How is passband bandwidth defined?

Passband bandwidth can be defined several ways. The minus 3 dB bandwidth uses the frequencies where the insertion loss is 3 dB below the minimum (half the power). This is the most common definition for general-purpose filters and amplifiers. The minus 1 dB bandwidth uses a tighter criterion and results in a narrower bandwidth. For cellular and wireless standards, the passband is defined by specification limits: for example, a 5G NR Band n77 filter must pass 3.3 to 4.2 GHz with less than 2 dB insertion loss and greater than 15 dB return loss. The ripple bandwidth defines the range over which insertion loss stays within the ripple specification (0.1 dB for a 0.1 dB Chebyshev filter). Fractional bandwidth is BW/f_center times 100 percent, which normalizes bandwidth to center frequency: a 100 MHz bandwidth at 2 GHz is 5 percent fractional bandwidth, while 100 MHz at 28 GHz is only 0.36 percent.

What is passband ripple and why does it matter?

Passband ripple is the peak-to-peak variation in insertion loss within the passband. Butterworth (maximally flat) filters have zero ripple by design, but achieve this with a gradual transition to the stopband. Chebyshev filters allow controlled ripple in exchange for steeper transitions: a 0.5 dB ripple Chebyshev has sharper rejection than a Butterworth of the same order. Elliptic (Cauer) filters have ripple in both passband and stopband, achieving the sharpest possible transition for a given order. Ripple matters because it causes gain flatness variation across the signal bandwidth, directly impacting EVM for wideband signals. For a 5G NR 100 MHz signal passing through a filter with 0.5 dB ripple, the EVM contribution is approximately 1 percent, which consumes a significant portion of the total EVM budget for 256-QAM.

How does passband relate to stopband and transition band?

A filter's frequency response has three regions: passband (signals pass with minimal loss), stopband (signals are attenuated by the required rejection, typically 20 to 80 dB), and transition band (the region between passband and stopband where attenuation increases from the passband insertion loss to the stopband rejection). The shape factor measures the transition steepness: SF = BW_stopband / BW_passband, where BW_stopband is measured at a specified rejection level (typically 30 or 60 dB). An ideal filter has SF = 1 (brick-wall response). Practical filters have SF of 1.5 to 5 depending on the technology and order. A 5th-order Chebyshev has SF approximately equal to 2 at 30 dB, meaning the 30 dB bandwidth is twice the 3 dB bandwidth. Higher order filters have sharper transitions (SF closer to 1) but increased insertion loss, group delay variation, and physical size.

Filter Characteristics

Passband

/pas-band/

The frequency range a filter passes with minimal attenuation, defined by −3 dB bandwidth (half-power points) or specification limits. Key metrics: insertion loss (0.1-3 dB), return loss (>15-20 dB), ripple (0 dB Butterworth, 0.01-0.5 dB Chebyshev). Shape factor = BW_stopband/BW_passband measures transition steepness (ideal: 1, practical: 1.5-5). Fractional BW = BW/f_center × 100%. Ripple directly impacts EVM (~2.2%/dB).

Definition: −3 dB points

Ripple: 0-0.5 dB

Shape: SF 1.5-5

Understanding the Passband

The passband defines the useful operating range of a filter. Within this frequency range, the filter should ideally be transparent: zero insertion loss, perfect impedance match, flat amplitude, and constant group delay. In practice, every filter departs from this ideal, and the quality of these departures defines the filter's impact on the system. A filter with low insertion loss but high ripple may be worse for a wideband signal than one with higher average loss but flatter response.

The choice of filter approximation (Butterworth, Chebyshev, elliptic, Bessel) fundamentally determines the passband characteristics. Butterworth provides maximum flatness at the expense of gradual transition. Chebyshev trades flatness for transition steepness. Elliptic achieves the sharpest transition but with ripple in both passband and stopband. Bessel sacrifices both flatness and steepness to achieve the most constant group delay (best phase linearity). The right choice depends on the application.

Passband Equations

Bandwidth definitions:
BW_−3dB = f_upper − f_lower (at −3 dB)
Fractional BW = BW/f_center × 100%
f_center = √(f_upper × f_lower)

Shape factor:
SF = BW_xdB / BW_3dB
x = 30 or 60 dB rejection level
5th-order Chebyshev: SF ≈ 2 @ 30 dB

Q and bandwidth relationship:
Q_loaded = f_center / BW_−3dB
Higher Q = narrower passband
Q = 100 at 2 GHz: BW = 20 MHz

Insertion loss vs. Q and order:
IL ≈ 4.343 × n / (Q_u × FBW)
n = filter order, Q_u = unloaded Q

Filter Approximation Comparison

Approximation	Passband Ripple	Shape Factor	Group Delay	Best For
Butterworth	0 dB (flat)	Moderate	Moderate GDV	General purpose
Chebyshev I	0.01-0.5 dB	Sharp	High GDV at edges	Selectivity needed
Elliptic	0.01-0.5 dB	Sharpest	Very high GDV	Max rejection
Bessel	Non-flat	Gradual	Flat (constant)	Phase-sensitive
Gaussian	Rounded	Very gradual	Very flat	Pulse preservation

Common Questions

Frequently Asked Questions

How is bandwidth defined?

-3 dB BW: frequencies where IL = 3 dB below minimum. -1 dB BW: tighter, narrower. Specification BW: defined by standard (e.g., n77: 3.3-4.2 GHz, IL<2 dB). Ripple BW: range within ripple spec. Fractional BW = BW/f_center × 100%. 100 MHz at 2 GHz = 5%. 100 MHz at 28 GHz = 0.36%. Narrower FBW = higher Q required = more challenging.

Why does ripple matter?

Passband ripple = gain flatness variation across signal BW. Butterworth: 0 dB (flat) but gradual rolloff. Chebyshev: 0.01-0.5 dB ripple for sharper transition. 0.5 dB ripple ≈ 1% EVM contribution for wideband OFDM. 256-QAM EVM budget: 3.5% total. Filter ripple can consume significant portion. Elliptic: ripple in both passband and stopband, sharpest transition.

Passband vs. stopband vs. transition?

Passband: signals pass with minimal loss. Stopband: signals rejected (20-80 dB attenuation). Transition band: IL increases from passband to stopband. Shape factor = BW_stop/BW_pass. Ideal SF=1 (brick wall). 5th-order Chebyshev: SF≈2 at 30 dB. Higher order: sharper transition (SF→1) but more IL, GDV, and size.

Related Terms

← Output Power Patch Antenna →