The pyranometer measures the temperature elevation of a blackbody when absorbing the light energy, under a double cupola of glas or quartz (greenhouse effect). Its response is well linear, covers a very large spectrum (flat response until up to 3 micrometers), and has a very good angular acceptance. The accuracy may attain 1-2% with very well calibrated instruments.
The reference cell has a short-circuit current well proportionnal to the irradiance (with a little temperature dependence of +0.05%/°C), but it only measures the irradiance in the spectra according to the sensitivity of the PV cell (for crystalline silicon, limited to 1.12 eV, i.e. less than about 1.1 micrometer). Moreover the usual devices are covered by a flat glas, therefore the angular sensitivity is affected by the IAM (incidence angle modifier) reflexion losses. The accuracy is limited by these two phenomena, and cannot be better than, say, 5%.
The answer to the question now depends on the use of the results.
For measuring the performance ratio (PR) and its stability, the plane-of-array (POA) irradiance is required.
The fact that the reference cell behaves in the same way as the array itself (sames biases, spectral and IAM) will provide a good reference for this measurement: the fluctuations of the PR discrepancies will be low. Moreover some reference cells also provide a measurement of the array temperature (by measuring the Voc).
However the simulation process is waiting for an "absolute" irradiance input, i.e. without these biases ("meteo" measurement quality). It will include a correction for the IAM (PVsyst neglects the spectral effect for crystalline modules), and the PV module's specified behaviour (on which is based the model) is related to the full spectra (STC are for AM 1.5).
Therefore, if you are only interested in the stability of the PV system along the time, the reference cell is well suited.
But if you want to provide prognosis, or analyse and quantify the data in more details using the simulation process, you should use a Pyranormeter as reference.
Precautions about irradiance measurements
The irradiance measurement are rather difficult, and require a great care.
- We strongly advise to perform them in the horizontal plane, not the POA (Plane of Array).
- The horizontality of the instrument should be carefully adjusted.
- The calibration given by the manufacturers is not always reliable. Moreover it may vary along the time (especially during the first years): please carefully check it, if possible periodically, against a reference instrument.
- The instrument should be kept clean, without dust.
- Usually the datalogger will perform a measurement every some few seconds, and average over a predefined time step. It is very important that this time step is well defined, and namely its stability along the life of the measurement. Otherwise this may induce time shifts when importing in PVsyst.
- Check that there are no shades on the instrument at any time of the year (even the diffuse may be affected by shading objects).