The requirement to deal with more complex engine technologies and to achieve higher fidelity in engine control maps suggests that identified dynamic models may in future be preferred to the current pseudo-dynamic fixed operating-point models of current practice. The result of such identification is however known to be significantly affected by the choice of input test signal. Accordingly, input test signals require to be designed to maximise the information content from the identification, subject to the practical limit and rate constraints of the test environment. In this paper, as an exemplar application, an optimal test signal is developed for identification of a dynamic spark mapping using a nonlinear MIMO model, based on a peak-pressure position (PPP) approach. The signal is designed to minimise the parameter covariance subject to the test constraints, and the performance index is selected such that the sensitivity of outputs to unknown parameters is maximized.The approach extends current analytic local-optimisation input design procedures to a numerical global optimisation framework whilst including the limit and rate constraints. The proposed input design method is described and verified on both a linear test system and on the nonlinear PPP spark mapping. For comparison, different numerical optimisation algorithms are employed within the scheme. The optimal test signals are compared in simulation to a large population of non-optimal signals to demonstrate the improved result. Finally the results from the global optimal test signal are compared to PRBS, PRTS and random-walk test signals in a dynamometer based experimental validation, and are found to give significantly superior fit and easier control over the limit and rate constraints.