Finite-precision effects in digital filters
• Supervisor: G.D. Halikias
• Content: Analytical, Programming (Matlab), Simulation
• Suitable for: EE, XCF students
• Useful third-year options: Digital Signal Processing
Digital filters designed under the assumption of infinite-precision arithmetic sometimes exhibit serious
performance or stability degradation when implemented on a microprocessor or a dedicated DSP chip.
This is due to signal quantisation errors and finite-precision arithmetic effects. In this project it will
be attempted to design robust digital filters that suffer minimal degradation when finite-precision
effects are taken into account. The proposed method is based on certain recent ideas of robust control
which is a systematic method for analysing and designing dynamic uncertain systems. In essence, the
finite precision effects will be interpreted as parametric uncertainty in the filter’s coefficients, and the
design will follow a “worst-case” approach. In addition, it will be attempted to identify optimal filter
structures. Background reading for this project includes linear digital systems and Z-transforms.
1