The quality control of photovoltaic modules in terms of the output power to satisfy the technical specification is of great importance for producers as well as consumers and also represents a major issue of certification procedures. Previous work focused on one-sided specification limits to reject underperforming samples (lots) of photovoltaic modules or solar cells. In the present paper, we generalize the classic acceptance sampling methodology and derive sampling plans on the basis of two-sided specification limits. Those sampling plans can be constructed for arbitrary output power distributions by making use of flash data tables. For the out-of-spec setting, the sampling plans are solutions of rather involved nonlinear equations. Explicit formulas, which resemble known sampling plans, can only be obtained under symmetry assumptions. Further, the solution depends on the ratio of overperforming modules to underperforming modules. We investigate by numerical studies to which extent the required sample size depends on that ratio and the shape of the underlying output power distribution. The application to real examples indicates that in practice, the new approach often results in substantially smaller control samples than classic approaches. Copyright © 2012 John Wiley & Sons, Ltd.