Quiz 7 Study Guide: Active Filters

EEL 3123C Linear Circuits II · Quiz 7 (labeled "Homework 6" in the PDF) · Textbook: Nilsson & Riedel, Chapter 15

This guide assumes you know nothing about filters yet. Read it top to bottom once. Section 0 is a map that names things like $H(s)$, $K$, $\beta$ and $Q$; don't worry about them yet, because every symbol is defined in §1 and explained in the sections after it. Yellow boxes are the equations and ideas you must memorize. Blue boxes explain why. Red boxes are traps.

0. What Quiz 7 asks you to do

Every problem on Quiz 7 is one of four jobs. Learn these four jobs and you can do the whole quiz:

  1. Derive $H(s)$: find the circuit's transfer function, usually with $H(s)=-Z_f/Z_i$ or with KCL.
  2. Identify the filter type (low-pass, high-pass, band-pass, band-reject) by checking what happens at very low and very high frequency.
  3. Read off the parameters (gain $K$, cutoff $\omega_c$, center $\omega_o$, bandwidth $\beta$, quality $Q$) by matching $H(s)$ to a standard form.
  4. Design: pick resistor and capacitor values that give the required numbers.
ProblemTopicJobsGuide section
1Narrowband band-pass (op amp, two C's)Derive, identify, design§12, solution
2First-order high-pass design (15.10)Design§8, solution
3Band-pass from one op amp (15.32)Derive, read off§6, solution
4Identify filter type from a pictureIdentify§7, solution
5First-order low-pass design, gain in dB (15.3)Design§8, solution
6Twin-T notch (band-reject)Derive, identify, design§12, solution
7Two-buffer second-order low-pass (P15.55)Derive, identifysolution
8Parallel band-reject from a Bode plot (Ex. 15.6)Read a plot, design§10, solution
9R‖C in and R‖C feedback (P15.6)Derive, limitssolution
10Cascade band-pass design (15.30)Design§10, solution
113rd-order Butterworth low-pass designDesign§11, solution
How this connects to your earlier homework

Contents

  1. Vocabulary: every symbol
  2. Impedance and the capacitor trick
  3. Transfer function, magnitude and dB
  4. The four filter types
  5. The op amp toolkit
  6. Standard forms: reading $K,\omega_o,\beta,Q$
  7. The $s\to0$ / $s\to\infty$ test
  8. First-order active LP and HP
  9. Scaling
  10. Broadband BP (cascade) and BR (parallel)
  11. Higher order and Butterworth
  12. Narrowband BP and twin-T notch
  13. Quiz 7 solved step by step
  14. One-page formula sheet and checklist

1. Vocabulary: every symbol you will see

SymbolNameMeaning in plain wordsUnits
$v_i(t)$, $V_i(s)$input voltageThe signal you feed in. Lowercase $v(t)$ is in time; uppercase $V(s)$ is the same signal in the Laplace ($s$) domain.V
$v_o(t)$, $V_o(s)$output voltageWhat comes out of the filter.V
$f$frequencyCycles per second.Hz
$\omega$angular frequency ("omega")The same frequency in radians per second. $\omega = 2\pi f$.rad/s
$s$Laplace variableFor steady sine waves, replace $s$ with $j\omega$.1/s
$j$imaginary unit$j=\sqrt{-1}$ (engineers use $j$ because $i$ means current). So $j^2=-1$, which means $s^2=(j\omega)^2=-\omega^2$.none
DCdirect currentZero frequency: a constant signal. "DC gain" = $H(0)$, i.e. $s=0$.
corner frequencyWhere a filter's response bends; for $\frac{\cdots}{s^2+bs+c}$ it is $\sqrt c$. For a Butterworth filter it equals the cutoff.rad/s
Bode plotA graph of gain in dB (up) against frequency on a log axis (right). On a log axis, each factor of 10 gets equal space, so the gridlines between 100 and 1000 are 200, 300, …, 900 and get closer together as you go right.
$Z$impedance"Resistance" that can depend on frequency. $Z_i$ = input-side impedance, $Z_f$ = feedback impedance.Ω
$H(s)$transfer function$V_o(s)/V_i(s)$: what the circuit multiplies the input by.none (V/V)
$|H(j\omega)|$magnitude / gainHow much a sine wave of frequency $\omega$ is scaled.none, or dB
$K$ (or $k$)passband gainThe gain in the frequencies the filter lets through. A negative sign only means the output is flipped (inverted).none
$\omega_c$, $f_c$cutoff frequencyThe edge of the passband: where the gain has dropped to $K/\sqrt2 \approx 0.707K$ (−3 dB).rad/s, Hz
$\omega_{c1}$, $\omega_{c2}$lower / upper cutoffThe two edges of a band-pass or band-reject filter. $\omega_{c1}<\omega_{c2}$.rad/s
$\omega_o$center (resonant) frequencyMiddle of the band (peak of a band-pass, notch of a band-reject).rad/s
$\beta$bandwidth ("beta")Width of the band: $\beta=\omega_{c2}-\omega_{c1}$.rad/s (or Hz)
$Q$quality factorHow narrow the band is: $Q=\omega_o/\beta$. Big $Q$ = narrow, sharp.none
$n$orderHighest power of $s$ in the denominator. Sets how steeply the gain falls outside the passband.none
$k_m$, $k_f$magnitude / frequency scale factorMultipliers that convert a "prototype" design (1 Ω, 1 F, 1 rad/s) into real values.none
$R$, $C$resistor, capacitorActive filters here use only R, C and op amps (no inductors).Ω, F
LPF, HPF, BPF, BRFfilter namesLow-pass, high-pass, band-pass, band-reject (also called "notch" or "bandstop").
$$\boxed{\;\omega = 2\pi f\;}\qquad\qquad \boxed{\;s = j\omega \text{ for sine waves}\;}$$ Problems give frequencies in Hz or rad/s. Formulas with $RC$ always produce rad/s. Convert with $2\pi$ every time.
Mixing Hz and rad/s is the most common way to lose points. "Cutoff frequency of 4 kHz" means $\omega_c = 2\pi(4000) = 25{,}133$ rad/s, not 4000 rad/s.

