a total irradiance larger than 1000 W/m 2 ). [29][30][31] Due to the limited number of measurement comparisons specific for terrestrial non-concentrating MJ PV devices, a first round-robin (RR) test on thin-film MJ PV cells was organised within the FP7 infrastructure SOPHIA project. 32 However, the test could not be completed due to technical issues with the circulated samples, which were not encapsulated and thus easily subject to mechanical damage.
Within the European FP7 project CHEETAH, 33 a second RR on thin-film MJ (tandem) PV cells was organised between 13 testing laboratories, which partly differed from the participants to the SOPHIA testing. The RR was organised as much as possible in the form of a proficiency test (PT), taking the ISO/IEC 17043 34 36 where RR PT organisational aspects still to be improved were also discussed. In this paper, we aim at discussing in more detail all the RR PT results, not only in terms of comparison of the participants' submitted values towards ESTI calibration, but also highlighting the main sources that could explain some of the largest deviations and that can originate from missing steps in the testing procedure.

| Devices
Five PV devices were tested in the RR PT reported here. Among them, three were monolithic double-junction (or tandem) PV cells made of amorphous silicon (a-Si) deposited on top of micro-crystalline silicon (μc-Si). The other two devices were SJ PV cells made of a-Si and μc-Si, respectively. The inclusion of SJ cells in this RR PT on tandem PV devices was driven by the fact that even testing thin-film SJ cells involves expertise and specific steps in the measurement procedure that may not be correctly or completely available at all laboratories.
The additional request to test separately also SJ PV cells of the same technology as those composing the tandem cells seemed quite reasonable in this RR PT, because checking the participants' capabilities in testing thin-film PV technologies was one of the targets of the RR PT.
Regarding the three tandem cells, they were prepared in such a way as to have a different ratio (i.e., balance) of the short-circuit current of the top junction as compared to the bottom junction. As well known, every series-connected PV device is limited by the current produced by the cell that less effectively responds to the actual operating conditions. From this point of view, a MJ PV device can be looked at as a special series-connected PV device, where one half of the cells (in the case of two junction types) is built and electrically contacted on top of the other half. However, contrary to seriesconnected SJ PV, in a MJ PV device the SR representative of one junction type is usually quite different from the other(s), with the result that the electrical limitation of the device is attributed generically to one of the junctions (called the limiting junction) rather than to a single cell. A currents' ratio or balance (CB) can then be calculated, 11 conventionally taking the junctions in the order in which they see the incoming light (i.e., top towards bottom). Usually, this ratio is referred to STC in order to set the reference conditions in the same way it is done for the electrical performance of the DUT. When the top junction is the limiting one, the CB value is smaller than 1. On the contrary, when the bottom junction is limiting, the CB value is larger than 1. In the case both junctions (ideally) deliver the same current, the ratio equals 1, and the PV device is defined matched. Of the three tandem cells included in this RR PT, one was identified by the producer as top-limited, one was nominally bottom-limited and the third was nominally matched. Table 1 lists the five PV cells together with the PV technology and the nominal junctions balance (where applicable) that characterise them. ESTI codes are used to identify them.
Each cell had a nominal active area of 1 × 1 cm 2 and was mounted in a metallic robust case under a glass window of about 3 × 3 cm 2 , to ensure the mechanical protection of the cell during both transportation and testing ( Figure 1). In particular, the case was provided with standard LEMO connectors (see Figure 1A) to make the connection operations easier and safer than with bare cells (used in the previous RR). The four-wire connection configuration (Kelvin probe) was used for them. Moreover, the solid-metal case assured good thermal conductivity between the cell and the external surface, facilitating its temperature control during I-V and SR measurements.
A Pt100 temperature sensor was also integrated inside the case close to the solar cell and was connected to the outside via a LEMO connector. In order to avoid misconnections, the LEMOs were labelled PV for the cell and RTD for the Pt100 (see Figure 1A). Some connecting cables and adapters to banana connectors were provided as well, in order to improve the reproducibility of the measurements and broaden the connection options for the participants.
The five PV cells were made by Jülich Research Centre according to previously published procedures. 37

