Isfahan University of Technology
In the first step, a noisy signal that has a single-tone noise was generated in MATLAB. Then, we designed a suitable FIR filter for removing the single-tone noise entirely. After that, we investigated the effect of coefficient quantization on the performance of digital filters based on direct and cascade structures. In this case, our target was finding optimum sizes of memories for saving the filters’ coefficients. We concluded that the cascade-based filters require less memory than the direct-based filters.
Software: MATLAB
Course: Digital Signal Processing
In a cascade system, the zeroes and poles of the overall transfer function are divided between several cascade systems.
Filter Without Quantizing the Coefficients:
Direct-based system with 23-bit quantizing coefficient:
Cascade-based system with 14-bit quantizing coefficient:
- Quantizing the transfer function coefficients shifts the location of zeros and poles
- The minimum required bit to quantize the coefficients is 14 bits in cascaded-based system, whereas it is 23 bits in direct-based system.
- The cascaded-based system shows a better performance