Classic Filter Bank — Rigorous DSP Implementation User Guide

Mathematically precise digital implementation of five classical analog filter types with exact -3dB cutoff, pole-zero plots, and optimized processing for accurate frequency/phase responses.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 0.2 (Rigorous DSP Edition, 2025) License: MIT License Repo: https://github.com/ShaiCohen-ops/Praat-plugin_AudioTools

Contents:

What this does Quick start Rigorous DSP Principles Filter Implementations Z-Plane Analysis Optimizations Parameters Applications

What this does

This script implements a rigorously accurate digital version of classical analog filters with mathematically precise -3dB cutoff frequencies, correct pole-zero placements, and optimized processing algorithms. Unlike simplified approximations, this implementation maintains exact analog-to-digital transformation using bilinear transform with pre-warping, proper normalization for filters with zeros (Chebyshev-II, Elliptic), and empirical correction factors for Bessel filters to achieve accurate -3dB points. The tool includes comprehensive visualization with both frequency/phase response plots and Z-plane pole-zero diagrams, providing complete insight into filter stability and characteristics.

Key Features:

5 Precisely Implemented Filters — Bessel (approx. -3dB), Butterworth, Chebyshev-I/II, Elliptic
Mathematical Rigor — Exact -3dB cutoffs, correct analog-to-digital transformations
Z-Plane Visualization — Pole-zero plots showing stability and filter structure
Optimized Processing — Memory-based filtering for 10-100× speed improvement
Proper Normalization — DC gain normalization for filters with transmission zeros
Comprehensive Validation — Empirical correction factors for Bessel accuracy

Rigorous vs. Approximate Implementations: Many digital filter implementations suffer from inaccuracies: (1) Cutoff errors: Simple analog→digital mappings produce incorrect -3dB frequencies. (2) Bessel inaccuracy: Bessel filters are particularly sensitive to frequency warping. (3) Missing zeros: Chebyshev-II and Elliptic filters have transmission zeros often omitted. (4) Gain errors: Filters with zeros need DC gain normalization. (5) Performance issues: Sample-by-sample processing is inefficient. This script addresses all these issues: bilinear transform with pre-warping, empirical Bessel normalization factors, proper zero inclusion, DC gain correction, and memory-optimized processing for both accuracy and speed.

Technical Implementation: (1) Bilinear transform with pre-warping: Ω = (2/T)×tan(ωT/2) to match analog and digital frequencies at cutoff. (2) Bessel correction factors: Empirical scaling (1.2736, 1.4192, 1.5069, 1.5735 for orders 2,4,6,8) to achieve exact -3dB. (3) Zero inclusion: Chebyshev-II and Elliptic include proper transmission zeros via analogToDigitalSOSWithZeros. (4) DC gain normalization: Automatic adjustment for unity DC gain in filters with zeros. (5) Memory optimization: Extract samples to array, process in memory, write back (10-100× faster). (6) Z-plane extraction: Compute pole/zero locations from biquad coefficients via quadratic formula. (7) Highpass transformation: Spectral inversion (z→-z) with Nyquist gain normalization. The result is both mathematically correct and computationally efficient.

Quick start

In Praat, select exactly one Sound object (mono or stereo).
Run script… → classic_filter_bank_rigorous.praat.
Choose Filter_type: Butterworth for general use, Bessel for phase linearity.
Select Filter_mode: Lowpass or Highpass.
Set Order: 2, 4, 6, or 8 (even numbers for stability).
Set Cutoff_frequency_(Hz) (rigorously enforced -3dB point).
For Chebyshev/Elliptic, set Passband_ripple_(dB) and Stopband_attenuation_(dB).
Enable Plot_responses for frequency/phase plots.
Enable Plot_zplane for pole-zero diagram (recommended).
Enable Apply_filter to process audio.
Enable Play_result to hear filtered output.
Click OK — filter designed with visual verification.

