← About PHYS 252

PHYS 252 · Electronics

Lecture 03 · Capacitors & RC filters

1 Where Lab 3 Sits

Lab 3 brings a new part to the bench: the capacitor. First you measure three capacitors and work out how they behave at a low and a high frequency. Then you build one filter from a resistor and a capacitor, and you measure how it treats each frequency: where it cuts off, how much it shifts the phase, and its full Bode plot.

Lab 3 in three rows. Know the parts: Task 1, measure three capacitors and find their reactance. Build a filter: Task 2, design a high-pass filter and drive it at 20 Hz. Measure it: Task 3, find the cutoff frequency and the gain there, then Task 4, the phase at the cutoff, then Task 5, seven points for a Bode plot.
Figure 1. Lab 3 in three rows: know the parts (Task 1), build a filter (Task 2), measure it (Tasks 3, 4 and 5).

1.1 Task map

Lab 3 taskWhat you doConcept & section
1Measure three capacitors; compare polarized and unpolarized; find their reactance at two frequenciesCapacitors (§2), reactance (§4)
2Design and build a high-pass filter; drive it at a low frequencyThe low-pass (§5), the high-pass and design (§6)
3Measure the cutoff frequency; find the gain there in V/V and in dBCutoff and decibels (§5), finding fcf_c on the scope (§7)
4Measure the phase at the cutoff; compare with calculation and simulationPhase (§3, §5), phase from a time shift (§7)
5Measure seven points and draw a Bode plotThe Bode plot (§5, §6), from data to a plot (§7)

Every task needs an error analysis. Section 8 collects the main error sources for each kind of measurement.

2 Capacitors

2.1 A capacitor stores charge

A capacitor is two conductors separated by an insulator. Connect it to a voltage VV and charge flows until one plate holds +Q+Q and the other −Q-Q. No charge crosses the insulator; the charge on one plate pushes the same amount off the other. The charge is proportional to the voltage:

Q=CVQ = CV(1)

The constant CC is the capacitance, in farads (F). One farad stores one coulomb per volt, which is a lot: the parts on your bench run from picofarads (10−1210^{-12} F) to a few thousand microfarads (10−610^{-6} F).

A sketch of a parallel-plate capacitor connected to a battery V. The top plate carries plus signs and the label +Q; the bottom plate carries minus signs and the label minus Q. An insulator fills the gap, with field arrows pointing from the top plate to the bottom one. Below: Q = CV.
Figure 2. A capacitor connected to a voltage VV: +Q+Q on one plate, −Q-Q on the other, and an insulator between them.
Worked example: charge on a 47 µF capacitor

At 9 V, a 47 µF capacitor holds Q=47×10−6 F×9 V=4.23×10−4Q = 47 \times 10^{-6}\ \text{F} \times 9\ \text{V} = 4.23 \times 10^{-4} C, or 423 µC.

2.2 Polarized and non-polarized

Capacitors come in two families. A non-polarized capacitor, such as a ceramic or film part, works with either lead at the higher voltage. A polarized capacitor, such as an aluminum electrolytic or a tantalum part, has a + lead that must stay at the higher voltage.

Assorted capacitors on a gray background: large and small blue and black aluminum electrolytic cans marked with microfarads and volts, a metal-cased paper capacitor, orange and blue ceramic discs marked 103K, 151, 561 and 10, a green film capacitor marked 224K 100V, and two variable capacitors with ceramic frames.
Figure 3. Electrolytic cans (blue and black), ceramic discs (orange and blue), a film capacitor (green) and two variable capacitors. Photo: Eric Schrader, CC BY-SA 2.0.

The reason is in how an electrolytic is made. Its insulator is a thin oxide layer grown on aluminum foil. A thin insulator gives a large capacitance in a small can, but the oxide holds only while the foil is positive. Reverse the voltage and a leakage current flows through it: the part heats up and can bulge or vent. The can is marked with a stripe on the − side, and the − lead is usually the shorter one.

Two capacitor symbols. Left, non-polarized (ceramic, film): two straight parallel plates. Right, polarized (electrolytic, tantalum): one curved plate and one straight plate, with a plus sign on the straight side.
Figure 4. Schematic symbols. Left: non-polarized, two straight plates. Right: polarized, with a curved − plate and a + mark.

2.3 Three kinds side by side

The table compares one common series of each kind. Each row comes from that series' datasheet.

Kind (series)RangeTolerancePolarized?Used for
Aluminum electrolytic (Nichicon UVR)1 to 22 000 µF±20 %yeslarge values: smoothing a supply, storing charge
Ceramic, C0G or X7R (Vishay K series)10 pF to 1 µF±5, 10 or 20 %nocoupling and decoupling; C0G for stable values
Film, polyester (WIMA MKS 2)0.01 to 4.7 µF±5, 10 or 20 %nocoupling, decoupling, timing
A close-up row of small capacitors: a blue ceramic disc marked 561, small orange ceramic discs, a green film capacitor marked 224K 100V, a larger orange ceramic disc marked 103K 500V, and a small black electrolytic capacitor marked 22 microfarads 35 V.
Figure 5. Small capacitors up close: ceramic discs marked 561, 10, 103K, 7, 151 and 15, a film capacitor marked 224K 100V, and a 22 µF, 35 V electrolytic. Photo: Eric Schrader, CC BY-SA 2.0, cropped.

C0G ceramics keep their value over temperature. X7R ceramics pack more capacitance into the same size, but their value drifts more with temperature and with the voltage across them. Film parts sit in between and stay close to their value.

