Lecture on FIR Filter Design and Parks-McClellan Algorithm

Introduction

• Presenter: Dr. S. Chesu, Professor of ECE at Hyderabad
• Topic: FIR filter design and Parks-McClellan algorithm using MATLAB

Key Concepts

Linear Time Invariant (LTI) Systems

• Systems described by constant coefficient difference equations
• Implementable with finite adds, multipliers, and delays
• H(z): Rational function used in filter design to determine the number and locations of zeros and poles, or equivalently, filter coefficients
• Design involves choosing number and values of filter coefficients

Digital Filters

• Types: Finite Impulse Response (FIR) and Infinite Impulse Response (IIR)
• FIR: N = 0, all zeros, linear phase
• IIR: N > 0, involves lower side lobes for the same number of coefficients

General Considerations

1. Filter Size: Filter length (N) and computation/storage trade-offs
2. Causality: Poles must be inside the unit circle for stability and causality
3. Real-time Implementation: Focus on causal filters for real-time applications
4. Input-Output Relationship: Output depends on current and past input values
5. System Function: Number of finite zeros ≤ number of finite poles
6. Frequency Response: Analyze H(ω) based on frequency performance

Practical Facts

• Causal FIR Filters: H(ω) cannot be zero over any band of frequency except in trivial cases
• Error Analysis: H(ω) magnitude (|H(ω)|) cannot be flat over any finite band
• Magnitude and Phase Response: HR(ω) and HI(ω) are interdependent
• Ideal filters for finite bands of zero response cannot be perfectly implemented

Practical Filters Design Considerations

• Ripple: Pass band and stop band ripples must be acceptable
• Transitions: No infinitely sharp transitions between pass and stop bands
• Frequency Response: Must consider practical applications (e.g., audio speaker design)

FIR vs. IIR Filters

• FIR: Achieve linear phase response, requires many coefficients, high computational demand
• IIR: Uses feedback, mimics analog filters, smaller number of coefficients

FIR Filter Design Methods

1. Simple Design Methods: Truncation and windowing (suboptimal frequency characteristics)
2. Parks-McClellan Algorithm: Optimal method using Remez exchange for FIR filter design

Parks-McClellan Algorithm Overview

• History: Developed by James McClellan and Tom Parks in 1972
• Goal: Minimize pass band and stop band errors using Chebyshev approximation
• Process: Uses a set of steps to iteratively achieve optimal filter coefficients

Steps of Parks-McClellan Algorithm

1. Initialization: Choose initial set of extrenmal frequencies (Ω)
2. Finite Set Approximation: Calculate best Chebyshev approximation on the current extremal set
3. Interpolation: Compute error function E(Ω) across the set of frequencies
4. Local Maximum Search: Find new extremal frequencies where |E(Ω)| is maximized
5. Iteration: Update the extremal set and repeat steps 2-4 until convergence
6. Finalization: Use interpolation formula and inverse discrete Fourier transform to determine FIR coefficients

Practical Example

• Application: Audio systems (e.g., tweeters and subwoofers)
• Performance Metrics: Passband ripple, stopband attenuation, band edge definitions
• Implementation: Practical issues and benefits of FIR design

Conclusion

• Key Takeaways: Understanding of LTI systems, FIR/IIR filter design, Parks-McClellan algorithm steps, and practical applications
• Next Steps: Implementation using MATLAB

Thank you! Like, share, and subscribe!