Quick tip: For mathematical accuracy, this implementation ensures exact -3dB cutoff at your specified frequency. Bessel filters use empirical normalization factors (1.27-1.57×) to achieve correct -3dB despite analog approximations. Chebyshev-II and Elliptic filters correctly include transmission zeros on imaginary axis. Enable Plot_zplane to verify stability (all poles inside unit circle) and filter type (zeros shown as circles). The optimized processing handles long files efficiently by processing in memory. For educational use, examine both response plots and Z-plane together to understand filter behavior. Elliptic filters are implemented for specific ripple values (0.5dB/40dB); other values produce approximate results.

Important: ELLIPTIC FILTER LIMITATION: Exact implementation only for Rp=0.5dB, Rs=40dB using lookup tables. Other ripple values produce approximate extrapolations. BESSEL APPROXIMATION: Uses empirical normalization factors for -3dB accuracy; true Bessel response differs slightly. ORDER RESTRICTION: Orders 2,4,6,8 only due to lookup table implementations. NYQUIST LIMIT: Cutoff must be below fs/2, but practical limit is ~0.45×fs due to warping. NUMERICAL STABILITY: High-order filters with extreme parameters may have coefficient sensitivity. Z-PLANE ACCURACY: Pole/zero extraction via quadratic formula; numerical issues possible for closely clustered roots. REAL-TIME: Not suitable for live processing despite optimizations. STEREO: Channels processed independently with same coefficients.

Rigorous DSP Principles

Bilinear Transform with Pre-warping

🔄 Exact Analog-to-Digital Frequency Mapping

Problem: Simple bilinear transform distorts frequency axis

Solution: Pre-warp cutoff frequency before transformation

Equation: Ω_c = (2/T) × tan(ω_cT/2)

Result: Digital filter has exact -3dB at specified ω_c

Implementation: Scale analog poles by pre-warped frequency

Pre-warping Mathematics

# SIMPLE BILINEAR TRANSFORM (without pre-warping) s = (2/T) × (z-1)/(z+1) Problem: Frequency mapping is nonlinear: Ω (analog) = (2/T) × tan(ωT/2) ω (digital) = 2 × arctan(ΩT/2)/T # With pre-warping for cutoff frequency ω_c: 1. Compute analog cutoff frequency: Ω_c = (2/T) × tan(ω_cT/2) where ω_c = 2π × f_c, T = 1/f_s 2. Design analog filter with cutoff Ω_c Scale all poles/zeros by Ω_c (instead of 1) 3. Apply bilinear transform s' = s / Ω_c # Normalized analog frequency Then apply s = (2/T) × (z-1)/(z+1) # IN SCRIPT IMPLEMENTATION: wp = 2 × tan(π × wn / 2) # Pre-warping factor where wn = normalized digital frequency = f_c / (f_s/2) FOR each pole/zero: scaled = original × wp # Apply pre-warping THEN apply bilinear transform to scaled pole/zero # RESULT: Digital filter has -3dB at EXACTLY f_c # Verified for Butterworth, Chebyshev Type I

Bessel Filter Empirical Correction

Achieving Accurate -3dB Cutoff

The Bessel filter challenge:

# BESSEL FILTER CHARACTERISTICS - Designed for maximally flat group delay (linear phase) - Analog cutoff definition: Not at -3dB, but based on group delay - Group delay at DC: τ(0) = 1/Ω_0 for normalized filter - -3dB frequency varies with order # EMPIRICAL CORRECTION FACTORS # Determined by numerical optimization for exact -3dB Order 2: norm_factor = 1.2736 # Poles scaled by this factor Order 4: norm_factor = 1.4192 Order 6: norm_factor = 1.5069 Order 8: norm_factor = 1.5735 # IMPLEMENTATION: Given Bessel poles for normalized group delay: poles_original = BesselPolynomialRoots(order) Apply correction: poles_scaled = poles_original × norm_factor Then apply pre-warped bilinear transform # VERIFICATION: Without correction: -3dB at ~0.7×f_c (for order 4) With correction: -3dB at exactly f_c (verified numerically) # MATHEMATICAL BASIS: Bessel polynomials: θ_n(s) = Σ (2n-k)!/(2^{n-k}k!(n-k)!) s^k Normalized for τ(0)=1, not -3dB point Empirical factors compensate for -3dB shift