2. Impedance and the capacitor trick

In the $s$-domain every component acts like a "resistance" called an impedance $Z$:

$$\boxed{Z_R = R}\qquad \boxed{Z_C = \frac{1}{sC}}\qquad (Z_L = sL \text{, not used in this chapter})$$

Combine impedances exactly like resistors:

$$\text{Series: } Z = Z_1+Z_2 \qquad\qquad \text{Parallel: } Z = \frac{Z_1Z_2}{Z_1+Z_2}$$ $$\boxed{R \text{ in series with } C:\; R+\frac{1}{sC} = \frac{1+sRC}{sC}}\qquad \boxed{R \parallel C:\; \frac{R\cdot\frac{1}{sC}}{R+\frac{1}{sC}} = \frac{R}{1+sRC}}$$
Where does $\frac{1}{sC}$ come from? For a sine wave of frequency $\omega$, a capacitor resists current by $\frac{1}{j\omega C}$; its size $\frac{1}{\omega C}$ shrinks as frequency rises. Writing $s$ instead of $j\omega$ is the Laplace shorthand you used in HW 2 to HW 4. So $Z_C=\frac1{sC}$ is just "the capacitor's frequency-dependent resistance."
Show the algebra for $R\parallel C$ Multiply top and bottom by $sC$: $$\frac{R\cdot\frac1{sC}}{R+\frac1{sC}}\cdot\frac{sC}{sC}=\frac{R}{sRC+1}.$$
The capacitor trick (use it on every problem):
At very low frequency ($\omega\to0$): $Z_C=\frac1{sC}\to\infty$, so a capacitor is an open circuit (a break in the wire).
At very high frequency ($\omega\to\infty$): $Z_C\to0$, so a capacitor is a short circuit (a plain wire).
A capacitor blocks slow signals and passes fast ones. Put it in series with the signal path and it blocks lows, which gives a high-pass. Put it from the signal to ground and it dumps the highs, which gives a low-pass.

3. Transfer function, magnitude and decibels

$$\boxed{H(s)=\frac{V_o(s)}{V_i(s)}}\qquad\text{(with all initial conditions zero)}$$

To find out what happens to a sine wave of frequency $\omega$, put $s=j\omega$ and take the magnitude $|H(j\omega)|$. If the input amplitude is 1 V and $|H(j\omega)|=3$, the output amplitude is 3 V.

Magnitude of a fraction = (magnitude of top) / (magnitude of bottom), and $|a+jb|=\sqrt{a^2+b^2}$.

Decibels (dB) are a log way of writing gain:

$$\boxed{\text{gain in dB} = 20\log_{10}|H|}\qquad\qquad \boxed{|H| = 10^{\,\text{dB}/20}}$$
Linear gain1$1/\sqrt2\approx0.707$233.16210100
dB0−36.029.54102040
Cutoff frequency = the −3 dB point. At $\omega_c$ the gain is $\dfrac{K}{\sqrt2}$, which is 3 dB below the passband gain. (Half the power gets through.)
Check with the first-order LPF $H=K\frac{\omega_c}{s+\omega_c}$ at $s=j\omega_c$: $\;|H|=\frac{K\omega_c}{|j\omega_c+\omega_c|}=\frac{K}{|1+j|}=\frac{K}{\sqrt{1^2+1^2}}=\frac{K}{\sqrt2}$.
"Passband gain of 10 dB" means $K = 10^{10/20}=3.162$, not $K=10$.

4. The four filter types

A filter lets some frequencies through (the passband) and blocks others (the stopband). The plots show gain (up) against frequency (right, log scale).

ωc
Low-pass (LPF): passes lows
ωc
High-pass (HPF): passes highs
ωo
Band-pass (BPF): passes a band
ωo
Band-reject (BRF / notch): blocks a band

Passive vs. active. Passive filters (HW 5) use only R, L, C. Their gain can never exceed 1, and connecting a load changes their behavior. Active filters add an op amp, which gives two advantages:

They also avoid bulky inductors.

5. The op amp toolkit

An op amp is a triangle with two inputs, $-$ (inverting) and $+$ (non-inverting), and one output. For this course it is always ideal:

Ideal op amp golden rules (valid when there is feedback from output to the $-$ input):
Rule 1: No current flows into either input: $i_+ = i_- = 0$.
Rule 2: The two inputs are at the same voltage: $v_- = v_+$. If $+$ is grounded, then $v_- = 0$ (the "virtual ground").

The one formula that solves most of Quiz 7: the inverting amplifier

If the $+$ input is grounded, the input goes through impedance $Z_i$ into the $-$ input, and $Z_f$ connects the $-$ input to the output, then:

$$\boxed{\;H(s)=\frac{V_o}{V_i}=-\frac{Z_f}{Z_i}\;}$$ $Z_i$ = everything between $V_i$ and the $-$ input. $Z_f$ = everything between the $-$ input and $V_o$.
Why? (2-line derivation) The $-$ input sits at 0 V (Rule 2). The current entering through $Z_i$ is $\frac{V_i-0}{Z_i}$. It cannot enter the op amp (Rule 1), so it all flows through $Z_f$: $\frac{V_i}{Z_i}=\frac{0-V_o}{Z_f}$. Rearrange: $\frac{V_o}{V_i}=-\frac{Z_f}{Z_i}$.
Other op amp blocksGainUsed for
Inverting amp with resistors $R_i$, $R_f$$-R_f/R_i$Adding gain to a filter
Voltage follower (buffer): output wired to $-$, signal into $+$$+1$, so $V_o = V_+$Copies a voltage without loading it (Problems 4a, 4b, 6, 7)
Inverting summing amp: inputs $V_1, V_2$ each through $R_i$, feedback $R_f$$V_o=-\frac{R_f}{R_i}(V_1+V_2)$Parallel band-reject (Problem 8)
Non-inverting amp$1+R_f/R_i$Positive gain

When a circuit is not a plain inverting amp (for example Problems 1, 6, 7), use KCL (Kirchhoff's current law: currents leaving a node add to zero) at each unknown node, along with the two golden rules.

6. Standard forms: how to read $K$, $\omega_c$, $\omega_o$, $\beta$, $Q$

Every answer key writes $H(s)$ in one of these shapes. Your job is to rearrange your $H(s)$ until it looks exactly like one, then read the constants.