2.4 Reading the datasheet

Figure 6 is the specification table of a common electrolytic series. Three lines matter here. The rated voltage is the most you may put across the part. The capacitance tolerance is ±20 %, measured at 120 Hz and 20 °C. The tangent of the loss angle, tan δ, measures the loss: it is the ratio of the part's internal series resistance to its reactance (§4), so a perfect capacitor has tan δ = 0. For this series it runs from 0.28 for 6.3 V parts down to 0.08 for 100 V parts.

Specifications table from the Nichicon UVR aluminum electrolytic capacitor datasheet: category temperature range minus 40 to plus 85 degrees C; rated voltage 6.3 to 450 V; rated capacitance 1 to 22000 microfarads; capacitance tolerance plus or minus 20 percent at 120 Hz and 20 degrees C; leakage current limits; and the maximum tangent of loss angle, tan delta, for each rated voltage, from 0.28 at 6.3 V to 0.08 at 100 V.
Figure 6. Specifications of the Nichicon UVR aluminum electrolytic capacitors. Source: Nichicon datasheet CAT.8100M, p. 1.

Small ceramic and film parts are too small for a written value, so they carry a three-digit code in picofarads: two digits, then the number of zeros. A letter after it gives the tolerance: J = ±5 %, K = ±10 %, M = ±20 %. Figure 7 shows the same code inside a part number.

Ordering code table from the Vishay K series ceramic capacitor datasheet, using the example K104K15X7RF53H5. Position 1, K, is the product type. Positions 2 to 4, 104, are the capacitance in picofarads: two significant digits and a multiplier digit. Position 5 is the tolerance: J for 5 percent, K for 10 percent, M for 20 percent. Then the size code, the dielectric (X7R), the rated voltage code (F for 50 V, H for 100 V, K for 200 V, L for 500 V), lead diameter, packaging, lead style and lead spacing.
Figure 7. The ordering code of the Vishay K series ceramic capacitors. Positions 2 to 4 are the capacitance code and position 5 the tolerance. Source: Vishay datasheet, document 45171, p. 2.
Worked example: reading 104K

The digits 1, 0 and then four zeros give 100 000 pF = 100 nF = 0.1 µF. K means ±10 %.

A code under 100 is the value itself in picofarads: a disc marked 15 is 15 pF.

2.5 Measuring C and checking the tolerance

Many DMMs have a capacitance range. Take the capacitor out of the circuit and discharge it first: short its leads through a resistor, since a charged electrolytic can damage the meter's input. Then compare the reading with the nominal value, as you did for resistors in Lab 1:

deviation=Cmeas−CnomCnom×100 %\text{deviation} = \dfrac{C_{\text{meas}} - C_{\text{nom}}}{C_{\text{nom}}} \times 100\,\%(2)
Worked example: a 47 µF, ±20 % electrolytic

The DMM reads 43.9 µF. The deviation is (43.9 − 47)/47 × 100 % = −6.60 %. A ±20 % part may lie anywhere from 0.8 × 47 = 37.6 µF to 1.2 × 47 = 56.4 µF, so this one is in tolerance.

A capacitance axis from 34 to 60 microfarads. A shaded band from 37.6 to 56.4 microfarads is the plus or minus 20 percent tolerance of a 47 microfarad capacitor, marked at its nominal value. A DMM reading of 43.9 microfarads, minus 6.6 percent, lies inside the band.
Figure 8. The ±20 % band of a 47 µF capacitor, 37.6 to 56.4 µF, and the reading of 43.9 µF inside it.

The meter's capacitance range has its own accuracy, usually a few percent plus some counts, and it is often worse on the smallest ranges. Look it up in your meter's manual for the range you used.

Practice 2

(a) How much charge does a 220 µF capacitor hold at 12 V?

(b) A film capacitor is marked 224K. What are its nominal value and its tolerance range?

(c) A 220 µF, ±20 % electrolytic reads 241 µF. What is the deviation? Is it in tolerance?

(d) Why must an electrolytic's + lead go to the higher voltage?

3 AC Signals and the Capacitor's Current

3.1 A time-dependent signal

Until now every voltage stayed constant. In this lab the generator makes a sine wave, so the voltage changes with time:

V(t)=Acos⁡(ωt−φ)ω=2πfV(t) = A\cos(\omega t - \varphi) \qquad \omega = 2\pi f(3)

AA is the amplitude, ff the frequency in hertz, ω\omega the angular frequency in radians per second, and φ\varphi the phase. A phase is a shift in time. One period T=1/fT = 1/f is 360°, so a phase of φ\varphi shifts the wave by

Δt=φ360° T\Delta t = \dfrac{\varphi}{360°}\, T(4)
Two sine waves at 1 kHz plotted against time from minus 0.25 to 2 ms. V1 is 1.0 V cos(omega t); V2 is 0.6 V cos(omega t minus 60 degrees), smaller and shifted later. Arrows mark the period T = 1 ms and the delay of 167 microseconds between a peak of V1 and the next peak of V2.
Figure 9. Two 1 kHz sine waves. The second is smaller and shifted by 60°, which is 60/360 of the 1 ms period: 167 µs.

Two numbers describe how a circuit changes a sine wave: the ratio of the amplitudes and the phase shift. A circuit of resistors and capacitors never changes the frequency.

