This paper introduces a new, robust, and reliable method to estimate ice cloud microphysical properties from cloud radar reflectivities and visible optical depth. The retrieval is formulated in an estimation theory framework that permits the introduction of optimal combinations of different measurements and a complete characterization of retrieval errors. The sensitivity of the retrieval to the assumed error statistics is assessed performing experiments with variable a priori, optical depth, and forward model uncertainties. Quantitative estimates of the uncertainties show that the average ice water content is retrieved with errors varying between 20 and 30%. The relative error on ice water path is of the same order of magnitude. The retrieval is applied to synthetic and real observations. Retrieved products are checked against other retrieval methods and in situ observations when available. The results compare well with results from other methods. In particular, the method described here is robust and can be applied successfully to a variety of cirrus clouds without suffering from the problems often encountered when using empirically based methods.