106 Chapter | 3 Calibration in Array Processing where joint modeling is nonunique without regularization, the array perturba- tions - 0 need to be "small enough." This is because the information from the data can only move 0 a small distance in the direction of the true array parameters, and thus can only reduce the effect of the modeling error on the estimation of to a limited extent as compared to using the nominal value 0 as if it were true. In contrast, when the joint parameterization is unique, the size of the array perturbations is essentially irrelevant provided the true minimum of the respective criteria can be found. Alternative approaches to MAP-based auto-calibration are presented in Jansson et al. (1998), Viberg and Swindlehurst (1994b), and Wahlberg et al. (1991), using a stochastic model of the signal waveforms. 3.5 CALIBRATION USING SOURCES AT KNOWN POSITIONS In the previous section we discussed methods that attempt to calibrate the array using the same data as used for DOA estimation. As was seen, this requires a parametric model with a known structure and relatively few parameters. This is not always a practical assumption. A useful remedy is to collect additional calibration data, normally in an anechoic chamber prior to the actual use of the antenna array. In most cases a single emitter is used, placed at several known