3.2 Complex form and Euler's formula

Sines are awkward to differentiate and add. Complex numbers make them easy. In electronics we write jj for −1\sqrt{-1}, since ii is already the current. Euler's formula links an exponential to a cosine and a sine:

ejφ=cos⁡φ+jsin⁡φe^{j\varphi} = \cos\varphi + j\sin\varphi(5)

So we write the signal as

V(t)=V0 ejωtV(t) = V_0\, e^{j\omega t}(6)

and keep in mind that the voltage you measure is its real part, V0cos⁡ωtV_0\cos\omega t. The number V0V_0 may be complex: its size is the amplitude and its angle is the phase.

Left: a unit circle with a rotating arrow e to the j omega t at angle omega t from the real axis; its projection on the real axis is cos omega t and its height is j sin omega t. Right: cos omega t plotted against omega t from 0 to 4 pi, with a dot at the angle of the arrow, showing that the real part traces the signal.
Figure 10. Left: ejωte^{j\omega t} is an arrow of length 1 turning at ω\omega radians per second. Its shadow on the real axis is cos⁡ωt\cos\omega t. Right: that shadow traced against ωt\omega t is the signal.

3.3 The current into a capacitor

Drive a capacitor with V(t)V(t). The charge follows the voltage, Q=CVQ = CV, and the current is the rate at which charge arrives:

I=dQdt=C dVdtI = \dfrac{dQ}{dt} = C\,\dfrac{dV}{dt}(7)

The familiar Q=ItQ = It holds only when the current is constant; here it is not, so we use the derivative.

An AC source V(t) connected directly across a capacitor C, with the current I(t) flowing from the source into the capacitor and the voltage V(t) marked across it. The bottom wire is grounded.
Figure 11. An AC source V(t)V(t) across a capacitor CC. The current I(t)I(t) charges and discharges it.

With the complex form, the derivative is one step. The derivative of ejωte^{j\omega t} is jω ejωtj\omega\, e^{j\omega t}, so

I=C ddt(V0 ejωt)=jωC V0 ejωtI = C\,\dfrac{d}{dt}\left(V_0\, e^{j\omega t}\right) = j\omega C\, V_0\, e^{j\omega t}(8)

The current has the same frequency as the voltage. Its amplitude is I0=ωCV0I_0 = \omega C V_0, and it grows with frequency: a capacitor lets more current through when the voltage changes faster.

Worked example: 1 µF at 5 V, 60 Hz

ω=2π×60=377.0\omega = 2\pi \times 60 = 377.0 rad/s, so I0=CωV0=10−6×377.0×5=1.885×10−3I_0 = C\omega V_0 = 10^{-6} \times 377.0 \times 5 = 1.885 \times 10^{-3} A = 1.885 mA. The period is 16.67 ms, and the current peaks a quarter period, 4.17 ms, before the voltage.

Voltage and current of a 1 microfarad capacitor driven by 5 V at 60 Hz, plotted against time for two periods (33.3 ms). The voltage, left axis, peaks at 5 V. The current, right axis, peaks at 1.885 mA a quarter period (4.17 ms) before each voltage peak.
Figure 12. The worked example to scale: 5 V at 60 Hz across 1 µF (left axis) and the current of 1.885 mA amplitude (right axis).

3.4 The current leads by 90°

The factor jj in Equation (8) is a phase. Since j=ej90°j = e^{j90°}, multiplying by jj turns a phasor 90° ahead. The current is therefore 90° ahead of the voltage: it peaks when the voltage crosses zero going up, where the voltage changes fastest, and it is zero at the voltage peaks, where the voltage stops changing for an instant.

A phasor diagram. The voltage phasor V = V0 lies along the real axis. The current phasor I = j omega C V0 points straight up along the imaginary axis, 90 degrees ahead. A note says that multiplying by j turns a phasor 90 degrees ahead.
Figure 13. The voltage phasor along the real axis and the current phasor jωCV0j\omega C V_0 straight up: the current leads by 90°.
Practice 3

(a) A 220 µF capacitor sits across a 120 Hz signal of 2 V amplitude. Find the amplitude of the current.

(b) For a 1 kHz signal, how long before the voltage peak does the capacitor's current peak?

(c) A sine at 2.5 kHz lags another by 40 µs. What is the phase shift in degrees?

(d) Why does Q=ItQ = It give the wrong answer for a capacitor on a sine wave?

4 Impedance and Reactance

4.1 Z = V/I

For a resistor, R=V/IR = V/I. For a capacitor on a sine wave we do the same and divide Equation (6) by Equation (8). The result is the impedance:

ZC=VI=V0 ejωtjωC V0 ejωt=1jωCZ_C = \dfrac{V}{I} = \dfrac{V_0\, e^{j\omega t}}{j\omega C\, V_0\, e^{j\omega t}} = \dfrac{1}{j\omega C}(9)
Two circuits side by side. Left: a source V drives current I through a resistor R, with Z_R = V/I = R. Right: a source V drives current I through a capacitor C, with Z_C = V/I = 1/(j omega C).
Figure 14. The same definition for both parts: Z=V/IZ = V/I. For a resistor it is RR; for a capacitor it is 1/(jωC)1/(j\omega C).

The factor ejωte^{j\omega t} cancels, so the impedance does not change with time. It changes with frequency: write ZZ or Z(ω)Z(\omega), not Z(t)Z(t). Its size is the reactance,

XC=∣ZC∣=1ωC=12πfCX_C = |Z_C| = \dfrac{1}{\omega C} = \dfrac{1}{2\pi f C}(10)

measured in ohms. Keep the two apart. XCX_C is a positive number of ohms; ZC=1/(jωC)=−jXCZ_C = 1/(j\omega C) = -jX_C is complex and carries the 90° phase.

