[en] The neutronics of reactors in operating (critical) states is generally modeled as a mathematical eigenvalue problem, with the multiplication constant, the eigenvalue of the solution, being the neutron 'balancing factor' that brings equality between neutron production and loss terms. In sub-critical assemblies, however, the mathematical problem is no longer an eigenvalue one, but rather an exactly determined linear algebra problem 'driven' by a fixed-source term. In Canada, the Refuelling Simulation Program (RFSP) is the industry-standard two-group neutron diffusion code for pressurized heavy water reactors. This paper describes a modification to RFSP to allow fixed-source calculations, and hence to allow the modelling of sub-critical systems. The new capability in RFSP was applied to a proposed new guaranteed shutdown system ('rod-based' GSS) for the Pickering B nuclear generating station. This system was modelled with both RFSP and a fixed-source Monte- Carlo model using MCNP. Comparisons are made between both the fixed-source and eigenvalue solutions for each code (RFSP and MCNP), and also between the RFSP fixed-source solution and the MCNP fixed-source solution. (authors)