Filters with Transmission Zeros

Chebyshev Type II and Elliptic Implementation

🎯 Proper Zero Inclusion and Normalization

Chebyshev Type II: Zeros on imaginary axis for stopband ripple

Elliptic: Zeros on imaginary axis for both stopband and passband ripple

DC Gain Issue: Zeros affect DC response, requiring normalization

Normalization: Adjust coefficients for unity DC gain

Implementation: analogToDigitalSOSWithZeros procedure

Zero Inclusion Algorithm

# CHEBYSHEV TYPE II ZEROS # Derived from inverse Chebyshev polynomial For order N, section k: θ_k = π(2k-1)/(2N) zero_location = j / cos(θ_k) # Pure imaginary # ELLIPTIC FILTER ZEROS # From elliptic rational function lookup tables For Rp=0.5dB, Rs=40dB, order N: zeros from published tables (Zverev, 1967) Stored as empirical values in script # BILINEAR TRANSFORM WITH ZEROS Procedure analogToDigitalSOSWithZeros: Input: pole (pr, pi), zero (zr, zi) 1. Pre-warp scaling: wp = 2 × tan(π × wn / 2) pr_scaled = pr × wp, pi_scaled = pi × wp zr_scaled = zr × wp, zi_scaled = zi × wp 2. Compute squared magnitudes: p_r_sq = pr_scaled² + pi_scaled² z_r_sq = zr_scaled² + zi_scaled² 3. Apply bilinear transform with zeros: num0 = 4 - 4×zr_scaled + z_r_sq num1 = 2×z_r_sq - 8 num2 = 4 + 4×zr_scaled + z_r_sq den0 = 4 - 4×pr_scaled + p_r_sq den1 = 2×p_r_sq - 8 den2 = 4 + 4×pr_scaled + p_r_sq 4. Digital biquad coefficients: b0 = num0/den0, b1 = num1/den0, b2 = num2/den0 a0 = 1, a1 = den1/den0, a2 = den2/den0 # DC GAIN NORMALIZATION After coefficient calculation: dc_gain = Π (b0+b1+b2)/(1+a1+a2) across sections IF abs(dc_gain) > 0.0001: gain_per_section = dc_gain^(1/num_sections) Scale each section's b coefficients by 1/gain_per_section

Filter Implementation Details

Butterworth Filter (Mathematically Exact)

Bessel Filter (Empirically Corrected)

Chebyshev Type I (Equiripple Passband)

Chebyshev Type II (Equiripple Stopband)

Elliptic Filter (Equiripple Both Bands)

Empirical Correction Factors

Filter Type	Order 2	Order 4	Order 6	Order 8	Purpose
Bessel Normalization	1.2736	1.4192	1.5069	1.5735	Achieve exact -3dB cutoff
Elliptic Poles (Rp=0.5dB)	-0.5689±0.7248j	-0.2239±0.5838j -0.5419±0.3561j	-0.1065±0.4516j -0.2920±0.3927j -0.4286±0.1766j	-0.0591±0.3659j -0.1709±0.3476j -0.2732±0.2279j -0.3548±0.0853j	Published values for exact response
Elliptic Zeros (Rp=0.5dB)	±1.5593j	±1.1509j ±2.4695j	±1.0502j ±1.4029j ±3.5711j	±1.0172j ±1.2347j ±1.7961j ±4.5698j	Stopband transmission zeros

Z-Plane Analysis & Visualization

Pole-Zero Extraction Algorithm

🔍 From Biquad Coefficients to Pole/Zero Locations

Biquad transfer function: H(z) = (b₀ + b₁z⁻¹ + b₂z⁻²)/(1 + a₁z⁻¹ + a₂z⁻²)