4.2 Reactance falls as frequency rises

Equation (10) says XCX_C is inversely proportional to both ff and CC. At DC (ff = 0), XCX_C is infinite: a capacitor blocks DC. At high frequency, XCX_C is small: the capacitor acts almost like a wire. On log-log axes each capacitor gives a straight line that falls by a factor of ten per decade of frequency.

Capacitive reactance against frequency from 10 Hz to 1 MHz on log-log axes for 47 microfarads, 1 microfarad and 22 nanofarads: three parallel straight lines falling a factor of ten per decade. Dashed lines at 60 Hz and 20 kHz mark the worked values.
Figure 15. XCX_C against frequency for 47 µF, 1 µF and 22 nF. The dots mark 60 Hz and 20 kHz.
Worked example: three capacitors at 60 Hz and 20 kHz
CapacitorXCX_C at 60 HzXCX_C at 20 kHz
47 µF electrolytic1/(2π⋅60⋅47×10−6)1/(2\pi \cdot 60 \cdot 47 \times 10^{-6}) = 56.44 Ω0.1693 Ω
1 µF film1/(2π⋅60⋅10−6)1/(2\pi \cdot 60 \cdot 10^{-6}) = 2653 Ω7.958 Ω
22 nF ceramic1/(2π⋅60⋅22×10−9)1/(2\pi \cdot 60 \cdot 22 \times 10^{-9}) = 120.6 kΩ361.7 Ω

From 60 Hz to 20 kHz the frequency grows by 20 000/60 = 333.3, so each reactance shrinks by the same factor.

4.3 R and C in series: the impedance triangle

Impedances in series add, like resistors, but as complex numbers:

Z=R+1jωC=R−jXCZ = R + \dfrac{1}{j\omega C} = R - jX_C(11)

Draw RR along the real axis and −jXC-jX_C straight down from its tip. The impedance ZZ is the hypotenuse of a right triangle:

∣Z∣=R2+XC2tan⁡θ=−XCR|Z| = \sqrt{R^2 + X_C^2} \qquad \tan\theta = -\dfrac{X_C}{R}(12)
The impedance triangle of a series RC circuit. R points right along the real axis; from its tip, minus j X_C points straight down; the hypotenuse Z = R minus j X_C points down and to the right at angle theta below the R axis. Beside it: the magnitude of Z equals the square root of R squared plus X_C squared, tan theta = minus X_C over R, and theta is negative.
Figure 16. The impedance triangle of R and C in series. ZZ points below the R axis, so its angle θ\theta is negative.

The angle is measured from the R axis, so the side opposite it is XCX_C and the side next to it is RR. Writing tan⁡θ=R/XC\tan\theta = R/X_C gives the other acute angle of the triangle, the one at the tip of ZZ.

Worked example: 3.3 kΩ and 10 nF in series at 10 kHz

XC=1/(2π×104×10−8)=1591.5X_C = 1/(2\pi \times 10^4 \times 10^{-8}) = 1591.5 Ω. ∣Z∣=33002+1591.52=3663.7|Z| = \sqrt{3300^2 + 1591.5^2} = 3663.7 Ω. θ=−arctan⁡(1591.5/3300)=−25.75°\theta = -\arctan(1591.5/3300) = -25.75°.

With arctan⁡(3300/1591.5)\arctan(3300/1591.5) you would get 64.25°, the complementary angle, which is not the angle of ZZ.

The impedance triangle drawn to scale for 3.3 kilohms and 10 nanofarads at 10 kHz: R = 3300 ohms along the real axis, X_C = 1591.5 ohms downward, the magnitude of Z = 3663.7 ohms, and theta = minus 25.75 degrees.
Figure 17. The worked example drawn to scale.
Practice 4

(a) Find the reactance of a 220 µF capacitor at 120 Hz.

(b) At what frequency does a 22 nF capacitor have a reactance of 1 kΩ?

(c) Find ∣Z∣|Z| and θ\theta for 10 kΩ in series with 1 nF at 40 kHz.

5 The RC Low-Pass Filter

5.1 A divider with a capacitor

Put a capacitor in place of the lower resistor of a voltage divider and take the output across it. At low frequency XCX_C is large, the capacitor is nearly an open circuit, and almost all of VinV_{\text{in}} appears at the output. At high frequency XCX_C is small, the capacitor is nearly a short, and the output falls toward zero. The circuit passes low frequencies and stops high ones: it is a low-pass filter.

An RC low-pass filter: the source V_in drives a 3.3 kilohm resistor R in series with a 10 nanofarad capacitor C to ground; the output V_out is taken across the capacitor.
Figure 18. An RC low-pass filter: RR = 3.3 kΩ in series, CC = 10 nF to ground, output across CC.

5.2 The transfer function

The divider rule of Lecture 01 still holds, with impedances in place of resistances. The ratio of output to input is the voltage gain, also called the transfer function:

Av=VoutVin=ZCR+ZC=1/(jωC)R+1/(jωC)=11+jωRCA_v = \dfrac{V_{\text{out}}}{V_{\text{in}}} = \dfrac{Z_C}{R + Z_C} = \dfrac{1/(j\omega C)}{R + 1/(j\omega C)} = \dfrac{1}{1 + j\omega RC}(13)
A voltage divider made of two impedance boxes: Z_R = R on top and Z_C = 1/(j omega C) below, driven by V_in, with V_out taken at the node between them.
Figure 19. The divider in impedances: ZR=RZ_R = R on top, ZC=1/(jωC)Z_C = 1/(j\omega C) below.

