Frequency Response and Bandwidth

Introduction and Context

Electronic amplifiers are often introduced through a midband voltage gain, but practical circuits do not amplify all frequencies equally. Coupling capacitors, bypass capacitors, transistor junction capacitances, wiring capacitances, device transit times, and feedback networks cause the gain and phase to vary with frequency. The study of this variation is the frequency response of a circuit.

Frequency-response analysis is central to amplifier design because it determines signal fidelity, usable bandwidth, stability, noise performance, and dynamic response. A voltage amplifier may have adequate gain at 1 kHz but severe attenuation or phase shift at 10 Hz or 10 MHz. In communication systems, frequency response also determines whether a modulated signal passes without unacceptable distortion.

Learning Objectives

After completing this chapter, the reader should be able to:

1. Define frequency response, bandwidth, cutoff frequency, gain, and phase response.
2. Explain the physical origins of low-, mid-, and high-frequency limitations.
3. Derive and apply first-order high-pass and low-pass response equations.
4. Construct approximate Bode magnitude and phase plots.
5. Determine bandwidth, gain-bandwidth product, and rise time under stated assumptions.
6. Analyse multistage amplifiers and identify dominant poles.
7. Evaluate frequency-response measurements and practical design trade-offs.
8. Apply these concepts to an amplifier design problem.

1. Fundamental Description of Frequency Response

For a linear time-invariant circuit, the sinusoidal steady-state response at angular frequency ω is represented by the transfer function

where V_i and V_o are phasor input and output voltages, j = √−1, and ω = 2πf is angular frequency in rad/s. The transfer function is generally complex and may be written as

Here |H| is the magnitude gain and φ is the phase shift. For a voltage amplifier, the magnitude is often expressed in decibels:

The factor 20 applies to a voltage or current ratio. For a power ratio, the corresponding expression is P_dB = 10 log₁₀(P_o/P_i).

A frequency-response curve normally includes:

• Magnitude response: gain as a function of frequency.
• Phase response: phase shift as a function of frequency.
• Low-frequency region: gain rises toward the midband value.
• Midband region: gain is approximately constant.
• High-frequency region: gain falls because of parasitic and device capacitances.

The bandwidth of a band-limited amplifier is the frequency interval over which the gain remains above its specified lower limit. For the conventional voltage-amplifier definition, the limits are the lower and upper −3 dB frequencies, f_L and f_H:

The −3 dB condition means

where A_M is the midband gain and f_c is a cutoff frequency. Since power is proportional to voltage magnitude squared for a fixed resistance, the −3 dB point corresponds to half power.

2. First-Order Networks as Frequency-Response Models

Most practical frequency responses can be understood initially as combinations of first-order poles and zeros. A pole introduces a frequency-dependent denominator; a zero introduces a frequency-dependent numerator. The pole frequency is a characteristic frequency at which the asymptotic magnitude slope changes by −20 dB per decade for a single pole. A single zero changes the slope by +20 dB per decade.

2.1 High-pass response

A series capacitor followed by a resistive load forms a high-pass network. With source resistance and load resistance combined into an equivalent resistance R, the transfer function across R is

The cutoff angular frequency and cutoff frequency are

The magnitude and phase are

These expressions assume an ideal source, a linear resistor-capacitor network, and a load represented by R. At frequencies much lower than f_L, the capacitor impedance is large and the output is attenuated. At frequencies much higher than f_L, the capacitor approximates a short circuit and the gain approaches unity.

2.2 Low-pass response

A resistor followed by a shunt capacitor produces a low-pass network:

with

The magnitude and phase are

The low-pass response is approximately constant below f_H and decreases at −20 dB per decade above f_H. At f_H, the phase is −45° and the magnitude is 0.707 of its low-frequency value.

2.3 Bode-plot interpretation

A Bode magnitude plot uses a logarithmic frequency axis and expresses gain in decibels. For a first-order low-pass system, the asymptotic plot is horizontal below the pole and has slope −20 dB/decade above it. The exact curve lies 3 dB below the low-frequency asymptote at the pole.

Figure: Approximate Bode magnitude response of a first-order low-pass network.

The phase transition extends over approximately two decades: from about +45° below the pole to −45° above it for a low-pass response, with −90° as the high-frequency limit. Exact phase plots should be calculated when phase margin or waveform fidelity is important.