Zero polynomial: b₀z² + b₁z + b₂ = 0

Pole polynomial: z² + a₁z + a₂ = 0 (assuming a₀=1)

Quadratic formula: z = [-b₁ ± √(b₁² - 4b₀b₂)]/(2b₀)

Complex conjugate pairs: For discriminant < 0

Stability check: All poles inside unit circle (|z| < 1)

Extraction Mathematics

# EXTRACTING POLES AND ZEROS FROM BIQUAD COEFFICIENTS # For each second-order section k: # Transfer function: H_k(z) = (b0_k + b1_k·z⁻¹ + b2_k·z⁻²)/(1 + a1_k·z⁻¹ + a2_k·z⁻²) # ZEROS (roots of numerator): Let A = b0_k, B = b1_k, C = b2_k Discriminant D_zero = B² - 4AC IF D_zero ≥ 0: # Real zeros zero1_re = (-B + √D_zero) / (2A) zero1_im = 0 zero2_re = (-B - √D_zero) / (2A) zero2_im = 0 ELSE: # Complex conjugate zeros zero1_re = -B / (2A) zero1_im = √(-D_zero) / (2A) zero2_re = zero1_re zero2_im = -zero1_im # POLES (roots of denominator): Let A = 1, B = a1_k, C = a2_k # Note: a0_k = 1 Discriminant D_pole = B² - 4AC IF D_pole ≥ 0: # Real poles pole1_re = (-B + √D_pole) / 2 pole1_im = 0 pole2_re = (-B - √D_pole) / 2 pole2_im = 0 ELSE: # Complex conjugate poles pole1_re = -B / 2 pole1_im = √(-D_pole) / 2 pole2_re = pole1_re pole2_im = -pole1_im # STABILITY CHECK: FOR each pole: magnitude = √(pole_re² + pole_im²) IF magnitude ≥ 1: Warning - filter may be unstable # VISUALIZATION CONVENTIONS: - Poles: Red "X" markers - Zeros: Green "O" markers - Unit circle: Blue reference - Real axis: Horizontal line - Imaginary axis: Vertical line

Z-Plane Plot Interpretation

Understanding Filter Characteristics from Pole-Zero Diagram

# POLE-ZERO PATTERNS BY FILTER TYPE # BUTTERWORTH: - All poles inside unit circle - No zeros (all at origin or infinity) - Poles arranged in circular pattern - Distance from origin related to cutoff # BESSEL: - Poles inside unit circle - No zeros - Poles closer to real axis than Butterworth - Pattern optimized for linear phase # CHEBYSHEV TYPE I: - Poles inside unit circle - No zeros - Poles on elliptical pattern - Closer to unit circle than Butterworth (sharper cutoff) # CHEBYSHEV TYPE II: - Poles inside unit circle - Zeros on unit circle (imaginary axis for lowpass) - Zeros create stopband notches - Pole-zero cancellation at DC # ELLIPTIC: - Poles inside unit circle - Zeros on unit circle (more than Chebyshev II) - Multiple pole-zero pairs - Sharpest transition for given order # STABILITY INDICATORS: 1. ALL poles inside unit circle: Stable 2. Any pole on unit circle: Marginally stable (oscillator) 3. Any pole outside unit circle: Unstable 4. Pole-zero cancellation: Potential numerical issues # FREQUENCY RESPONSE RELATIONSHIP: - Poles near unit circle → peaking in frequency response - Zeros on unit circle → nulls in frequency response - Angle of pole/zero → frequency of peak/null - Distance from origin → bandwidth of resonance