Use the complex ZCZ_C in this step. The two impedances are 90° apart, so they do not add like numbers on a line. If you put the reactance XCX_C into the divider instead, XC/(R+XC)X_C/(R + X_C), you would get 0.5 at the cutoff instead of the correct 0.707.

5.3 The cutoff frequency

The product RCRC has units of time. It is the time constant τ=RC\tau = RC, and its inverse is the cutoff:

ωc=1τ=1RCfc=ωc2π=12πRC\omega_c = \dfrac{1}{\tau} = \dfrac{1}{RC} \qquad f_c = \dfrac{\omega_c}{2\pi} = \dfrac{1}{2\pi RC}(14)

With it, Equation (13) becomes

Av=11+j ω/ωc=11+j f/fcA_v = \dfrac{1}{1 + j\,\omega/\omega_c} = \dfrac{1}{1 + j\,f/f_c}(15)

At the cutoff, ωRC=1\omega RC = 1, which is the same as XC=RX_C = R. Below fcf_c the capacitor's reactance is the larger one and most of the input appears across it; above fcf_c the resistor takes most of it.

Worked example: R = 3.3 kΩ, C = 10 nF

τ=3300×10×10−9=33\tau = 3300 \times 10 \times 10^{-9} = 33 µs. ωc=1/τ=30 303\omega_c = 1/\tau = 30\,303 rad/s. fc=30 303/2π=4822.9f_c = 30\,303/2\pi = 4822.9 Hz, about 4.82 kHz. A simulation in ngspice puts the −3 dB point at the same 4822.9 Hz.

Reactance of a 10 nanofarad capacitor and a 3.3 kilohm resistance against frequency from 100 Hz to 1 MHz on log-log axes. The falling reactance line crosses the flat R line at f_c = 4823 Hz, where X_C = R. Below f_c, X_C is larger than R and most of V_in is across C; above f_c, X_C is smaller than R and C shorts the output.
Figure 20. XCX_C of 10 nF and RR = 3.3 kΩ against frequency. They cross at fcf_c = 4823 Hz.

5.4 Magnitude and phase in polar form

A complex gain is easier to read in polar form, as a size times a phase. For a>0a > 0, 1/(a+jb)=e−jarctan⁡(b/a)/a2+b21/(a + jb) = e^{-j\arctan(b/a)}/\sqrt{a^2 + b^2}. With a=1a = 1 and b=f/fcb = f/f_c, Equation (15) becomes

Av=11+(f/fc)2  e−jarctan⁡(f/fc)A_v = \dfrac{1}{\sqrt{1 + (f/f_c)^2}}\; e^{-j\arctan(f/f_c)}(16)

So the output amplitude is ∣Av∣|A_v| times the input amplitude, and the output lags the input by θ=−arctan⁡(f/fc)\theta = -\arctan(f/f_c). At fcf_c: ∣Av∣=1/2=0.7071|A_v| = 1/\sqrt{2} = 0.7071 and θ=−45°\theta = -45°.

Worked example: the 4.82 kHz low-pass at four frequencies
fff/fcf/f_c∣Av∣|A_v|θ\theta
1 kHz0.20730.9792−11.71°
4823 Hz10.7071−45.00°
10 kHz2.0730.4344−64.25°
50 kHz10.370.0960−84.49°
Two plots against f over f_c on a log axis from 0.01 to 100. Left, magnitude: 1/sqrt(1 + (f/f_c) squared), flat at 1, then falling, with 0.707 marked at f_c and 0.434 at 10 kHz. Right, phase: minus arctan(f/f_c), from 0 to minus 90 degrees, with minus 45 degrees at f_c and minus 64.25 degrees at 10 kHz.
Figure 21. Magnitude and phase of the low-pass against f/fcf/f_c, with fcf_c and 10 kHz marked.

5.5 Decibels and the −3 dB point

Gains span many powers of ten, so we often give them in decibels:

AdB=20log⁡10∣Av∣A_{\text{dB}} = 20\log_{10}|A_v|(17)

A gain of 1 is 0 dB, 0.5 is −6.02 dB, 0.1 is −20 dB and 0.01 is −40 dB. For the low-pass, Equation (16) gives

AdB=20log⁡1011+(f/fc)2=−10log⁡10 ⁣[1+(f/fc)2]A_{\text{dB}} = 20\log_{10}\dfrac{1}{\sqrt{1 + (f/f_c)^2}} = -10\log_{10}\!\left[1 + (f/f_c)^2\right](18)

At f=fcf = f_c this is −10log⁡102=−3.01-10\log_{10}2 = -3.01 dB. That is why the cutoff is also called the −3 dB point. It holds for every first-order RC filter, whatever its R and C: only f/fcf/f_c enters Equation (18). The same values in dB from the worked example above: −0.18 dB at 1 kHz, −7.24 dB at 10 kHz, −20.35 dB at 50 kHz.

5.6 The Bode plot

A Bode plot shows a filter's whole behavior in two graphs that share a log frequency axis: the gain in dB, and the phase in degrees. On these axes the low-pass looks like two straight lines joined by a bend. Far below fcf_c the gain is 0 dB. Far above, f/fc≫1f/f_c \gg 1, so AdB≈−20log⁡10(f/fc)A_{\text{dB}} \approx -20\log_{10}(f/f_c): the gain falls 20 dB for every factor of ten in frequency (a decade), or 6.02 dB for every factor of two (an octave). The two straight lines, the asymptotes, meet at fcf_c, where the real curve sits 3.01 dB below them.

