Scaling and round-off noise in multistage interpolators and decimators
Mattias Olsson, Håkan Johansson, Per Löwenborg · 2004
In this paper we present novel methods for scaling and roundoff noise calculation in multistage interpolators and decimators. When a signal is upsampledby $L$ it becomes Cyclo-Wide-Sense Stationary (CWSS) which prevents the use of common design algorithms for scaling. When a signal is downsampled by $M$ the situation is simpler since the stationarity is preserved, however, that case will still be treated for completeness. Finally, we present methods for scaling and noise calculation when the sample rate of a signal is changed by a rational factor $L/M$. The methods are based on multirate identities and polyphase decomposition.