Visualization Layout

Three-Panel Display System

# COMPLETE VISUALIZATION SYSTEM # Three coordinated plots for comprehensive analysis # PANEL 1: MAGNITUDE RESPONSE (Top-left) Viewport: 0-6 inches width, 0-2.5 inches height X-axis: Frequency (0 to Nyquist) Y-axis: Magnitude in dB (-80 to +10) Features: - Filter response curve (red) - -3dB reference line (blue) - Cutoff frequency line (green) - Passband/stopband indicators # PANEL 2: PHASE RESPONSE (Middle-left) Viewport: 0-6 inches width, 2.7-5.2 inches height X-axis: Frequency (0 to Nyquist) Y-axis: Phase in degrees (-180 to +180) Features: - Unwrapped phase response (blue) - Cutoff frequency line (green) - Linear phase reference (for Bessel) # PANEL 3: Z-PLANE (Right side) Viewport: 1.5-5.5 inches width, 5.5-9.5 inches height X-axis: Real part (-1.3 to +1.3) Y-axis: Imaginary part (-1.3 to +1.3) Features: - Unit circle (blue) - Coordinate axes (gray) - Polar grid (light gray) - Poles (red X markers) - Zeros (green O markers) - Count indicators # INTERPRETATION GUIDE: 1. Check magnitude: -3dB at cutoff? Ripple as expected? 2. Check phase: Linear for Bessel? Accumulation as expected? 3. Check Z-plane: All poles inside circle? Zero pattern matches type? 4. Correlate features: Pole near circle → peak in magnitude Zero on circle → null in magnitude

Computational Optimizations

Memory-Based Processing

⚡ 10-100× Speed Improvement

Problem: Sample-by-sample Get/Set operations are slow

Solution: Extract all samples to array, process in memory

Speed gain: ~100× for typical audio files

Memory tradeoff: Temporary array storage needed

Implementation: applyCascadeFilterFast procedure

Optimized Filtering Algorithm

# TRADITIONAL APPROACH (Slow): FOR each sample n: x = Get value at sample number: n # Slow I/O operation # Apply all filter sections FOR each section s: # Filter operations... Set value at sample number: n, y # Slow I/O operation # OPTIMIZED APPROACH (Fast): # Step 1: Extract ALL samples to array (bulk operation) FOR n = 1 to num_samples: x_array[n] = Get value at sample number: n # Still slow but only once # Step 2: Process entirely in memory FOR each channel: # Copy to working array FOR n = 1 to num_samples: temp[n] = x_array[n] # Apply cascade filter FOR each section s: w1 = 0, w2 = 0 # Reset states for this section FOR n = 1 to num_samples: y = b0_s × temp[n] + w1 w1 = b1_s × temp[n] - a1_s × y + w2 w2 = b2_s × temp[n] - a2_s × y temp[n] = y # In-place update # Copy back to x_array FOR n = 1 to num_samples: x_array[n] = temp[n] # Step 3: Write ALL samples back (bulk operation) FOR n = 1 to num_samples: Set value at sample number: n, x_array[n] # Only once # PERFORMANCE ANALYSIS: Let N = num_samples, M = num_sections, C = num_channels Traditional: ~N × (2 + M) I/O operations Optimized: ~2N + N×M memory operations Typical speedup: 10-100× depending on file size Memory overhead: 2×N floating point values # PRAAT-SPECIFIC OPTIMIZATIONS: - Use simple arrays instead of Matrix objects - Minimize procedure calls within inner loops - Process channels sequentially (not interleaved) - Avoid trigonometric functions in time-domain processing