Simulated Bode plot from ngspice of the 3.3 kilohm, 10 nanofarad low-pass, 10 Hz to 1 MHz. Top: gain in dB, flat at 0 dB, minus 3.01 dB at f_c = 4823 Hz, then falling at minus 20 dB per decade, with dashed straight-line asymptotes. Bottom: phase from 0 to minus 90 degrees, minus 45 degrees at f_c.
Figure 22. Bode plot of the 3.3 kΩ, 10 nF low-pass, simulated in ngspice from 10 Hz to 1 MHz. Dashed: the asymptotes. The netlist is ideal parts only.

The phase follows the same pattern: near 0° far below fcf_c, −45° at fcf_c, and close to −90° far above it. Most of the change happens within a decade on each side of fcf_c.

Worked example: a Bode table for the 4.82 kHz low-pass
ffHz∣Av∣|A_v|dBθ\theta
fc/20f_c/20241.10.9988−0.01−2.86°
fc/3f_c/31607.60.9487−0.46−18.43°
fcf_c4822.90.7071−3.01−45.00°
3fc3f_c14 4690.3162−10.00−71.57°
20fc20f_c96 4580.0499−26.03−87.14°

Between 3fc3f_c and 20fc20f_c the gain falls 16.03 dB over a factor of 6.67, which is log10 6.67 = 0.82 of a decade: 19.5 dB per decade, close to the asymptote's 20.

The gain in dB of the 4823 Hz low-pass against frequency, with five points marked: f_c/20 near 0 dB, f_c/3 at minus 0.46 dB, f_c at minus 3.01 dB, 3 f_c at minus 10.00 dB and 20 f_c at minus 26.03 dB.
Figure 23. The five points of the table on the gain curve.
Practice 5

(a) A low-pass uses 10 kΩ and 1 nF. Find fcf_c, and find ∣Av∣|A_v|, AdBA_{\text{dB}} and θ\theta at 40 kHz.

(b) A low-pass has a gain of −12 dB at 25 kHz. What is its fcf_c?

(c) Why must Equation (13) use ZCZ_C and not XCX_C? Show what goes wrong at fcf_c.

6 The RC High-Pass Filter and Design

6.1 Swap R and C

Exchange the two parts: the capacitor in series, the resistor to ground, and the output across the resistor. Now the capacitor blocks low frequencies and passes high ones. This is a high-pass filter.

An RC high-pass filter: the source V_in drives a 10 nanofarad capacitor C in series with a 3.3 kilohm resistor R to ground; the output V_out is taken across the resistor.
Figure 24. An RC high-pass filter: CC = 10 nF in series, RR = 3.3 kΩ to ground, output across RR.

The divider rule gives

Av=RR+ZC=jωRC1+jωRC=j f/fc1+j f/fcA_v = \dfrac{R}{R + Z_C} = \dfrac{j\omega RC}{1 + j\omega RC} = \dfrac{j\,f/f_c}{1 + j\,f/f_c}(19)

with the same fc=1/(2πRC)f_c = 1/(2\pi RC). In polar form,

∣Av∣=f/fc1+(f/fc)2θ=90°−arctan⁡(f/fc)|A_v| = \dfrac{f/f_c}{\sqrt{1 + (f/f_c)^2}} \qquad \theta = 90° - \arctan(f/f_c)(20)

The output leads the input. The lead is close to +90° far below fcf_c, +45° at fcf_c, and falls toward 0° far above it. At fcf_c the gain is again 0.7071, or −3.01 dB.

6.2 Designing for a cutoff

A design starts from the cutoff you want. Equation (14) has two unknowns, so choose one part and compute the other. Choose the capacitor first, because capacitors come in fewer values than resistors. Then compute R=1/(2πfcC)R = 1/(2\pi f_c C), round it to a standard value, and recompute fcf_c with the value you will use.

Keep RR between about 1 kΩ and 100 kΩ. Much below 1 kΩ, the filter draws enough current that the generator's 50 Ω output resistance starts to matter (Lecture 01). Much above 100 kΩ, the scope's 1 MΩ input sits in parallel with the output and changes the circuit.

Worked example: a 5 kHz filter with 10 nF

R=1/(2π×5000×10×10−9)=3183.1R = 1/(2\pi \times 5000 \times 10 \times 10^{-9}) = 3183.1 Ω. The nearest E24 value is 3.3 kΩ, which gives fc=1/(2π×3300×10×10−9)=4823f_c = 1/(2\pi \times 3300 \times 10 \times 10^{-9}) = 4823 Hz, 3.5 % below the target. These are the parts of Figures 18 and 24.

A design chart on log-log axes: resistance R from 10 ohms to 10 megohms against capacitance C from 1 nanofarad to 10 microfarads. Three falling lines show the pairs that give f_c = 50 Hz, 5 kHz and 50 kHz. A shaded band marks R from 1 kilohm to 100 kilohms. A marked point at 10 nanofarads: the exact R is 3183.1 ohms, and the E24 value 3.3 kilohms gives f_c = 4823 Hz.
Figure 25. Pairs of RR and CC for three cutoffs: 50 Hz, 5 kHz and 50 kHz. The shaded band is the useful range of RR. The dot marks 10 nF with 3.3 kΩ.