3. Frequency Limits in Transistor Amplifiers

3.1 Low-frequency effects

Low-frequency roll-off in a transistor amplifier is commonly caused by:

• input coupling capacitors;
• output coupling capacitors;
• emitter or source bypass capacitors;
• transformer magnetizing inductance;
• bias networks interacting with coupling components.

Each capacitor forms a high-pass section with the resistance seen by that capacitor. For a capacitor C_i, the approximate corner frequency is

where R_i is the Thévenin resistance seen by the capacitor when independent sources are deactivated. In a multisection amplifier, the overall lower cutoff is not always exactly the largest individual f_i, because the sections interact. However, if the sections are sufficiently separated or weakly interacting, the largest corner frequency often provides a useful first estimate.

For cascaded independent high-pass sections,

At a frequency near one section's cutoff, other sections may already contribute attenuation. Consequently, simply selecting the largest individual cutoff can underestimate the true system lower cutoff.

3.2 Midband operation

In the midband region, coupling and bypass capacitors are treated as shorts, while transistor internal capacitances are treated as open circuits. The resulting gain is the midband gain, A_M. This approximation is useful only when the operating frequency is sufficiently above all low-frequency corners and below all high-frequency corners.

3.3 High-frequency effects

At high frequencies, transistor junction capacitances, package capacitances, wiring capacitances, and the input capacitance of following stages become important. In a common-emitter or common-source stage, the input capacitance is increased by the Miller effect. If C_μ is the feedback capacitance between input and output, then, for an inverting gain A_v,

Thus, if A_v is negative, the effective input capacitance can be much larger than C_μ. Including the transistor's intrinsic input capacitance C_π and stray capacitance C_s gives approximately

This approximation assumes a sufficiently low frequency that a lumped capacitance model is valid and that A_v is treated as approximately constant over the region being estimated. At very high frequencies, the frequency dependence of A_v must be included in a full hybrid-π or small-signal model.

The high-frequency pole is often estimated by

where R_seen is the resistance seen by the relevant capacitance and C_total is the effective capacitance. Reducing source resistance, minimizing stray capacitance, reducing unnecessary gain per stage, or using cascode and neutralization techniques can increase the upper cutoff frequency.

4. Multistage Amplifiers and Overall Bandwidth

For cascaded stages,

In decibels, multiplication becomes addition:

The overall bandwidth is generally narrower than the bandwidth of an individual stage. For n identical, non-interacting, single-pole low-pass stages with individual upper cutoff f_Hs, the normalized magnitude is

Setting this equal to 1/√2 gives the overall upper −3 dB frequency:

For n = 2, f_H,total ≈ 0.644 f_Hs. This result assumes identical poles and no loading interaction. Real multistage amplifiers may require pole-zero analysis using the complete small-signal circuit.

A useful design strategy is pole separation: the dominant pole is deliberately placed much lower than the remaining poles. The response then resembles a single-pole system over the principal operating range, simplifying compensation and improving predictability.

5. Gain–Bandwidth Product and Operational Amplifiers

For a voltage-feedback operational amplifier modeled as a single dominant-pole device, the open-loop gain may be approximated by

where A_0 is the low-frequency open-loop gain and f_p is the dominant pole. The unity-gain frequency or gain-bandwidth product is approximately

For a stable voltage-feedback amplifier operated with noise gain N, the closed-loop bandwidth is approximately

Noise gain is the gain experienced by the amplifier's input-referred voltage noise and is not necessarily the same as the signal gain. The approximation applies when one dominant pole controls the response and additional poles do not materially affect the closed-loop bandwidth.

A designer must distinguish signal bandwidth from noise bandwidth. A low-pass filter with transfer function H(f) has equivalent noise bandwidth

For a first-order low-pass filter with 3 dB frequency f_H, B_ENBW = (π/2)f_H. Therefore, thermal and amplifier noise calculations should not automatically use the −3 dB bandwidth as the integration bandwidth.