TypeStandard formWhat the letters mean
1st-order LPF$H(s)=K\dfrac{\omega_c}{s+\omega_c}$$K$ = gain at DC, $\omega_c$ = cutoff
1st-order HPF$H(s)=K\dfrac{s}{s+\omega_c}$$K$ = gain at high freq, $\omega_c$ = cutoff
2nd-order LPF$H(s)=k\dfrac{c}{s^2+bs+c}$$k$ = DC gain, $\sqrt c$ = corner freq
2nd-order HPF$H(s)=k\dfrac{s^2}{s^2+bs+c}$$k$ = high-freq gain, $\sqrt c$ = corner freq
Band-pass$H(s)=k\dfrac{bs}{s^2+bs+c}$$k$ = gain at $\omega_o$, $\beta=b$, $\omega_o=\sqrt c$
Band-reject$H(s)=k\dfrac{s^2+c}{s^2+bs+c}$$k$ = passband gain, $\beta=b$, $\omega_o=\sqrt c$
For any 2nd-order BP or BR in the form $\dfrac{\cdots}{s^2+bs+c}$: $$\boxed{\omega_o=\sqrt{c}}\qquad \boxed{\beta=b}\qquad \boxed{Q=\frac{\omega_o}{\beta}}$$ $$\boxed{\omega_{c1}=-\frac{\beta}{2}+\sqrt{\left(\frac{\beta}{2}\right)^2+\omega_o^2}}\qquad \boxed{\omega_{c2}=+\frac{\beta}{2}+\sqrt{\left(\frac{\beta}{2}\right)^2+\omega_o^2}}$$ And going the other way: $\boxed{\omega_o=\sqrt{\omega_{c1}\omega_{c2}}}$ (geometric mean, not the average) and $\boxed{\beta=\omega_{c2}-\omega_{c1}}$.

The recipe for reading constants:

  1. Make the $s^2$ term in the denominator have coefficient 1. Divide top and bottom if needed.
  2. Denominator is now $s^2+bs+c$. Read $b$ and $c$.
  3. For a band-pass, rewrite the numerator as (number)$\times bs$. The number in front is $k$. Example: $\dfrac{-250s}{s^2+70s+1000}=\dfrac{-250}{70}\cdot\dfrac{70s}{s^2+70s+1000}$, so $k=-3.57$.
Why is $k$ the gain at the center for a band-pass? At $s=j\omega_o$ (with $\omega_o=\sqrt c$), the $s^2$ and $c$ parts of the denominator cancel: $s^2+c=-\omega_o^2+\omega_o^2=0$. The denominator $s^2+bs+c$ is left with only $bs$. So $H=k\frac{bs}{bs}=k$. The same trick shows the numerator of a band-reject is 0 at $\omega_o$, which is the notch.

7. The $s\to0$ / $s\to\infty$ test: identify any filter

Evaluate $H$ at very low frequency ($s\to0$) and very high frequency ($s\to\infty$):
$|H|$ as $s\to0$$|H|$ as $s\to\infty$Type
nonzero0Low-pass
0nonzeroHigh-pass
00Band-pass
nonzerononzero (and a dip in between)Band-reject

Two ways to do it:

A band-reject and the "shelf" circuit in Problem 9 both have nonzero gain at both ends. A band-reject has a notch (numerator $s^2+c$, which is zero at $\omega_o$). A shelf just moves from one level to another.

8. First-order active low-pass and high-pass filters

Fig 15.1 first-order low-pass: R1 input, R2 parallel C feedback

Fig 15.1: LPF. $R_1$ in, $R_2\parallel C$ in the feedback

Fig 15.4 first-order high-pass: R1 series C input, R2 feedback

Fig 15.4: HPF. $R_1$ + $C$ in series at the input, $R_2$ in the feedback

Low-pass derivation: $Z_i=R_1$, $Z_f=R_2\parallel C=\frac{R_2}{1+sR_2C}$:

$$H(s)=-\frac{Z_f}{Z_i}=-\frac{R_2}{R_1}\cdot\frac{1}{1+sR_2C}$$

To get the standard form, divide top and bottom of $\frac{1}{1+sR_2C}$ by $R_2C$:

$$H(s)=-\frac{R_2}{R_1}\cdot\frac{1/R_2C}{s+1/R_2C}$$

High-pass derivation: $Z_i=R_1+\frac1{sC}=\frac{1+sR_1C}{sC}$, $Z_f=R_2$:

$$H(s)=-\frac{R_2}{Z_i}=-\frac{R_2\,sC}{1+sR_1C}$$

Divide top and bottom by $R_1C$: the top becomes $\frac{R_2}{R_1}s$ and the bottom $s+\frac{1}{R_1C}$:

$$H(s)=-\frac{R_2}{R_1}\cdot\frac{s}{s+1/R_1C}$$
Transfer functionGainCutoff
LPF (Fig 15.1)$H(s)=-K\dfrac{\omega_c}{s+\omega_c}$$\boxed{K=\dfrac{R_2}{R_1}}$$\boxed{\omega_c=\dfrac{1}{R_2C}}$ (C is with $R_2$)
HPF (Fig 15.4)$H(s)=-K\dfrac{s}{s+\omega_c}$$\boxed{K=\dfrac{R_2}{R_1}}$$\boxed{\omega_c=\dfrac{1}{R_1C}}$ (C is with $R_1$)
Memory hook: the cutoff uses the resistor that touches the capacitor.

Design recipe (given $f_c$, gain, and $C$):

  1. Convert: $\omega_c=2\pi f_c$. If the gain is in dB: $K=10^{\text{dB}/20}$.
  2. Solve the cutoff formula for the resistor that touches $C$: LPF $R_2=\frac{1}{\omega_cC}$, HPF $R_1=\frac{1}{\omega_cC}$.
  3. Get the other resistor from $K=R_2/R_1$.
In the HPF, changing $R_2$ changes only the gain, because $\omega_c=1/R_1C$ does not contain $R_2$. In the LPF, changing $R_1$ changes only the gain. This is exactly Problem 2(b).

