Navigation receiver

A method of improving accuracy in a navigation system uses an error model (208) having a first term that models noise and a second term that models indirect signal propagation. The first term may be modelled by a Gaussian distribution (pdf, probability density function) and the second term by an exponential distribution. The distributions may be location (elevation) dependent, or depend on the single-to-noise ratio. By maximising the likelihood of observed data (210) (MLE maximum likelihood estimation), given a number of known error model parameters (208), a number of unknown model parameters can be found. These unknown model parameters can provide a more accurate location, velocity or time estimate for the receiving device. The method and apparatus can be used in global satellite navigation systems such as the global positioning system (GPS) or terrestrial systems where many of the received signals are subject to multipath. Processing may be at the receiver or at a remote server.