| Organisation and protocol of the RR PT
As mentioned above, the purpose of the RR PT was to evaluate the capability of the participating laboratories in testing thin-film MJ PV cells (described in Section 2.1) by using their procedures and facilities.
As the IEC standards for MJ PV devices 11,12 were not yet published at the beginning of the RR (although under advanced stage in the approval process 40,41 ), the RR also aimed at collecting information about the possible improvements to be adopted in the measurement procedures as applied to MJ PV devices by laboratories active in European projects.
The RR was organised within the CHEETAH project 33 as a PT for 13 participating laboratories, also building on a failure analysis of the SOPHIA project's RR and on prior experience in interlaboratory comparisons. [21][22][23]25 The reference measurements were independently provided by ESTI, which calibrated all devices at the beginning of the PT (see Section 2.4). The devices were again shipped to ESTI at the end of the RR PT for a final verification and calibration. The standard ISO/IEC 17043 34 was taken as guideline for the evaluation of the participants' results, as already done in previous similar exercises for PV. 23,25 As initiator of the RR within the CHEETAH project, Helmholtz-Zentrum Berlin (HZB) acted as the coordinator (but not as the official provider in the sense of ISO/IEC 17043 34 ) of the RR PT, although it was also a PT participant. This double role of one participant is usually avoided in PTs organised by official PT providers for testing laboratories, in order to assure unbiased results assessment. However, two elements allowed considering HZB position not detrimental to the good development of this specific RR. First, HZB's double role had been taken into account within the CHEETAH project, with no objection by any participant. Second, although the participants' results were submitted to both the coordinator and ESTI, the data analysis of the overall RR PT was made by ESTI, with no disclosure of the RMPT calibration value before the end of the project.
A guideline was circulated to the participants before the beginning of the RR, with instructions on how to handle the DUTs and which type of information was needed in order to compare each participant's results to the RMPT values by ESTI. The participants were required to test each DUT at STC according to their standard procedure for the specific type of DUT. In particular, they were asked to apply their usual procedure for testing MJ PV devices, with possibly incomplete or incorrect steps in the specific applied procedure. However, identifying such insufficiencies was also part of the study, allowing a realistic verification of the capability of the participants in terms of facilities and procedures and the identification of possible areas for measurement improvements and best practices sharing.
In addition, as not all laboratories could perform an adequate preconditioning of the PV cells before testing them, it was agreed that all participants should measure the DUTs without further stabilisation (other than the one performed at ESTI before the reference calibration). This could affect the measurements at successive laboratories if changes related to metastability of the devices occurred, but at the time of the RR organisation it was considered a balanced approach to verify the measurement procedures at the laboratories without affecting the duration (and the cost) of the overall RR PT. To monitor possible exposure to high temperatures during the shipments, which might alter the stabilisation state of the a-Si based PV cells, the devices were provided with irreversible thermal-sensitive labels attached to their case (see Figure 1B) in Section 2.1), and the participants were asked for reporting any indication of heat excess. No mechanical monitoring was used to identify mechanical shocks during transportation, because the latter was considered a less likely threat for these robustly encapsulated thin-film cells.
In general, the procedure to attain STC for I-V measurement can be done either a priori, by adjusting both intensity and spectrum of the solar simulator with a spectrally-matched reference cell (RC) (in this case, necessary for all types of DUTs regardless of the number of junctions), or a posteriori, by applying the SMM correction, 6 which in the case of MJ PV has to be the one calculated for the limiting junction (defined in Section 2.1). In the case of an a posteriori correction, the SMM had to be submitted, too. It was also requested to Note: For the junction-limitation certified by ESTI, see Table S1 in the supporting information. Abbreviation: PV, photovoltaic.
indicate which junction was found limiting for the tandem cells.
Beside I-V curve measurements, SR data were required in the case SR measurement was part of the laboratory's procedure. For the MJ PV cells, the information to be submitted for SR measurements included the bias voltage applied to the DUT as well as the bias light spectrum.
Finally, all the four main I-V parameters, namely, short-circuit current I SC , open-circuit voltage V OC , maximum power P max and fill factor FF had to be reported together with a measurement uncertainty (UC) estimate.
The guideline included also a description of the devices to be measured (see Section 2.1) with their nominal characteristics as assigned by the producer. In this way, a suitable voltage limitation could be used during the I-V measurements in order to avoid damaging the thin-film cells by reverse overcurrent.