9. Scaling: from "prototype" to real parts

Textbook designs start from a prototype: all resistors 1 Ω, all capacitors 1 F, cutoff 1 rad/s. Those values are unrealistic, so you scale them:

$$\boxed{R' = k_m R}\qquad \boxed{C'=\frac{C}{k_m k_f}}\qquad \boxed{L'=\frac{k_m}{k_f}L}$$ $k_f=\dfrac{\omega_{\text{new}}}{\omega_{\text{old}}}$ moves the frequency (prototype: $k_f=\omega_c$). $k_m$ changes impedance level without changing frequency or gain.

How to pick them: $k_f$ comes from the frequency you need. Then choose $k_m$ so one given part comes out right:

$RC$ products set frequencies. Scaling $R$ up by $k_m$ and $C$ down by $k_m$ leaves $RC$ unchanged, so the frequency stays put. Dividing $C$ by $k_f$ shrinks $RC$, so the frequency goes up by $k_f$.

10. Broadband band-pass (cascade) and band-reject (parallel)

Band-pass = low-pass THEN high-pass (in series, "cascade")

Cascade: first-order low-pass stage followed by first-order high-pass stage

The signal must survive both filters. The LPF removes everything above $\omega_{c2}$, and the HPF removes everything below $\omega_{c1}$. What is left is the band between them.

Cascade BPF design (each stage is a unity-gain first-order filter, $R$'s equal): $$\boxed{\text{LPF stage sets the UPPER cutoff: } \omega_{c2}=\frac{1}{R_LC_L}}\qquad \boxed{\text{HPF stage sets the LOWER cutoff: } \omega_{c1}=\frac{1}{R_HC_H}}$$ Optional third stage (inverting amp, $R_f/R_i$) sets the passband gain.
Works only if broadband: $\omega_{c2}\gg\omega_{c1}$ (roughly $Q<\tfrac12$).

When given $\omega_o$ and $\beta$ instead of the cutoffs, solve the pair $\omega_{c1}\omega_{c2}=\omega_o^2$ and $\omega_{c2}-\omega_{c1}=\beta$. This gives the boxed $\omega_{c1,2}$ formulas from §6.

Show how (quadratic formula) From the second equation, $\omega_{c2}=\omega_{c1}+\beta$. Substitute into the first: $\omega_{c1}(\omega_{c1}+\beta)=\omega_o^2$, i.e. $$\omega_{c1}^2+\beta\,\omega_{c1}-\omega_o^2=0$$ Quadratic formula, keeping the positive root (frequencies are positive): $$\omega_{c1}=\frac{-\beta+\sqrt{\beta^2+4\omega_o^2}}{2}=-\frac\beta2+\sqrt{\left(\tfrac\beta2\right)^2+\omega_o^2},\qquad \omega_{c2}=\omega_{c1}+\beta=\frac\beta2+\sqrt{\left(\tfrac\beta2\right)^2+\omega_o^2}$$
The Low-pass sets the upper edge, and the High-pass sets the lower edge. It feels backwards, so double-check every time.

Band-reject = low-pass AND high-pass side by side, then added (parallel)

Fig 15.13 parallel band-reject: low-pass and high-pass in parallel feeding a summing amplifier

The LPF keeps lows, the HPF keeps highs, and the summing amp adds them. Frequencies in the middle are dropped by both, which leaves a notch.

Parallel BRF design: $$\boxed{\text{LPF sets the LOWER cutoff: } \omega_{c1}=\frac{1}{R_LC_L}}\qquad\boxed{\text{HPF sets the UPPER cutoff: } \omega_{c2}=\frac{1}{R_HC_H}}\qquad \boxed{\text{gain}=\frac{R_f}{R_i}}$$
For the band-reject the roles flip compared with the band-pass. The LPF's passband has to end at the bottom of the notch, so the LPF gets the lower cutoff.

11. Higher-order filters and Butterworth

A first-order filter falls off slowly. Higher order means a steeper fall-off:

$$\boxed{\text{roll-off} = -20\,n \text{ dB/decade}}\qquad (n=\text{order})$$ −20 dB/dec → $n=1$; −40 → $n=2$; −60 → $n=3$; −80 → $n=4$. ("Decade" = a factor of 10 in frequency.)

Simply cascading $n$ identical first-order stages makes the corner droopy and moves the cutoff. A Butterworth filter picks the stages so the passband is as flat as possible and the cutoff stays exactly at $\omega_c$:

$$|H(j\omega)|=\frac{1}{\sqrt{1+(\omega/\omega_c)^{2n}}}$$
Normalized Butterworth polynomials (denominator of $H(s)$ for $\omega_c=1$ rad/s)
$n$Denominator
1$(s+1)$
2$(s^2+\sqrt2\,s+1)=(s^2+1.414s+1)$
3$(s+1)(s^2+s+1)$
4$(s^2+0.765s+1)(s^2+1.848s+1)$
Odd order: one first-order stage $(s+1)$ plus second-order stages. Even order: second-order stages only.
For a real cutoff $\omega_c$, replace $s$ with $s/\omega_c$ and clear the fractions. Example, $n=2$: $\;s^2+\sqrt2\,\omega_c s+\omega_c^2$. If you see a denominator with $b=\sqrt2\sqrt c$, it is a 2nd-order Butterworth.
Fig 15.21 second-order Butterworth low-pass cell

Fig 15.21: 2nd-order LP cell (R = 1 Ω)

Fig 15.25 second-order Butterworth high-pass cell

Fig 15.25: 2nd-order HP cell (C = 1 F)

The LP cell (Fig 15.21) has, with $R=1\,\Omega$ (derived in the textbook with KCL at $V_a$ and at the $+$ input; given, memorize the result):

$$H(s)=\frac{1/(C_1C_2)}{s^2+\frac{2}{C_1}s+\frac{1}{C_1C_2}}$$

To make it equal $\dfrac{1}{s^2+b_1s+1}$, match coefficients:

$$\text{LP cell (R=1 Ω):}\quad \boxed{C_1=\frac{2}{b_1}}\qquad\boxed{C_2=\frac{1}{C_1}}\qquad\qquad \text{HP cell (C=1 F):}\quad \boxed{R_2=\frac{2}{b_1}}\qquad\boxed{R_1=\frac{1}{R_2}}$$ HP cell: $H(s)=\dfrac{s^2}{s^2+\frac{2}{R_2}s+\frac{1}{R_1R_2}}$, with $R_1$ in the feedback path and $R_2$ to ground (Eq. 15.17 and 15.18).
Signs: these cells are built around a voltage follower (output wired to $-$), so their gain is $+1$ and they don't flip the signal. Inverting stages (Fig 15.1, Fig 15.4, the inverting gain stage) each contribute a $-1$. Multiply the signs of all stages to get the overall sign.