6.3 The two filters on one Bode plot

The same R and C, wired both ways round, give mirror images about fcf_c. The high-pass at fc/nf_c/n has the same gain as the low-pass at nfcn f_c, and the same phase with the opposite sign. Below fcf_c, the high-pass gain rises at +20 dB per decade.

Simulated Bode plots from ngspice of the low-pass and the high-pass built from the same 3.3 kilohms and 10 nanofarads. Top: the low-pass gain is flat then falls; the high-pass gain rises then is flat; they cross at f_c, both at minus 3.01 dB. Bottom: the low-pass phase goes from 0 to minus 90 degrees and the high-pass phase from plus 90 to 0 degrees, minus 45 and plus 45 degrees at f_c.
Figure 26. Low-pass and high-pass from 3.3 kΩ and 10 nF, simulated in ngspice. Both are −3.01 dB at fcf_c; the phases are −45° and +45° there.
Worked example: the 4.82 kHz high-pass at four points

At fc/20f_c/20: 0.0499 at +87.14°. At fc/3f_c/3: 0.3162 at +71.57°. At 3fc3f_c: 0.9487 at +18.43°. At 20fc20f_c: 0.9988 at +2.86°. Compare the low-pass table in §5.6.

6.4 Driving a high-pass far below its cutoff

Worked example: the 4.82 kHz high-pass at 300 Hz, 1.0 Vpp in

f/fc=300/4822.9=0.0622f/f_c = 300/4822.9 = 0.0622. Equation (20) gives ∣Av∣=0.0622/1+0.06222=0.06208|A_v| = 0.0622/\sqrt{1 + 0.0622^2} = 0.06208, or 20log⁡100.06208=−24.1420\log_{10}0.06208 = -24.14 dB, and θ=90°−arctan⁡0.0622=+86.44°\theta = 90° - \arctan 0.0622 = +86.44°.

The output is 0.06208×1.0=62.080.06208 \times 1.0 = 62.08 mV peak to peak, or 31.04 mV peak. One period at 300 Hz is 3.333 ms, so the output leads by 86.44/360×3.333=0.8086.44/360 \times 3.333 = 0.80 ms. An ngspice transient run of the same circuit gives 62.08 mVpp.

A simulated two-channel scope screen at 1 ms per division. CH1 at 200 mV per division shows the 1.00 V peak-to-peak, 300 Hz input. CH2 at 20 mV per division shows the high-pass output, 62.1 mV peak-to-peak, leading the input by 86.44 degrees, 0.80 ms. The gain is 0.0621, minus 24.14 dB.
Figure 27. The worked example as a scope would show it, simulated in ngspice. CH2 is set ten times more sensitive than CH1 so the small output fills the screen.

Far below its cutoff, a high-pass makes a small output that is nearly 90° ahead. The output is small, so set the CH2 scale to fit it and say which scale you used.

Practice 6

(a) Design a high-pass for 20 kHz with a 1 nF capacitor. Choose an E24 resistor and recompute fcf_c.

(b) A high-pass with fcf_c = 8 kHz is driven at 12 kHz with 1.0 Vpp. Find the output amplitude and the phase.

(c) A classmate builds a high-pass from 220 Ω and 1 µF. What could go wrong when the generator drives it?

7 Measuring a Filter on the Scope

7.1 CH1 on the input, CH2 on the output

To measure a filter, watch both ends at once. Connect CH1 to the input node and CH2 to the output node, and clip both probe grounds to the circuit ground. Trigger on CH1, the input, so the time axis stays fixed while you compare the two traces. Read both amplitudes at every frequency.

Measuring a filter on the scope: a generator V_in drives R then C to ground. CH1 connects to the input node before R, CH2 to the output node between R and C, and both probe grounds connect to the circuit ground.
Figure 28. Measuring a filter: the generator drives R and C; CH1 reads the input and CH2 the output; both grounds go to circuit ground.

7.2 Finding the cutoff

At the cutoff, Vout/Vin=0.707V_{\text{out}}/V_{\text{in}} = 0.707. Set the generator near your calculated fcf_c, then change the frequency until the ratio of the two amplitudes is 0.707. Use the ratio, not VoutV_{\text{out}} alone. The filter's input impedance changes with frequency, so the generator's 50 Ω output resistance (Lecture 01) lets VinV_{\text{in}} shift a little as you sweep.

A scope screen at 0.5 V per division and 50 microseconds per division with the low-pass at f = 4823 Hz. CH1, the input, is 2.00 V peak to peak; CH2, the output, is 1.414 V peak to peak and slightly later. Dotted lines mark the output peaks. The ratio 1.414 over 2.00 is 0.707.
Figure 29. The 4.82 kHz low-pass at its cutoff: 2.00 Vpp in, 1.414 Vpp out, a ratio of 0.707.

Near fcf_c the ratio changes slowly with frequency (§8 works out how slowly), so read the amplitudes as precisely as the scope allows: use the scope's amplitude measurements or the cursors, and spread each trace over most of the screen.

7.3 Phase from the time shift

The scope shows time, not phase, so measure the time shift Δt\Delta t between matching points of the two traces, such as their zero crossings going down. Equation (4) turns it into a phase:

θ=−360°×ΔtT\theta = -360° \times \dfrac{\Delta t}{T}(21)

The minus sign is for an output that comes later (a lag). For a high-pass the output comes earlier, and the phase is positive.

