Care should be taken with models where one decides to use multiple phases to run one time-history signal (when for example the time history signal is very long). In principle in each new phase the kernel should continue from the (dynamic) end time/latest step of the previous phase. However, due to an issue in the kernel, in each dynamic phase that follows a previous dynamic phase (with a continued time-history signal) the kernel does not take the last step of the previous phase as starting point but the second last step. The resulting time-history may lead to wrong results. The severity of the issue is dependent on a.o. the used time interval for each phase and the used Newmark alpha and beta parameters.
Note that this issue may occur for time-history files specified as ASCII or SMC files.
For now there is no workaround available. Users should try to use where possible one dynamic phase per time-history signal. Note that you may specify up to 10,000 additional steps per phase.
This issue is fixed in version 2D2011.02.