Butterworth design recipe (low-pass):

  1. Find the order $n$ from the roll-off ($-20n$ dB/dec) or from the stopband formula below.
  2. Look up the polynomial; split into $(s+1)$ and $(s^2+b_1s+1)$ factors.
  3. Each $(s+1)$ → prototype first-order LPF (1 Ω, 1 Ω, 1 F). Each $(s^2+b_1s+1)$ → LP cell with $C_1=2/b_1$, $C_2=1/C_1$.
  4. Scale: $k_f=\omega_c$ (in rad/s), $k_m$ from the required R or C.
  5. Need gain $\ne1$? Add an inverting amplifier stage with $R_f/R_i=$ gain.
Finding the order from a stopband spec (Eq. 15.15 and 15.16) Spec: gain must be at least $A_p$ dB at $\omega_p$ (passband edge) and at most $A_s$ dB at $\omega_s$ (stopband edge). Both $A$'s are negative numbers. $$\sigma_p=\sqrt{10^{-0.1A_p}-1},\quad \sigma_s=\sqrt{10^{-0.1A_s}-1},\quad n=\frac{\log_{10}(\sigma_s/\sigma_p)}{\log_{10}(\omega_s/\omega_p)}$$ If $A_p=-3$ dB (so $\sigma_p\approx1$): $\;n\approx\dfrac{-0.05\,A_s}{\log_{10}(\omega_s/\omega_p)}$. Always round up to the next whole number.

12. Narrowband (high-Q) band-pass and the twin-T notch

The cascade/parallel tricks only work when the band is wide. For a narrow band ($Q$ large) use these single-op-amp circuits.

Narrowband band-pass (Fig 15.27)

Fig 15.27 narrowband bandpass: R1, R2 to ground, two capacitors, R3 feedback
$$H(s)=\frac{-\frac{1}{R_1C}\,s}{s^2+\frac{2}{R_3C}s+\frac{1}{R_{eq}R_3C^2}},\qquad R_{eq}=R_1\parallel R_2$$ $$\boxed{\beta=\frac{2}{R_3C}}\qquad\boxed{\omega_o^2=\frac{1}{R_{eq}R_3C^2}}\qquad \boxed{K=-\frac{R_3}{2R_1}}$$ Prototype design ($\omega_o=1$, $C=1$ F): $R_1=\dfrac{Q}{|K|}$, $R_2=\dfrac{Q}{2Q^2-|K|}$, $R_3=2Q$. Then scale (use $k_f=\omega_o$ and $k_m=\frac{1}{k_fC'}$).
$H(s)$ comes from KCL at $V_a$ and at the $-$ input, the same method as Quiz Problem 1. Given, memorize.
Quiz Problem 1 is this exact circuit with $R_2$ removed ($R_2=\infty$, so $R_{eq}=R_1$), and with $R_1=R_a$, $R_3=R_b$.

Twin-T notch (band-reject), Fig 15.30, Quiz Problem 6

Twin-T notch: two capacitors with R/2 shunt, two resistors with 2C shunt, two followers and sigma divider

The "twin T" is two T-shaped networks side by side:

The middle legs return to $\sigma V_o$, a fraction of the output made by the $(1-\sigma)R$ / $\sigma R$ voltage divider and buffered by the lower op amp.

$$H(s)=\frac{s^2+\frac{1}{R^2C^2}}{s^2+\frac{4(1-\sigma)}{RC}s+\frac{1}{R^2C^2}}$$ $$\boxed{\omega_o=\frac{1}{RC}}\qquad\boxed{\beta=\frac{4(1-\sigma)}{RC}}\qquad \boxed{Q=\frac{1}{4(1-\sigma)}}$$ $\sigma$ (between 0 and 1) is set by the voltage divider $(1-\sigma)R$ and $\sigma R$ on the output. Larger $\sigma$ gives a narrower notch.

13. Quiz 7 solved step by step

Try each problem yourself first, then open the steps.

Problem 1 narrowband BP

Problem 1 circuit: Ra to node A, capacitor C from A to inverting input, capacitor C from A to output, Rb from inverting input to output

Asked: (A) $H(s)=V_o/V_i$ in one of the four forms. (B) Filter type. (C) Pick $R_a$, $R_b$, $C$ so that $\beta=12{,}500$ rad/s and gain = 8 at resonance, with every resistor ≥ 1 kΩ.

Know first: golden rules, KCL, the band-pass form $k\frac{bs}{s^2+bs+c}$.

Step 1: Label nodes

Call the junction of $R_a$ and the two capacitors node $A$ (voltage $V_a$). The $-$ input is at 0 V because $+$ is grounded (Rule 2). This is not a plain inverting amp, because node A has a third branch (the top C to $V_o$). So use KCL.

Step 2: KCL at the $-$ input

Current arrives from $A$ through the lower $C$ and leaves through $R_b$ to $V_o$. No current enters the op amp:

$$(V_a-0)\,sC=\frac{0-V_o}{R_b}\;\Rightarrow\; V_a=-\frac{V_o}{sCR_b}$$
Step 3: KCL at node A

