Recursive digital filters with low coefficient sensitivity
M. McCallig, Richard R. Kurth, R. Steel
Abstract
M. McCallig, Richard R. Kurth, R. Steel
Abstract
This paper describes a technique for designing IIR digital filters to a desired magnitude-squared frequency response with auxiliary constraints to control coefficient quantization sensitivity. The procedure can be used to design IIR filters for direct form implementation with a reasonable number of bits, e.g., 12 or 16, in contrast to standard IIR filters which usually require parallel or cascade form realization because of quantization sensitivity. The method will design filters with response characteristics that provide a continuous link between conventional IIR and FIR filters. The relative control simplicity associated with the direct form is shown to have advantages in special-purpose hardware implementations of low-sensitivity IIR filters using LSI arithmetic function circuits. A design example and its 12-bit hardware realization are described.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper describes a technique for designing IIR digital filters to a desired magnitude-squared frequency response with auxiliary constraints to control coefficient quantization sensitivity. The procedure can be used to design IIR filters for direct form implementation with a reasonable number of bits, e.g., 12 or 16, in contrast to standard IIR filters which usually require parallel or cascade form realization because of quantization sensitivity. The method will design filters with response characteristics that provide a continuous link between conventional IIR and FIR filters. The relative control simplicity associated with the direct form is shown to have advantages in special-purpose hardware implementations of low-sensitivity IIR filters using LSI arithmetic function circuits. A design example and its 12-bit hardware realization are described.
Key concepts: Infinite impulse response, Realization (probability), 2D Filters, Quantization (signal processing), Digital filter, Computer science, Finite impulse response, Sensitivity (control systems)