| Participants
The laboratories participating in the RR PT were 13 (see Appendix A).
Each of them had its own facilities and expertise to test MJ PV devices. All of them used a solar simulator to measure the I-V curves.
Some of them did not adjust the solar simulator other than for total irradiance as read by a c-Si RC.

| Pre-conditioning before calibration
Before the actual calibration measurements, the cells went through the ESTI procedure for their pre-conditioning, which was carried out according to IEC 61646 39 between December 2015 and January 2016. Such a procedure is based on international standards developed at the IEC. In particular, at the time of the start of the RR, the standard IEC 61646 39 for design qualification of thin-film PV technologies was in place and therefore applied before the RMPT. As first step, STC measurements at the Wacom steady-state solar simulator of ESTI were performed to set the initial state of each DUT. Although it was not required by the pre-conditioning procedure according to IEC 61646, 39 the SR before the pre-conditioning was measured as well in order to (i) calculate the SMM to be applied to the I-V measurements, (ii) identify the limiting junction of the tandem DUTs before preconditioning and (iii) verify if any change to the limiting junction could occur due to the light-soaking procedure. Each DUT was then connected to a resistor, which was sized to keep the DUT approximately at the maximum power point as calculated from the initial I-V curves.
The DUTs were finally mounted on a dedicated holding structure and put in a ventilated light-soaking chamber for a total of about 234 kWh, divided in three steps. The temperature of each DUT was constantly monitored by a calibrated Pt100 sensor attached to the DUT case and connected to a calibrated temperature reader. The temperature was maintained at (45 ± 5) C for the whole pre-conditioning.
The first pre-conditioning step had to be particularly long (about 133 kWh) due to temporary unavailability of the Wacom solar simulator. The second and the third lasted about 50 kWh each. At the end of each light-soaking step, the five DUTs were disconnected from their load, tested at STC at the same solar simulator and with the same RC as in the initial measurements and finally repositioned in the lightsoaking chamber to continue the pre-conditioning. As these intermediate measurements are only relative measurements and the Wacom irradiance is known to be stable from periodic characterisation, no SMM correction was applied to them, and the criterion for stability check 39 was verified by comparing the maximum power values as calculated directly from the measured I-V curves.
For the sake of completeness and guidance to the reader, it is worthy to note that the CB value for a MJ DUT can generally be expected to change whenever the SRs of the junctions of which it's composed change relatively to each other during the pre-conditioning.
However, the relevance of this possibility and the need for its verifica-  Table S2 of the supporting information together with the Z i factor defined in IEC 60904-1-1. 11 The absolute SR of the stable DUTs is shown in Figure 2; the normalised EQEs of the DUTs calculated from ESTI SR measurements before and after the preconditioning are reported in Section 3.2 (Figures 8, 9 and 10), which is specifically dedicated to the SR results of the RR PT. The two spectra of the Wacom and the AM1.5 reference spectrum are shown in Figure S1 of the supporting information. Finally, it must be noted that no attempt was made to improve the balance of the Wacom's spectrum beyond what was already its status in the case of the measurements before the pre-conditioning, as the I-V measurements on the as-received DUTs were just used for stabilisation purposes and ESTI did not issue any calibration certificate on the basis of those measurement results.

