Butterworth vs Chebyshev vs Elliptic Filters: Key Differences & Uses

In signal processing, Butterworth filters have a smooth frequency response with no ripple, Chebyshev filters allow ripple in the passband or stopband for a steeper roll-off, and Elliptic filters have ripple in both bands but provide the sharpest transition. Each filter balances trade-offs between flatness, sharpness, and complexity.

Quick Comparison

Here is a quick table comparing the main features of Butterworth, Chebyshev, and Elliptic filters.

Feature	Butterworth	Chebyshev	Elliptic
Frequency Response	Maximally flat (no ripple)	Ripple in passband (Type I) or stopband (Type II)	Ripple in both passband and stopband
Roll-off Steepness	Moderate	Steeper than Butterworth	Steepest
Complexity	Lowest	Moderate	Highest
Phase Response	Smoothest	Less smooth	Least smooth
Use Case	When smooth response is needed	When sharper cutoff is needed with some ripple	When sharpest cutoff is critical
Design Parameters	Cutoff frequency only	Cutoff frequency + ripple	Cutoff frequency + ripple in both bands

Key Differences

Butterworth filters are designed to have a flat frequency response in the passband, meaning no ripples. This makes them ideal when you want a smooth and natural signal without distortion in the passband. However, their transition from passband to stopband is gradual, so they have a moderate roll-off.

Chebyshev filters allow ripple either in the passband (Type I) or stopband (Type II). This ripple lets the filter achieve a steeper roll-off than Butterworth filters, meaning it can better separate frequencies near the cutoff. The trade-off is some variation in signal amplitude within the ripple region and a less smooth phase response.

Elliptic filters, also called Cauer filters, have ripple in both the passband and stopband. This design achieves the steepest roll-off for a given filter order, making them very efficient at separating frequencies. The downside is the most complex design and the least smooth phase response, which can cause signal distortion in some applications.

Code Comparison

Here is how to design a Butterworth lowpass filter using Python's SciPy library.

python

from scipy.signal import butter, freqz
import matplotlib.pyplot as plt

# Butterworth filter design
order = 4
cutoff = 0.3  # normalized frequency (0 to 1)
b, a = butter(order, cutoff, btype='low', analog=False)

# Frequency response
w, h = freqz(b, a)

plt.plot(w / 3.14159, abs(h))
plt.title('Butterworth Filter Frequency Response')
plt.xlabel('Normalized Frequency')
plt.ylabel('Gain')
plt.grid(True)
plt.show()

Output

A plot showing a smooth, flat passband gain that gradually rolls off after the cutoff frequency.

Chebyshev Equivalent

Here is how to design a Chebyshev Type I lowpass filter with 1 dB ripple in the passband using Python's SciPy.

python

from scipy.signal import cheby1, freqz
import matplotlib.pyplot as plt

# Chebyshev Type I filter design
order = 4
ripple = 1  # dB ripple in passband
cutoff = 0.3
b, a = cheby1(order, ripple, cutoff, btype='low', analog=False)

# Frequency response
w, h = freqz(b, a)

plt.plot(w / 3.14159, abs(h))
plt.title('Chebyshev Type I Filter Frequency Response')
plt.xlabel('Normalized Frequency')
plt.ylabel('Gain')
plt.grid(True)
plt.show()

Output

A plot showing ripple in the passband and a steeper roll-off after the cutoff frequency compared to Butterworth.

When to Use Which

Choose Butterworth when you need a smooth, ripple-free passband and can accept a slower transition from passband to stopband. It is best for audio and applications where signal distortion must be minimal.

Choose Chebyshev when you want a sharper cutoff than Butterworth and can tolerate some ripple in either the passband or stopband. It suits applications where sharper frequency separation is needed but some ripple is acceptable.

Choose Elliptic when the sharpest cutoff is critical and ripple in both bands is acceptable. It is ideal for systems with strict frequency separation requirements and where filter order (complexity) must be minimized.

Key Takeaways

Butterworth filters have no ripple and smooth phase but moderate roll-off.

Chebyshev filters trade ripple in one band for a steeper roll-off.

Elliptic filters offer the steepest roll-off with ripple in both bands.

Use Butterworth for smooth signals, Chebyshev for sharper cutoff with some ripple, Elliptic for maximum sharpness.

Filter choice depends on the balance between ripple tolerance, roll-off steepness, and phase smoothness.