Sum of currents leaving A = 0 (to $V_i$ through $R_a$, to the $-$ input through C, to $V_o$ through the top C):

$$\frac{V_a-V_i}{R_a}+V_a\,sC+(V_a-V_o)\,sC=0$$ Multiply by $R_a$: $V_a(1+2sCR_a)-sCR_aV_o=V_i$.
Step 4: Substitute and solve $$-\frac{V_o}{sCR_b}(1+2sCR_a)-sCR_aV_o=V_i$$ Multiply through by $sCR_b$: $$-V_o\left[1+2sCR_a+s^2C^2R_aR_b\right]=sCR_b\,V_i$$ $$H(s)=\frac{-sCR_b}{s^2C^2R_aR_b+2sCR_a+1}$$ Divide top and bottom by $C^2R_aR_b$ so the $s^2$ coefficient is 1: $$H(s)=\frac{-\frac{1}{CR_a}s}{s^2+\frac{2}{CR_b}s+\frac{1}{C^2R_aR_b}}$$
Step 5: Match to $k\dfrac{bs}{s^2+bs+c}$ $b=\dfrac{2}{CR_b}$, $c=\dfrac{1}{C^2R_aR_b}$. The numerator must equal $k\cdot b\cdot s$: $\;k\cdot\dfrac{2}{CR_b}=-\dfrac{1}{CR_a}\Rightarrow k=-\dfrac{R_b}{2R_a}$.
Step 6: Type $s\to0$: numerator → 0, so $H\to0$. $s\to\infty$: $\sim\frac{s}{s^2}\to0$. Zero at both ends → band-pass. (Picture check: at DC both caps open, so nothing reaches the op amp. At high frequency the top cap shorts the output to node A, and the lower cap shorts A to virtual ground, so $V_o=0$.)
Step 7: Design

Gain at resonance = $|k|$ = 8: $\frac{R_b}{2R_a}=8\Rightarrow R_b=16R_a$.

Bandwidth: $\beta=b=\frac{2}{CR_b}=12{,}500$.

That is 2 equations and 3 unknowns, so you get to choose one. Pick the smallest allowed resistor: $R_a=1\text{ k}\Omega$. Then $R_b=16\text{ k}\Omega$ and

$$C=\frac{2}{\beta R_b}=\frac{2}{12{,}500\times16{,}000}=10\text{ nF}$$

(Bonus check: $\omega_o=\frac{1}{C\sqrt{R_aR_b}}=25{,}000$ rad/s, $Q=\omega_o/\beta=2$.)

(A) $H(s)=k\dfrac{bs}{s^2+bs+c}$, with $k=-\dfrac{R_b}{2R_a}$, $b=\dfrac{2}{CR_b}$, $c=\dfrac{1}{C^2R_aR_b}$   (B) Band-pass   (C) $R_a=1$ kΩ, $R_b=16$ kΩ, $C=10$ nF

Problem 2 (15.10) 1st-order HPF design

Problem 15.10 statement

Asked: design a HPF with $f_c=4$ kHz, gain 8, $C=250$ nF. Then: what changes if only the feedback resistor changes?

Steps
  1. Use Fig 15.4: $\omega_c=\frac1{R_1C}$, $K=\frac{R_2}{R_1}$.
  2. $\omega_c=2\pi(4000)=25{,}133$ rad/s.
  3. $R_1=\dfrac{1}{\omega_cC}=\dfrac{1}{25{,}133\times250\times10^{-9}}=159.2\ \Omega$
  4. $R_2=K R_1=8\times159.2=1273\ \Omega$
  5. (b) The feedback resistor is $R_2$, which appears only in $K$. So the passband gain changes and the cutoff frequency stays the same.
Solution circuit: 250 nF and 159 ohm in series input, 1273 ohm feedback
$R_1\approx159\ \Omega$ (in series with the 250 nF capacitor), $R_2\approx1273\ \Omega$ in the feedback. (b) Gain changes, cutoff unchanged.

Problem 3 (15.32) BP from one op amp

Problem 15.32: 400 ohm and 50 uF in series input, 5k parallel 10 uF feedback

Asked: show it is band-pass. Find $\omega_o$, $\beta$, gain $K$, the cutoffs and $Q$.

Step 1: Impedances $Z_i=400+\frac{1}{s(50\mu)}=\frac{1+0.02s}{s(50\times10^{-6})}$ (since $R_1C_1=400\times50\mu=0.02$).
$Z_f=5000\parallel10\mu\text{F}=\frac{5000}{1+0.05s}$ (since $R_2C_2=5000\times10\mu=0.05$).
Step 2: $H=-Z_f/Z_i$ $$H(s)=-\frac{5000}{1+0.05s}\cdot\frac{50\times10^{-6}s}{1+0.02s}=\frac{-0.25s}{(1+0.05s)(1+0.02s)}$$ Factor: $1+0.05s=0.05(s+20)$ and $1+0.02s=0.02(s+50)$, and $0.05\times0.02=0.001$: $$H(s)=\frac{-250s}{(s+20)(s+50)}=\frac{-250s}{s^2+70s+1000}$$
Step 3: Read off $c=1000\Rightarrow\omega_o=\sqrt{1000}=31.6$ rad/s. $\;b=\beta=70$ rad/s.
$k$: $-250s=k(70s)\Rightarrow k=-3.57$.
$Q=31.6/70=0.45$.
$\omega_{c1,2}=\mp35+\sqrt{35^2+1000}=\mp35+47.17$, so $\omega_{c1}=12.17$ and $\omega_{c2}=82.17$ rad/s.
$H(s)=\dfrac{-250s}{s^2+70s+1000}$ (band-pass: 0 at both ends). $\omega_o=31.6$ rad/s, $\beta=70$ rad/s, $K=-3.57$, $\omega_{c1}=12.17$ rad/s, $\omega_{c2}=82.17$ rad/s, $Q=0.45$.

Problem 4 identify type

(a) R series then C to ground into follower; (b) C series then R to ground into follower(c) Ri and Ci series input, Rf parallel Cf feedback
Steps (capacitor trick)

(a) R in series, C to ground, then a follower. Low $f$: C open, so $v_o=v_i$ (passes). High $f$: C shorts the node to ground, so $v_o=0$. → LPF.

(b) C in series, R to ground. Low $f$: C open, so nothing gets through. High $f$: C is a wire, so $v_o=v_i$. → HPF.

(c) Low $f$: $C_i$ open, so no input current and $v_o=0$. High $f$: $C_f$ shorts output to the $-$ input (0 V), so $v_o=0$. Zero at both ends → BPF. (Problem 3 is this same circuit with numbers.)