| Electrical parameters assessment
For each laboratory and device, the percentage deviations in I SC , V OC , P max and FF from the relevant RMPT value were calculated; they are reported in one graph per device in the following subsections. In the graphs, the UC of the RMPT value is shown as a shaded area symmetrically drawn around the x-axis. The RMPT's UC is shown only for P max (light-grey area) and FF (dark-grey area) to limit the complexity of the graphs while preserving the relevant information. Indeed, in the graph, the area of the ESTI UC for I SC would be slightly narrower than that of P max , and the one for V OC would almost disappear within the x-axis line.
In addition, also the measurement UC stated by each laboratory,   For the remaining four laboratories (i.e., #2, #4, #7 and #13), which were not affected by the mentioned electrical issue, the deviation in P max is strongly connected to the deviation in I SC , which is well beyond ±5% of the RMPT value. This is likely related to issues with SMM correction. They can originate from two sources. The first is the use of non-spectrally matched RC together with no or insufficient SMM correction (e.g., due to the test spectrum used to calculate SMM). The second is the incorrect or incomplete adjustment of the solar simulator spectrum, which in turn can be linked again to issues with the RC(s) used to adjust it. Laboratory #2 did not measure the EQE and did not adjust the spectral irradiance of the solar simulator used for the I-V curve measurements. Therefore, no spectral correction was applied neither a priori nor a posteriori. Laboratory #4 provided UCs, which are however so small that they are not visible in the graph. At this laboratory, there seems to be no issue neither with V OC nor with FF, so the large deviation in P max and I SC is extremely likely to be due to SMM correction (either a priori or a posteriori) or incorrect calibration of the RC. Laboratories #7 and #13 corrected (either a priori or a posteriori) the I-V curve measurements for SMM, but it is clear that the procedure to achieve the SMM correction, the RC calibration value or a combination of these two did not lead to a satisfactory result.
Finally, for the sake of completeness, it has to be noted that no pre-conditioning of the DUT was done at the participants' premises because not everybody could do it. As a-Si is intrinsically unstable and the RR lasted more than 1 year, it could also be expected that the degradation of the a-Si stable state achieved at ESTI at the beginning of the RR partially affected the results of some of the last laboratories in the RR progression. A deviation of −3.6% in P max from the RMPT value was measured at ESTI (EU in Figure 3) at the end of the RR, when the DUT returned there, and before any further preconditioning to take it back to a stabilised state. The additional stabilisation brought this DUT to a P max value −10.9% (ES in Figure 3 Figure 4 shows the results for RR82, which is the μc-Si DUT. The change in P max measured at ESTI at the end of the RR PT was +3.9% (EU in Figure 4), which after further stabilisation went down to +0.5% (ES in Figure 4). The latter value is well within the UC of ESTI measurement. However, from the temporal sequence of the laboratories (not shown here), none of them submitted a deviation that can be explained by such a degradation in relation to their temporal positioning.

| RR82 (μc-Si SJ DUT)
Finally, it is worthy to note here that this DUT has SR closest to the SR of a typical c-Si RC in comparison to all other DUTs (see RR82 in Figure 2). However, it would be wrong to assume that no SMM correction were needed, as SMM is a measurement of both how far the SR of the DUT is from that of the RC as well as how much the test spectrum differs from the reference AM1.5. This is evident for laboratory #1, which did not correct for spectral deviations from STC neither a priori nor a posteriori and whose measurement UCs are not sufficient to explain the deviation from the RMPT value for neither P max nor I SC . Figure 5 reports the results for RR83, which was nominally labelled as the top-limited DUT (see Table 1) and certified by ESTI as such at the beginning of the RR (see CB in Table S2 in the supporting information The P max change for this cell during the RR PT was of +4.5%

| RR83 (a-Si/μc-Si MJ DUT)
(EU in Figure 5), which returned to +0.0% after the final stabilisation (ES in Figure 5). None of the laboratories that show deviations beyond ±5% can relate them to this because of their positioning in the RR PT temporal sequence. Figure 6 gives the results for RR84, which was nominally labelled as bottom-limited (see Table 1) and certified by ESTI as such at the beginning of the RR (see CB in Table S2). However, this cell showed  Figure 6 correspond to the ESTI measurements made at the end of the RR PT before and after the pre-conditioning, respectively.

