Personal weather stations (PWS) are gaining popularity since they provide open meteorological data with high spatial resolution. However, a number of factors may affect the quality of the measurements of these stations. PWS provide irradiance measurements using silicon-photodiode sensors, which are a low-cost and lower-maintenance option compared to thermopile-type pyranometers. Silicon-photodiode sensors present, however, several limitations related to the spectral bandwidth, the device's temperature and other affecting factors, which increase the error in the measurements. This article evaluates the accuracy of irradiance measurements from PWS under all-sky conditions. We propose a calibration method to reduce the uncertainty in the measurements based on the solar zenith angle, the temperature of the device and the clear-sky index. The proposed calibration model is evaluated in 30 personal weather stations from different producers located in 5 climate zones worldwide over several years. An average reduction of the relative mean bias error from 18.4% to 2.8% is achieved based on 5-min instantaneous irradiance samples for clear-sky instances and it is within a +/- 5% tolerance for all-sky conditions, where other metrics are also improved 1-4%. This study helps understanding the nature of uncertainty of irradiance measurements in this increasingly used data source and provides a practical calibration method to increase accuracy of PWS data.