6. Time-Domain Consequences

Frequency response and transient response are Fourier-transform descriptions of the same linear system. A single-pole low-pass system has time constant

Its response to a step input approaches its final value according to

for t ≥ 0, where V_F is the final output value and zero initial conditions are assumed. The 10–90% rise time is approximately

This relationship applies to a single-pole or approximately Gaussian response. It should not be used indiscriminately for systems with significant overshoot, multiple closely spaced poles, or strong peaking.

Phase response is equally important. Two amplifiers can have the same magnitude bandwidth but different phase responses, producing different overshoot, ringing, or pulse distortion. In feedback systems, phase lag reduces phase margin and may cause instability.

7. Worked Example: Common-Emitter Amplifier

Consider an illustrative common-emitter amplifier with midband gain magnitude |A_M| = 100, an input coupling capacitor C_in = 1 μF, an output coupling capacitor C_out = 4.7 μF, and an emitter bypass capacitor C_E = 47 μF. Suppose the resistances seen by these capacitors are respectively 10 kΩ, 4 kΩ, and 100 Ω. Ignore interaction initially and assume a single effective high-frequency pole at 2 MHz.

The three low-frequency corners are

The emitter-bypass section has the highest individual corner, so a first estimate is f_L ≈ 34 Hz. Because the other sections also attenuate the signal at this frequency, the exact overall f_L will be somewhat higher and must be obtained by multiplying the three exact section responses or by simulation.

The estimated upper cutoff is f_H ≈ 2 MHz. Hence the approximate bandwidth is

At 1 kHz, the low-frequency sections are well above their corners and the high-frequency pole is far away, so the gain is close to 100 V/V, or

At 2 MHz, the high-frequency section alone reduces the gain to 0.707 × 100 = 70.7 V/V, or 37 dB relative to the input voltage ratio. This example demonstrates why the midband approximation is useful for design estimates but insufficient for accurate cutoff prediction when several poles interact.

8. Measurement and Design Practice

A practical frequency-response measurement uses a function generator, oscilloscope or network analyser, suitable terminations, and a controlled amplitude. The input amplitude should remain constant while frequency is swept. The measured gain is

and should be converted to decibels only after the ratio has been formed. Input and output probe capacitances must be included because they can significantly reduce f_H, especially in high-impedance nodes.

Recommended practice includes:

• establish the DC operating point before the sweep;
• use small signals so that the circuit remains linear;
• verify generator and oscilloscope bandwidth;
• account for loading and cable termination;
• measure phase as well as magnitude when feedback is present;
• repeat measurements at different signal amplitudes to detect slew-rate or clipping effects;
• compare measured poles with a small-signal model.

A frequency response that changes with signal level may indicate nonlinear distortion rather than a purely linear bandwidth limitation. Likewise, apparent peaking may be caused by probe loading, layout parasitics, supply decoupling, or an unstable feedback loop.

9. Applied Case: Broadband Sensor Interface

Illustrative case. A sensor interface must amplify signals from 20 Hz to 100 kHz. The required midband gain is 40 dB, and the output must remain within 1 dB of its nominal gain across the band.

Background and problem

A single high-gain stage would suffer from Miller multiplication and may not provide adequate upper bandwidth. Excessive low-frequency coupling-capacitor reactance could also attenuate the 20 Hz lower limit.

Analysis

A 40 dB gain corresponds to a voltage gain of 100 V/V. If two identical stages are used, each stage may provide approximately 10 V/V, or 20 dB. Splitting gain reduces the Miller capacitance associated with each stage. The coupling capacitors must be selected so that their corner frequencies are well below 20 Hz; placing each corner near 2 Hz provides margin, although the exact choice depends on the allowable cumulative attenuation.

For an operational-amplifier implementation with unity-gain frequency f_T = 10 MHz and noise gain N = 11, the approximate closed-loop bandwidth is

This comfortably exceeds 100 kHz, but stability and the amplifier's data-sheet small-signal response must still be checked. If the interface drives a capacitive cable, an isolation resistor or compensation network may be required.

Approach and solution

The design uses two moderate-gain stages, high-quality coupling capacitors, a defined input resistance, local supply bypassing, and a measured Bode response. The PCB layout keeps high-impedance input nodes short and separates sensor wiring from high-current output paths.

Outcome

The cascaded design meets the signal band with substantially lower sensitivity to Miller-induced bandwidth loss than one high-gain stage. The final validation includes gain flatness, phase margin, noise density, overload recovery, and response under production tolerances.