(a) LPF (b) HPF (c) BPF

Problem 5 (15.3) 1st-order LPF, dB gain

15.3 statement: low-pass, 10 dB gain, 2.5 kHz cutoff, 2.5 nF
Steps
  1. Fig 15.1: $\omega_c=\frac1{R_2C}$, $K=\frac{R_2}{R_1}$.
  2. Gain in dB → linear: $K=10^{10/20}=3.162$.
  3. $\omega_c=2\pi(2500)=15{,}708$ rad/s.
  4. $R_2=\dfrac{1}{\omega_cC}=\dfrac{1}{15{,}708\times2.5\times10^{-9}}=25{,}465\ \Omega$
  5. $R_1=\dfrac{R_2}{K}=\dfrac{25{,}465}{3.162}=8053\ \Omega$
Solution: 8053 ohm input, 25465 ohm parallel 2.5 nF feedback
$R_1=8053\ \Omega$, $R_2=25{,}465\ \Omega$, $C=2.5$ nF (with $R_2$ in the feedback).

Problem 6 twin-T notch

Twin-T notch filter with two followers and sigma divider

Asked: derive $H(s)$, name the filter, and design for a 60 Hz center frequency.

Read the picture:

Step 1: Three KCL equations $$\text{node }a:\;(V_a-V_i)sC+(V_a-V_o)sC+\frac{V_a-\sigma V_o}{R/2}=0$$ $$\text{node }b:\;\frac{V_b-V_i}{R}+\frac{V_b-V_o}{R}+(V_b-\sigma V_o)\,2sC=0$$ $$\text{top }+\text{ input (at }V_o\text{)}:\;(V_o-V_a)sC+\frac{V_o-V_b}{R}=0$$
Step 2: Solve (let $x=sRC$ to keep it short) Multiply each equation by $R$. Node a becomes $$x(V_a-V_i)+x(V_a-V_o)+2(V_a-\sigma V_o)=0\;\Rightarrow\;V_a(2x+2)=xV_i+(x+2\sigma)V_o$$ Node b becomes $$(V_b-V_i)+(V_b-V_o)+2x(V_b-\sigma V_o)=0\;\Rightarrow\;V_b(2+2x)=V_i+(1+2\sigma x)V_o$$ The top node becomes $x(V_o-V_a)+(V_o-V_b)=0$. So: $$V_a=\frac{xV_i+(x+2\sigma)V_o}{2(x+1)},\qquad V_b=\frac{V_i+(1+2\sigma x)V_o}{2(x+1)},\qquad (x+1)V_o=xV_a+V_b$$ Put the first two into the third and multiply by $2(x+1)$: $$2(x+1)^2V_o=(x^2+1)V_i+(x^2+4\sigma x+1)V_o$$ $$V_o\left[x^2+4(1-\sigma)x+1\right]=(x^2+1)V_i$$ Put back $x=sRC$ and divide by $R^2C^2$: $$H(s)=\frac{s^2+\frac1{R^2C^2}}{s^2+\frac{4(1-\sigma)}{RC}s+\frac1{R^2C^2}}$$
Step 3: Type and design Form $k\frac{s^2+c}{s^2+bs+c}$ with $k=1$ → band-reject (gain 1 at both ends, 0 at $\omega_o$).
$\omega_o=\frac1{RC}=2\pi(60)=120\pi=377$ rad/s. Choose a standard capacitor, $C=1\ \mu$F: $$R=\frac{1}{120\pi\times10^{-6}}=2653\ \Omega\approx2.65\text{ k}\Omega$$ ($\sigma$ is free here because no bandwidth was specified. It only sets $\beta=4(1-\sigma)\omega_o$.)
$H(s)=\dfrac{s^2+\frac{1}{R^2C^2}}{s^2+\frac{4(1-\sigma)}{RC}s+\frac{1}{R^2C^2}}$, band-reject (notch). $\frac1{RC}=120\pi$, so $C=1\ \mu$F and $R=2.65$ kΩ.

Problem 7 (P15.55) 2nd-order LPF

R1 into follower 1, C1 from its input to Vo, follower 1 output through R2 to node with C2 to ground, follower 2 gives Vo

Read the picture:

Steps

Follower 2: $V_o=V_B$. KCL at B: $\frac{V_B-V_A}{R_2}+V_BsC_2=0\Rightarrow V_A=V_o(1+sR_2C_2)$.

KCL at A (no current into follower 1): $\frac{V_A-V_i}{R_1}+(V_A-V_o)sC_1=0$. Multiply by $R_1$:

$$V_A(1+sR_1C_1)-sR_1C_1V_o=V_i$$ $$V_o\left[(1+sR_2C_2)(1+sR_1C_1)-sR_1C_1\right]=V_i$$ $$V_o\left[s^2R_1R_2C_1C_2+sR_2C_2+1\right]=V_i$$ Divide by $R_1R_2C_1C_2$: $$H(s)=\frac{\frac{1}{C_1C_2R_1R_2}}{s^2+\frac{1}{C_1R_1}s+\frac{1}{C_1C_2R_1R_2}}$$

$s\to0$: $H\to1$. $s\to\infty$: $H\to0$. → low-pass (form $k\frac{c}{s^2+bs+c}$ with $k=1$).

$H(s)=\dfrac{1/(C_1C_2R_1R_2)}{s^2+\frac{1}{C_1R_1}s+\frac{1}{C_1C_2R_1R_2}}$, a second-order low-pass filter.

Problem 8 (Example 15.6) parallel BRF from Bode plot

Bode magnitude: passband about 9.5 dB, -3 dB line at 6.54 dB, cutoffs near 100 and 2000 rad/s
Step 1: Read the plot
  • The −3 dB line is at 6.54 dB, so the passband is $6.54+3=9.54$ dB, and $K=10^{9.54/20}=3$.
  • The dashed line crosses the curve at $\omega_{c1}=100$ rad/s and $\omega_{c2}=2000$ rad/s.
  • $\omega_{c2}/\omega_{c1}=20$, which is broadband, so the parallel design works.
Step 2: Design (C = 0.5 µF everywhere)
  • LPF sets the lower cutoff: $R_L=\dfrac{1}{\omega_{c1}C}=\dfrac{1}{100\times0.5\times10^{-6}}=20$ kΩ.
  • HPF sets the upper cutoff: $R_H=\dfrac{1}{\omega_{c2}C}=\dfrac{1}{2000\times0.5\times10^{-6}}=1$ kΩ.
  • Summing amp gain $R_f/R_i=3$: pick $R_i=1$ kΩ, so $R_f=3$ kΩ.
$R_L=20$ kΩ, $R_H=1$ kΩ, $R_i=1$ kΩ, $R_f=3$ kΩ, all capacitors 0.5 µF.