| RR84 (a-Si/μc-Si MJ DUT)
The light-blue-shaded area helps in separating them from the results of the RR PT participants, too.  Figure 6), on the basis of the measurements performed at ESTI when the cells returned there at the end of the RR PT and before further stabilisation. The P max value measured at ESTI after the additional stabilisation was −2.0% of the RMPT value (ES in Figure 6), which can be partly related to the change found in the F I G U R E 6 Percentage deviations from ESTI RMPT value for I SC , V OC , P max and FF of RR84 (a-Si/μc-Si). RMPT value is represented by the x-axis, and its combined expanded UC (k = 2) is shown for simplicity only for P max (light-grey-shaded area) and FF (dark-grey-shaded area  Figure 7 shows the results for RR85, which was nominally a matched cell (see Table 1) and certified by ESTI as top-limited at the beginning of the RR (see CB in Table S2 value. In addition, the submitted FF values do not fall within the UC band of the relevant RMPT value, but it is reasonable to assume that in all cases a sensible UC estimate associated with both P max and FF might make these results agree with the RMPT value. A similar reasoning might be partly applicable to laboratory #9 and partly to laboratories #3 and #10; however, for all of them, the deviation in P max is accompanied by a similar deviation in I SC , which might be sign of incorrect or incomplete evaluation of the spectral correction.

| RR85 (a-Si/μc-Si MJ DUT)
The last five laboratories are all outside the arbitrary threshold of ±5% of the RMPT value, although laboratory #11 is just across it (+5.0%). However, this laboratory submitted UC together with the results (UC of P max hidden by the marker's size), and they are not sufficient to cover the deviations from the RMPT values. Therefore, it should analyse in more detail the causes at the origin of the resulting deviations, as they could likely derive from incomplete or incorrect spectral correction (either a posteriori or a priori), also looking at the deviation in I SC and FF. The deviation of laboratory #12 in P max , instead, is not connected to a similar deviation in I SC , which shows good agreement with the RMPT value, but essentially to the one in FF, which might indicate issues with the connections together with an incorrect balancing of the two relevant components of the spectral irradiance. For the remaining three laboratories (#1, #2 and #13), the deviation in P max is evidently associated to a similar deviation in I SC , thus suggesting that these laboratories, as others to minor extent, should carefully evaluate their procedures and/or calibration value of the RC(s) for MJ testing.
The change in P max of this DUT during the RR PT was +2.7% (EU in Figure 7), on the basis of the measurements performed at ESTI when the cells returned there at the end of the RR PT and before further stabilisation. The P max value measured at ESTI after the additional stabilisation was +0.4% of the RMPT value (ES in Figure 7), well within the UC of the measurement. Again, no laboratory of those outside the ±5% threshold can sensibly explain the deviation from the RMPT value in relation to its positioning within the temporal sequence of the RR PT.