Learning points

Bandwidth is a system property, not merely a transistor property. Gain distribution, loading, parasitic capacitance, compensation, and measurement technique must be considered together. A nominal −3 dB bandwidth alone does not guarantee acceptable pulse fidelity, stability, or noise performance.

10. Limitations, Trade-offs, and Contemporary Developments

Increasing bandwidth often requires reducing gain per stage, lowering impedance, increasing bias current, or accepting greater power consumption. Wideband designs may have higher integrated noise and greater sensitivity to layout. Compensation that improves stability may reduce speed. Conversely, bandwidth extension techniques such as feed-forward paths, cascodes, inductive peaking, active feedback, and digital calibration can improve performance but introduce complexity.

Modern integrated circuits increasingly use simulation and automated extraction of parasitic elements. Electromagnetic modelling, statistical process analysis, and corner simulations help predict frequency response across manufacturing, voltage, and temperature variations. Machine-learning tools may assist in surrogate modelling or component optimisation, but their outputs must be verified against physical small-signal models, stability criteria, measurement, and device reliability constraints. They do not replace engineering validation.

Professional design should also respect electromagnetic-compatibility regulations, specified emissions limits, safety isolation, and truthful performance claims. A circuit should not be described as “100 kHz bandwidth” without stating the gain criterion, loading, signal level, and measurement conditions.

Best-Practice Checklist

1. Define bandwidth using an explicit criterion, normally −3 dB relative to midband gain.
2. Separate low-frequency, midband, and high-frequency analyses.
3. Identify every significant pole, zero, loading path, and parasitic capacitance.
4. Use exact transfer functions near critical corners rather than only asymptotic slopes.
5. Treat multistage bandwidth as a product of responses, not as the bandwidth of one stage.
6. Check phase margin and transient response in feedback circuits.
7. Include noise bandwidth when calculating integrated noise.
8. Validate models with controlled measurements and realistic loading.
9. Report amplitude, phase, signal level, temperature, and termination conditions.

Chapter Summary

Frequency response describes how an amplifier's magnitude and phase vary with frequency. The −3 dB frequencies define conventional lower and upper bandwidth limits, but phase response, noise bandwidth, transient response, and stability are also essential performance measures. First-order high-pass and low-pass networks provide the basic analytical models. Low-frequency limitations arise mainly from coupling and bypass capacitors, whereas high-frequency limitations arise from junction, stray, and Miller-effect capacitances. Cascaded stages have multiplicative transfer functions and generally narrower overall bandwidth. Dominant-pole compensation, gain–bandwidth analysis, Bode plots, and time-domain relationships provide practical design tools. Accurate engineering requires attention to loading, parasitics, measurement conditions, process variation, and feedback stability.

Key Terms

Frequency response: Variation of a circuit’s magnitude and phase with frequency.

Transfer function: Ratio of output phasor to input phasor as a function of complex frequency.

Magnitude response: Absolute value of the transfer function versus frequency.

Phase response: Argument or phase angle of the transfer function versus frequency.

Midband gain: Approximately constant gain in the central operating-frequency region.

Bandwidth: Frequency interval satisfying a specified gain criterion.

Cutoff frequency: Frequency at which a response reaches a defined limit, commonly −3 dB.

−3 dB point: Frequency at which voltage magnitude is 1/√2 of its reference value.

Pole: A transfer-function denominator root that contributes negative magnitude slope and phase lag.

Zero: A transfer-function numerator root that changes magnitude slope and phase.

Bode plot: Log-frequency plot of magnitude in decibels and phase.

Miller effect: Apparent multiplication of a feedback capacitance by amplifier gain.

Dominant pole: Pole intentionally made low enough to control system response.

Gain–bandwidth product: Approximate product of closed-loop gain and bandwidth in a single-pole amplifier.

Noise bandwidth: Equivalent rectangular bandwidth that produces the same integrated noise power.

Phase margin: Stability measure based on the remaining phase before the loop reaches −180° at unity gain.

Parasitic capacitance: Unintended capacitance from devices, packages, wiring, or layout.

Rise time: Time required for a signal to move between specified percentages, often 10% and 90%.

