Perturbing a parameter in a 'small-signal' device simulation

Full computer analysis of the time-varying behaviour of semiconductor devices tends to be very expensive. The ‘small-signal’ method is much cheaper. It may be the only practical option, and in any case is a useful first step. Sometimes, the response of the small-signal analysis to a small parameter change is of interest (e.g. to find the temperature coefficient of an oscillator). Two runs at slightly different values of the parameter may not work, because of truncation errors. A perturbation approach that is cheaper to run and guarantees accuracy is described.