| SR measurements
Only a qualitative comparison was made for the SR of the DUTs because of the absence of results for some participants and, more importantly, because most of those available were submitted without UCs. The comparison is shown in the following as normalised EQE plots. The latter do serve to compare the qualitative agreement of the shape of the measured EQE (or SR) to the ESTI SR reference measurement, but they cannot be used to derive quantitative information on the SR of the DUTs in terms of absolute units. The shape of the EQE is enough, though, to evaluate the SMM correction of the I-V curve measurements for SJ PV DUTs; it is only one piece of the necessary information for MJ PV DUTs, for which the limiting junction should be also determined by quantitative assessment. In the same way, they cannot show on their own whether the SR of the DUT was affected in absolute terms by the pre-conditioning or by the long-term degradation of the cells during the entire RR PT. For the MJ PV cells, though, F I G U R E 7 Percentage deviations from ESTI RMPT value for I SC , V OC , P max and FF of RR85 (a-Si/μc-Si). RMPT value is represented by the x-axis and its combined expanded UC (k = 2) is shown for simplicity only for P max (light-grey-shaded area) and FF (dark-greyshaded area). The error bars shown represent the UC of the submitted result as stated by the participant. An inset is shown for the points falling outside the main range of the deviations. The EU and ES values correspond to the ESTI measurements made, respectively, before and after the pre-conditioning at the end of the RR PT. Note that the laboratory identification number may be different for the other DUTs [Colour figure can be viewed at wileyonlinelibrary.com] they can give a visual representation of the relative ratio between the two junctions. Where the laboratory originally measured SR data (as it is the case for ESTI, too) and submitted them as such, the data were converted to EQE values as per: where λ is the wavelength, h is the Planck's constant, c is the light speed and q is the elementary charge.
In all the graphs, ESTI measured data points are displayed in the same way as in Figure 2  1. the information on the ratio between the EQE of the two junctions is correctly preserved as submitted by the participants, as opposed to if both individual EQEs were normalised separately; 2. the possible metastability of the μc-Si junction is deemed to be less significant than the one the a-Si junction could have. Therefore, it is used as the relative reference between the two junctions.  In most of the cases, though, an important cause for the deviations in the SR measurement of the MJ PV cells can be related to the missing or incorrect application of the bias voltage, which is a crucial step in those measurements. This determines not only the correct shape of the SR of the individual junctions but also the correct ratio between them.

| Key points for good practice in SJ and MJ PV testing
Some important advice and warnings can be derived from the detailed discussion of the results that has been made in the previous sections for both SJ and MJ thin-film PV devices, which were tested during this RR PT. We summarise them here in form of check list:  A brief summary of the measurement procedures and facility capabilities at the time of the RR PT is given in the following for the 13 participating laboratories, alphabetically sorted by laboratory's acronym.
This sorting has been chosen to preserve as much as possible the anonymity of the reported results, while giving some information on the capabilities of each laboratory. The order of appearance in this list is therefore not connected in any way neither to the actual temporal sequence of the RR PT nor to the actual scoring of the laboratories as reported in this paper.

A.1 | AIT (Austria)
The procedure to attain STC for I-V measurement was done a priori, by adjusting both intensity and spectrum of the solar simulator with a spectrally-matched RC. Therefore, no specific measurement of the EQE was used. In detail, the SMM of solar simulator and DUTs was Two different bias-light sources were used: a blue LED peaked at 450 nm, for which the RC gave an I SC equivalent to 300 W/m 2 , and a dichroic lamp with a long-pass filter (cut-off at 760 nm), for which the RC gave an I SC equivalent to 100 W/m 2 .

A.4 | CREST (United Kingdom)
The testing procedure included both I-V curve and SR measurements.
The latter were performed under over-illumination, without bias voltage even for MJ PV cells and under a bias light of less than 100 W/m 2 in case only narrow-band lights were switched on to bias the junction not under test. The spectral irradiance under which the I-V curves were performed was measured by a spectroradiometer and was used together with the SR data of each DUT to calculate the relevant SMM. The latter was used to correct the I SC of the measured I-V curves as per IEC 60904-7. 6 In the case of the three MJ PV cells, the SMM of the limiting junction was applied.

A.5 | DTU (Denmark)
The testing procedure included both I-V curve and EQE measurements. Before I-V measurements, the total irradiance was adjusted to 1000 W/m 2 by using a c-Si reference device. The measured data were then corrected for the corresponding SMM, which was calculated for each DUT and between 300 nm and 900 nm (due to spectra measure- The EQE of all cells was measured using a double-grating monochromator with a beam spot smaller than the active cell area. Additional spectrally shaped bias light was used for the tandem devices to measure the EQE of the individual junctions. The I-V curves were measured under a multi-source solar simulator, whose spectral irradiance was adjusted for each junction by adjusting single lamps. 48