Worked example: the 4.82 kHz low-pass at 10 kHz

Theory gives θ=−64.25°\theta = -64.25° (§5.4). At 10 kHz, TT = 100 µs, so the expected delay is 64.25/360×100=17.8564.25/360 \times 100 = 17.85 µs. Suppose the cursors read Δt\Delta t = 17.8 µs: θ=−360°×17.8/100=−64.08°\theta = -360° \times 17.8/100 = -64.08°. At the cutoff of the same filter, TT = 207.35 µs and the −45° lag is a delay of T/8T/8 = 25.92 µs.

A scope screen at 0.5 V per division and 20 microseconds per division with the low-pass at 10 kHz. Dashed cursors mark the falling zero crossings of the input (CH1) and the smaller output (CH2); the output crosses 17.8 microseconds later. One period is 100 microseconds, so theta = minus 360 degrees times 17.8/100 = minus 64.1 degrees.
Figure 30. The low-pass at 10 kHz: the output crosses zero 17.8 µs after the input. One period is 100 µs, so θ\theta = −64.1°.

7.4 From measurements to a Bode plot

For each frequency, record ff, VinV_{\text{in}} and VoutV_{\text{out}} (and Δt\Delta t if you want the phase). Compute 20log⁡10(Vout/Vin)20\log_{10}(V_{\text{out}}/V_{\text{in}}) and plot it against ff on a log axis. Draw the theory curve for your calculated fcf_c on the same axes, and mark which points are measured.

A report Bode plot: gain in dB, 20 log of V_out over V_in, against frequency from 100 Hz to about 160 kHz on a log axis. The theory curve of the low-pass with f_c = 4823 Hz, and nine measured points from simulated example data that lie close to it, with a dashed line at minus 3 dB.
Figure 31. A report plot of simulated example data: nine "measured" points of the 4.82 kHz low-pass against its theory curve.

Spread the frequencies evenly on the log axis, and put more of them near fcf_c, where the curve bends.

Practice 7

(a) A low-pass is driven at 2 kHz, and its output crosses zero 40 µs after the input. Find θ\theta, then use θ=−arctan⁡(f/fc)\theta = -\arctan(f/f_c) to estimate fcf_c.

(b) At one frequency the scope reads 2.04 Vpp on CH1 and 1.40 Vpp on CH2. For a low-pass, is that frequency above or below fcf_c?

(c) Why trigger on CH1 rather than CH2 when you measure a phase?

8 Error Analysis

Every task in this course needs an error analysis. For each value you report, state its uncertainty, name the main sources of error, and compare it with the value you expected, simulated or took from a datasheet.

MeasurementMain error sources
CapacitanceThe DMM's capacitance accuracy for the range; the leads' own capacitance for small values; a part not fully discharged
Calculated fcf_c or XCX_CTolerance of R and C; use measured values where you have them
Measured fcf_cReading the two amplitudes; the slow change of the ratio near fcf_c; the generator's frequency display
PhaseReading Δt\Delta t (cursor placement, trace thickness, noise); the period TT
Bode pointsAmplitude readings at small outputs; probe and channel settings; the scope's input loading the output

8.1 How part tolerance moves the cutoff

Since fc=1/(2πRC)f_c = 1/(2\pi RC), the relative errors of R and C carry straight into fcf_c. If the errors are independent, add them in quadrature. For the worst case, push both the same way.

Worked example: R ±5 %, C ±10 %

In quadrature: 52+102=11.18 %\sqrt{5^2 + 10^2} = 11.18\,\%, so fcf_c = 4823 Hz ± 11.18 %.

Worst case: both low gives fc/(0.95×0.90)=5641f_c/(0.95 \times 0.90) = 5641 Hz (+16.96 %); both high gives fc/(1.05×1.10)=4176f_c/(1.05 \times 1.10) = 4176 Hz (−13.42 %). The range is not symmetric because fcf_c goes as 1/(RC)1/(RC).

Two plots near the cutoff of the 4823 Hz low-pass. With R plus or minus 5 percent and C plus or minus 10 percent, the gain curve (left) and the phase curve (right) can lie anywhere in a shaded band, and the cutoff can lie anywhere from 4176 to 5641 Hz, marked by two vertical lines. Dashed lines mark minus 3 dB and minus 45 degrees.
Figure 32. Gain and phase of the 4.82 kHz low-pass with R ±5 % and C ±10 %. The curve can lie anywhere in the shaded band, and the cutoff anywhere from 4176 to 5641 Hz.

Measure R and C before you build, and calculate fcf_c from the measured values. Then the comparison with your measured fcf_c tests the circuit, not the tolerance band.

8.2 How reading errors move the cutoff

Near fcf_c both curves are fairly flat, so a small reading error becomes a larger frequency error. With x=f/fcx = f/f_c, the slope of the magnitude at x=1x = 1 is d∣Av∣/dx=−x(1+x2)−3/2=−0.3536d|A_v|/dx = -x(1 + x^2)^{-3/2} = -0.3536. An error of 1 % in ∣Av∣|A_v| (0.00707) is then an error of 0.00707/0.3536=0.0200.00707/0.3536 = 0.020 in xx: 2.0 % in fcf_c. The phase has slope dθ/dx=−1/(1+x2)d\theta/dx = -1/(1 + x^2) rad = −28.65° per unit of xx at fcf_c, so a 1° error in phase is a 1/28.65=0.0351/28.65 = 0.035 error in xx: 3.5 % in fcf_c.

A difference between measured and expected is only meaningful when it is larger than the uncertainty. When it is, say which source most likely explains it, and check that the size fits.

← All course materials