Domain Mapping
Mathematical mapping from the continuous s-domain to the discrete z-domain allows analog filter designs to be converted into digital algorithms. The bilinear transform employs a first-order rational approximation of the natural logarithm to map the entire left half of the complex s-plane into the unit circle of the z-plane. This algebraic substitution preserves system stability while avoiding the aliasing effects that occur with alternative conversion methods.
The transformation requires frequency warping adjustments because the continuous and discrete frequency scales do not relate linearly, meaning high-frequency responses are compressed.
Frequency Warping
Non-linear compression of the continuous frequency axis demands pre-warping of the critical analog frequencies before applying the algebraic formula. The designer computes the target discrete frequency and scales the corresponding analog design frequency using a trigonometric tangent function. If this warping step is omitted, the resulting digital filter displays cutoff frequencies that are lower than intended.
This discrepancy becomes severe as the sampling rate approaches the Nyquist frequency.
Filter Implementation
Converting the continuous transfer function yields a set of discrete coefficients for use in difference equations on a microprocessor. The algorithm executes on digital signal processors or field programmable gate arrays using fixed-point or floating-point arithmetic. High coefficients precision prevents rounding errors from driving the discrete poles toward the boundary of instability.
Quantization Effect
Finite word length in the execution hardware introduces numerical noise during multiplication steps. In high-order filters, these rounding errors can cause limit cycle oscillations or alter the frequency response of the algorithm. System designers mitigate these effects by cascading second-order sections instead of implementing a single high-order direct form.