Problem 9 (P15.6) derive + limits

R1 parallel C1 at input, R2 parallel C2 in feedback, inverting op amp
Steps

Inverting amp with $Z_i=R_1\parallel C_1=\frac{R_1}{1+sR_1C_1}$ and $Z_f=R_2\parallel C_2=\frac{R_2}{1+sR_2C_2}$:

$$H(s)=-\frac{Z_f}{Z_i}=-\frac{R_2}{R_1}\cdot\frac{1+sR_1C_1}{1+sR_2C_2}$$ Pull out $R_1C_1$ on top and $R_2C_2$ on the bottom: $\frac{R_2}{R_1}\cdot\frac{R_1C_1}{R_2C_2}=\frac{C_1}{C_2}$, so $$H(s)=-\frac{C_1}{C_2}\cdot\frac{s+\frac{1}{R_1C_1}}{s+\frac{1}{R_2C_2}}$$

$\omega\to0$ ($s=0$): $H=-\frac{C_1}{C_2}\cdot\frac{R_2C_2}{R_1C_1}=-\frac{R_2}{R_1}$ (caps open → plain resistor amp).

$\omega\to\infty$: $H\to-\frac{C_1}{C_2}$ (caps dominate; impedance ratio $\frac{1/sC_2}{1/sC_1}$).

$H(s)=-\dfrac{C_1}{C_2}\dfrac{s+1/R_1C_1}{s+1/R_2C_2}$; $\;H(j0)=-\dfrac{R_2}{R_1}$; $\;H(j\infty)=-\dfrac{C_1}{C_2}$

Problem 10 (15.30) cascade BPF design

15.30: unity-gain BPF cascade, f0 = 200 Hz, bandwidth 1000 Hz, 5 uF
Step 1: Find the cutoffs (in Hz, same formulas work) $f_{c1}f_{c2}=f_o^2=40{,}000$ and $f_{c2}-f_{c1}=1000$: $$f_{c1}=-500+\sqrt{500^2+200^2}=38.5\text{ Hz},\qquad f_{c2}=500+\sqrt{500^2+200^2}=1038.5\text{ Hz}$$
Step 2: Resistors

LPF sets the upper cutoff: $R_L=\dfrac{1}{2\pi f_{c2}C}=\dfrac{1}{2\pi(1038.5)(5\times10^{-6})}=30.65\ \Omega$.

HPF sets the lower cutoff: $R_H=\dfrac{1}{2\pi f_{c1}C}=\dfrac{1}{2\pi(38.52)(5\times10^{-6})}=826.4\ \Omega$.

Unity gain: each stage uses equal resistors, so gain = (−1)(−1) = +1. No third stage needed.

$f_{c1}=38.5$ Hz, $f_{c2}=1038.5$ Hz, $R_L=30.65\ \Omega$, $R_H=826.4\ \Omega$.

Problem 11 3rd-order Butterworth LPF

Asked: design a LPF with gain 4, $f_c=1$ kHz, roll-off −60 dB/dec, using 10 kΩ resistors. The key says "verify with a simulator", so here is a full design you can build in LTspice.

Step 1: Order−60 dB/dec = $-20n$, so $n=3$. Butterworth $n=3$: $(s+1)(s^2+s+1)$.
Step 2: Prototype stages

Stage 1, $(s+1)$: prototype first-order LPF (Fig 15.1) with $R_1=R_2=1\ \Omega$ and $C=1$ F. Gain $-1$.

Stage 2, $(s^2+1s+1)$, so $b_1=1$: LP cell (Fig 15.21) with $R=1\ \Omega$, $C_1=2/b_1=2$ F, $C_2=1/C_1=0.5$ F. Gain $+1$.

Step 3: Scale

$k_f=2\pi(1000)=6283$. Resistors must be 10 kΩ, so $k_m=10{,}000$. Capacitors divide by $k_mk_f=6.283\times10^7$:

  • $1\text{ F}\to15.92$ nF
  • $2\text{ F}\to31.83$ nF
  • $0.5\text{ F}\to7.96$ nF
Step 4: Gain of 4 Add an inverting amp: $R_i=10$ kΩ, $R_f=40$ kΩ (four 10 kΩ in series). Total gain $=(-1)(+1)(-4)=+4$.
Stage 1: 10 kΩ / 10 kΩ / 15.92 nF first-order LPF. Stage 2: Fig 15.21 cell with R = 10 kΩ, $C_1=31.83$ nF (feedback cap), $C_2=7.96$ nF (cap to ground). Stage 3: inverting amp, gain 4. Check: |H| = 4 at DC, $4/\sqrt2=2.83$ at 1 kHz, −60 dB/dec beyond (verified numerically).

14. One-page formula sheet and checklist

$$\omega=2\pi f\qquad Z_C=\frac1{sC}\qquad R\parallel C=\frac{R}{1+sRC}\qquad H=-\frac{Z_f}{Z_i}\qquad K=10^{\text{dB}/20}$$ $$\text{LPF (15.1): }K=\frac{R_2}{R_1},\ \omega_c=\frac1{R_2C}\qquad \text{HPF (15.4): }K=\frac{R_2}{R_1},\ \omega_c=\frac1{R_1C}$$ $$\omega_o=\sqrt c=\sqrt{\omega_{c1}\omega_{c2}}\qquad \beta=b=\omega_{c2}-\omega_{c1}\qquad Q=\frac{\omega_o}{\beta}\qquad \omega_{c1,2}=\mp\frac\beta2+\sqrt{\left(\tfrac\beta2\right)^2+\omega_o^2}$$ $$R'=k_mR\qquad C'=\frac{C}{k_mk_f}\qquad \text{roll-off}=-20n\text{ dB/dec}\qquad C_1=\frac2{b_1},\ C_2=\frac1{C_1}$$ $$\text{Narrowband BP: }\beta=\frac{2}{R_3C},\ K=-\frac{R_3}{2R_1}\qquad \text{Twin-T: }\omega_o=\frac1{RC},\ \beta=4(1-\sigma)\omega_o$$
Before you hand in, check:

Next: Flashcards → Practice Problems → Practice Quiz