On using correlated sampling to simulate small changes in system response by MCNP

Correlated sampling is a developed technique that enables MCNP to simulate small changes in system response with an acceptable uncertainty at reasonable number of particle histories. In this work the performance of correlated sampling is investigated. It is found that under certain conditions using the default output format of MCNP and the original practice of batch statistics developed for correlated sampling may overestimate the uncertainty of a response change. The cause of the overestimation is analyzed; correspondingly two suggestions are proposed for correlated sampling to ensure correct estimation of the uncertainty. The performance of improved correlated sampling is compared with that of un-correlated sampling computation to demonstrate the effect of correlated sampling. Results show that correlated sampling indeed is a robust technique for MCNP to simulate small changes in system response.