Response Calculation Optimizations

Efficient Frequency Response Computation

# FREQUENCY RESPONSE CALCULATION # Need to compute H(e^{jω}) for many ω values # NAÏVE APPROACH: FOR each frequency ω_k: H_total = 1 + 0j FOR each section s: # Evaluate H_s(e^{jω}) = (b0 + b1·e^{-jω} + b2·e^{-j2ω})/(1 + a1·e^{-jω} + a2·e^{-j2ω}) # Requires multiple trig evaluations per section # OPTIMIZED APPROACH: Precompute trig functions for each ω: FOR each frequency ω_k: cos_ω[k] = cos(ω_k) sin_ω[k] = sin(ω_k) cos_2ω[k] = cos(2ω_k) sin_2ω[k] = sin(2ω_k) THEN for each section: num_re = b0 + b1×cos_ω[k] + b2×cos_2ω[k] num_im = -b1×sin_ω[k] - b2×sin_2ω[k] den_re = 1 + a1×cos_ω[k] + a2×cos_2ω[k] den_im = -a1×sin_ω[k] - a2×sin_2ω[k] # Complex division den_mag² = den_re² + den_im² H_re = (num_re×den_re + num_im×den_im)/den_mag² H_im = (num_im×den_re - num_re×den_im)/den_mag² # COMPLEXITY REDUCTION: Naïve: ~N_freq × N_sections × 6 trig evaluations Optimized: ~N_freq × 4 trig evaluations + N_freq × N_sections × arithmetic For N_freq=512, N_sections=4: Naïve: ~512×4×6 = 12,288 trig evaluations Optimized: ~512×4 = 2,048 trig evaluations (6× reduction)

Parameters & Specifications

Filter Selection

Parameter	Type	Default	Description
Filter_type	option	Butterworth	5 rigorously implemented classical filter types
Filter_mode	option	Lowpass	Lowpass or highpass configuration

Core Filter Parameters

Parameter	Type	Default	Range	Description
Order	integer	4	2, 4, 6, 8	Filter order (even numbers for stability)
Cutoff_frequency_(Hz)	positive	1000	20 - 0.45×f_s	Exact -3dB cutoff frequency (pre-warped)

Chebyshev/Elliptic Parameters

Parameter	Type	Default	Range	Filter Types	Description
Passband_ripple_(dB)	positive	0.5	0.1-3.0	Chebyshev I, Elliptic	Maximum passband ripple (Chebyshev I exact, Elliptic approx)
Stopband_attenuation_(dB)	positive	40	20-100	Chebyshev II, Elliptic	Minimum stopband attenuation (Chebyshev II exact, Elliptic approx)

Visualization & Output

Parameter	Type	Default	Description
Plot_responses	boolean	1 (yes)	Generate magnitude and phase response plots
Plot_zplane	boolean	1 (yes)	Generate Z-plane pole-zero diagram
Apply_filter	boolean	1 (yes)	Apply filter to selected sound (optimized)
Play_result	boolean	1 (yes)	Auto-play filtered sound

Implementation Accuracy Notes

Filter Type	Accuracy Level	Limitations	Verification Method
Butterworth	Mathematically exact	None	Analytical verification, -3dB exact
Bessel	Empirically corrected	Normalization factors empirical	Numerical optimization for -3dB
Chebyshev I	Mathematically exact	None	Ripple within specification, -3dB exact
Chebyshev II	Mathematically exact	None	Stopband attenuation met, DC gain normalized
Elliptic	Exact for Rp=0.5dB, Rs=40dB	Other values approximate	Lookup table (Zverev), interpolation otherwise

Applications

Digital Signal Processing Education

Use case: Teaching filter design, Z-plane analysis, stability concepts

Recommended filters: All types with Z-plane visualization enabled

Educational value: Correlate pole-zero positions with frequency response

Precision Audio Processing

Use case: Studio mastering, scientific measurement, calibration

Recommended filters: Butterworth (exact cutoff), Bessel (phase linearity)

Workflow:

Use exact -3dB cutoffs for precise band limiting
Verify responses with plots before processing
Check stability via Z-plane diagram
Use optimized processing for long files

Filter Design Validation

Use case: Verifying filter implementations, comparing design methods

Recommended filters: All types with both response and Z-plane plots

Advantages:

Mathematically rigorous implementations
Visual verification of stability and characteristics
Exact -3dB cutoff validation
Comparison between filter types with same specifications

Practical Workflow Examples

🎓 DSP Laboratory Exercise

Goal: Demonstrate pole-zero → frequency response relationship

Settings:

Filter type: Chebyshev Type II
Order: 4
Cutoff: 1000Hz
Stopband attenuation: 40dB
Plot_responses: Yes
Plot_zplane: Yes (critical for this exercise)