Review Questions

Short Answer Questions

1. What is meant by the frequency response of an amplifier?
2. Why is the −3 dB frequency important in bandwidth specification?
3. State two physical causes of low-frequency roll-off.
4. Explain the Miller effect in one or two sentences.
5. What is the difference between magnitude response and phase response?

Long Answer Questions

1. Derive the magnitude and phase response of a first-order RC low-pass network and explain its Bode plot.
2. Discuss the origins of low- and high-frequency limitations in a common-emitter amplifier.
3. Explain why cascaded amplifier stages generally reduce the overall bandwidth.
4. Derive the approximate gain–bandwidth relationship for a dominant-pole operational amplifier.
5. Discuss the relationship between frequency response, rise time, overshoot, and phase margin.

Analytical / Application Questions

1. A high-pass coupling network has R = 22 kΩ and C = 0.47 μF. Calculate its cutoff frequency and determine whether a 10 Hz signal is substantially attenuated.
2. Two identical first-order amplifier stages each have an upper cutoff frequency of 1 MHz. Calculate the approximate overall upper −3 dB frequency using the cascaded-stage expression and explain the assumption involved.
3. Design a conceptual sensor amplifier for a 20 Hz–100 kHz signal band. Specify how you would distribute gain, select coupling-corner frequencies, assess high-frequency loading, and validate stability experimentally.

Multiple Choice Questions with Answers

Question 1

At the −3 dB frequency of a voltage amplifier, the voltage magnitude is approximately what fraction of its reference midband value?

A. 0.5

B. 0.707

C. 1.414

D. 2.0

Correct Answer: B. 0.707

Question 2

For a first-order RC low-pass network, the cutoff frequency is:

A. f = 2πRC

B. f = RC/(2π)

C. f = 1/(2πRC)

D. f = 1/(RC²)

Correct Answer: C. f = 1/(2πRC)

Question 3

A single first-order pole contributes what asymptotic magnitude slope above its corner frequency?

A. +20 dB per decade

B. −3 dB per decade

C. −10 dB per decade

D. −20 dB per decade

Correct Answer: D. −20 dB per decade

Question 4

Which component most directly causes low-frequency roll-off in a transistor amplifier?

A. Input coupling capacitor

B. Collector supply voltage alone

C. Metal film resistor tolerance alone

D. DC bias current source at zero frequency

Correct Answer: A. Input coupling capacitor

Question 5

For an inverting amplifier, the approximate Miller contribution of a feedback capacitance Cμ to input capacitance is:

A. C_μ(1 + A_v)

B. C_μ(1 − A_v)

C. C_μ/A_v

D. A_v/C_μ

Correct Answer: B. C_μ(1 − A_v)

Question 6

If an amplifier has a midband voltage gain of 100 V/V, its gain in decibels is:

A. 10 dB

B. 20 dB

C. 40 dB

D. 100 dB

Correct Answer: C. 40 dB

Question 7

For a dominant-pole operational amplifier with unity-gain frequency 10 MHz and noise gain 11, the approximate closed-loop bandwidth is:

A. About 91 kHz

B. About 0.91 MHz

C. About 11 MHz

D. About 110 MHz

Correct Answer: B. About 0.91 MHz

Question 8

For two identical, non-interacting, single-pole stages, the overall upper −3 dB frequency is approximately:

A. Higher than the individual stage cutoff by √2

B. Equal to twice the individual cutoff

C. About 0.644 times the individual stage cutoff

D. Equal to the individual stage cutoff in all cases

Correct Answer: C. About 0.644 times the individual stage cutoff

Question 9

The approximate 10–90% rise time of a single-pole low-pass system is:

A. t_r ≈ 0.35/f_H

B. t_r ≈ 3.5f_H

C. t_r ≈ f_H/0.35

D. t_r ≈ 2πf_H

Correct Answer: A. t_r ≈ 0.35/f_H

Question 10

Why should noise bandwidth not automatically be set equal to the −3 dB bandwidth?

A. Noise is independent of frequency in every circuit

B. Integrated noise depends on the squared magnitude response over frequency

C. The −3 dB point is defined only for current amplifiers

D. Noise bandwidth is always exactly twice the −3 dB bandwidth

Correct Answer: B. Integrated noise depends on the squared magnitude response over frequency