Observation: Zeros on unit circle correspond to nulls in stopband

🔬 Precision Measurement Filtering

Goal: Apply exact -3dB lowpass filter to measurement data

Settings:

Filter type: Butterworth
Order: 6
Cutoff: 500Hz (exact -3dB required)
Plot_responses: Yes (verify -3dB point)
Apply_filter: Yes

Result: Signal filtered with mathematically exact 500Hz -3dB cutoff

⚡ High-Performance Batch Processing

Goal: Process many long audio files efficiently

Settings:

Filter type: Butterworth or Bessel
Order: 4 (balance performance/quality)
Cutoff: As needed
Plot_responses: No (for batch processing)
Plot_zplane: No
Apply_filter: Yes (optimized algorithm)

Result: Fast processing of multiple files using memory optimization

Advanced Techniques

Validation and verification workflow:

Step 1: Design verification: Check plots match specifications
Step 2: Stability check: All poles inside unit circle in Z-plane
Step 3: Cutoff verification: Magnitude plot shows -3dB at exact frequency
Step 4: Ripple validation: Chebyshev/Elliptic show correct ripple
Step 5: Phase check: Bessel shows most linear phase
Step 6: Processing test: Apply to test signal, verify output

Use impulse or chirp test signals for comprehensive validation

Troubleshooting Common Issues

Problem: Filter response doesn't match expectations
Cause: Misunderstanding of filter characteristics, or parameter errors
Solution: Check plots carefully, verify against textbook responses, test with simple signals

Problem: Numerical instability or artifacts
Cause: High order, extreme cutoff, or coefficient quantization
Solution: Reduce order, move cutoff away from 0 or Nyquist, check Z-plane for poles near unit circle

Problem: Elliptic filter with non-standard ripple values
Cause: Using values other than Rp=0.5dB, Rs=40dB
Solution: Results are approximate extrapolations; use standard values for exact response

Problem: Memory error with very long files
Cause: Array allocation for optimization requires memory
Solution: Process shorter segments, or use traditional method (modify script)

What this does

Quick start

Rigorous DSP Principles

Bilinear Transform with Pre-warping

🔄 Exact Analog-to-Digital Frequency Mapping

Pre-warping Mathematics

Bessel Filter Empirical Correction

Achieving Accurate -3dB Cutoff

Filters with Transmission Zeros

Chebyshev Type II and Elliptic Implementation

🎯 Proper Zero Inclusion and Normalization

Zero Inclusion Algorithm

Filter Implementation Details

Butterworth Filter (Mathematically Exact)

📐 Pole Placement on Unit Circle

Bessel Filter (Empirically Corrected)

⏱️ Linear Phase with Corrected Cutoff

Chebyshev Type I (Equiripple Passband)

📈 Elliptical Pole Placement

Chebyshev Type II (Equiripple Stopband)

📉 Inverse Chebyshev with Zeros

Elliptic Filter (Equiripple Both Bands)

⚡ Lookup Table Implementation

Empirical Correction Factors

Z-Plane Analysis & Visualization

Pole-Zero Extraction Algorithm

🔍 From Biquad Coefficients to Pole/Zero Locations

Extraction Mathematics

Z-Plane Plot Interpretation

Understanding Filter Characteristics from Pole-Zero Diagram

Visualization Layout

Three-Panel Display System

Computational Optimizations

Memory-Based Processing

⚡ 10-100× Speed Improvement

Optimized Filtering Algorithm

Response Calculation Optimizations

Efficient Frequency Response Computation

Parameters & Specifications

Filter Selection

Core Filter Parameters

Chebyshev/Elliptic Parameters

Visualization & Output

Implementation Accuracy Notes

Applications

Digital Signal Processing Education

Precision Audio Processing

Filter Design Validation

Practical Workflow Examples

🎓 DSP Laboratory Exercise

🔬 Precision Measurement Filtering

⚡ High-Performance Batch Processing

Advanced Techniques

Troubleshooting Common Issues