Electrical Engineering › Section N · Decibels, filters, and Bode plots › Topic N.5
Limited-attenuation filters and crossover networks
Not every filter should cut a signal away completely. Adding one resistor makes an R-C filter level off at a fixed attenuation, a shape used for tone controls and equalisers, and simple first-order filters also steer bass and treble to the right loudspeakers.
Filters that level off
The R-C filters of R-C low-pass and high-pass filters keep falling at 20 dB per decade, so far enough into the stop-band the output approaches zero. Often that is not wanted: a treble control may need to cut high frequencies by 15 dB and no more. Adding a second resistor gives a limited-attenuation filter: its gain falls between two corner frequencies and then levels off at a fixed value, a shape called a shelf. Bode plots handle these filters well, because each one is just a pair of standard factors.
A low-pass filter with limited attenuation
Connect the source between node a and ground. Put \(R_1\) from a to b, then \(R_2\) in series with C from b to ground, with \(V_o\) across that \(R_2\)-C branch. At low frequencies C is nearly open, no current flows and \(V_o = V_i\). At high frequencies C is nearly a short, leaving a resistive divider, so \(V_o\) settles at \(\frac{R_2}{R_1 + R_2}V_i\) instead of zero. The gain is
with \(f_1 = \frac{1}{2\pi(R_1 + R_2)C}\) and \(f_2 = \frac{1}{2\pi R_2 C}\). The denominator gives a low-pass corner at \(f_1\), and the numerator a corner at \(f_2\) where the slope returns to zero.
Reading its Bode plot
The asymptotes are flat at 0 dB up to \(f_1\), fall at 20 dB per decade from \(f_1\) to \(f_2\), then run flat again at \(20\log_{10}\frac{R_2}{R_1 + R_2}\). The two corners are linked: \(f_2/f_1 = (R_1 + R_2)/R_2\), so the deeper the shelf, the more widely spaced the corners. The phase is 0° at both extremes, where the circuit is resistive; in between \(V_o\) lags, most at \(\sqrt{f_1 f_2}\). When the corners are close, their corrections partly cancel, so at each corner the curve sits a little less than 3 dB from the asymptotes; calculate exact values where they matter.
Worked example: a limited low-pass filter
In the circuit just described, \(R_1\) = 10 kΩ, \(R_2\) = 1.0 kΩ and C = 47 nF. 1. Lower corner: \(f_1 = \frac{1}{2\pi(11{,}000)(47 \times 10^{-9})} = 308\) Hz. 2. Upper corner: \(f_2 = \frac{1}{2\pi(1000)(47 \times 10^{-9})} = 3.39\) kHz. 3. Shelf level: \(20\log_{10}\frac{1.0}{11} = -20.8\) dB. 4. Check the slope: from 308 Hz to 3.39 kHz is a ratio of 11, about 1.04 decades, and \(1.04 \times 20 = 20.8\) dB, matching the shelf. 5. Exact values: at 1.02 kHz, the geometric mean, the gain is −10.4 dB, halfway down in dB, with the largest lag, 56.4°; at 30 kHz it is −20.8 dB with only 5.9° of lag.
A high-pass filter with limited attenuation
Now let the series element be \(R_1\) in parallel with C, from node a to node b, with \(R_2\) from b to ground and \(V_o\) across \(R_2\). At low frequencies C is open, so \(V_o = \frac{R_2}{R_1 + R_2}V_i\): the low end is attenuated but not removed. At high frequencies C shorts out \(R_1\) and \(V_o = V_i\). The gain is
with \(f_1 = \frac{1}{2\pi R_1 C}\) and \(f_2 = \frac{1}{2\pi (R_1 \parallel R_2) C}\). The asymptotes sit flat at the shelf level up to \(f_1\), rise at 20 dB per decade to \(f_2\), then stay at 0 dB. Here \(V_o\) leads in the middle region.
Worked example: a limited high-pass filter
In that circuit, \(R_1\) = 22 kΩ, \(R_2\) = 4.7 kΩ and C = 10 nF. 1. Shelf: \(20\log_{10}\frac{4.7}{26.7} = -15.1\) dB. 2. First corner: \(f_1 = \frac{1}{2\pi(22{,}000)(10 \times 10^{-9})} = 723\) Hz. 3. Parallel resistance: \(R_1 \parallel R_2 = \frac{(22)(4.7)}{26.7} = 3.87\,\text{k}\Omega\). 4. Second corner: \(f_2 = \frac{1}{2\pi(3870)(10 \times 10^{-9})} = 4.11\) kHz. 5. Sketch: flat at −15.1 dB below 723 Hz, rising 20 dB per decade to 4.11 kHz, flat at 0 dB above. 6. Exact check: at 1.72 kHz, the geometric mean, the gain is −7.5 dB with the output leading by 44.5°.
The standard factors at a glance
Every response in this section is built from a few factors, with \(f_1\) a corner frequency:
Unlike \(1 + j(f/f_1)\), which is flat below its corner, \(j(f/f_1)\) rises at every frequency. Add the factors' dB and phase to get the plot.
Crossover networks
No single loudspeaker reproduces the whole audio range, about 20 Hz to 20 kHz, well, so a crossover network steers bass to a woofer, treble to a tweeter and, in three-way systems, the middle to a midrange driver. The simplest is first order: an inductor L in series with the woofer forms an R-L low-pass filter, and a capacitor C in series with the tweeter forms an R-C high-pass filter, each driver acting roughly as a resistance R, often a nominal 8 Ω. At the crossover frequency \(f_x\) both outputs are 3 dB down, so each driver receives half the power, and ideally the two powers always add up to the full power.
Worked example: a two-way crossover
Design a first-order crossover at 2.5 kHz for an 8 Ω woofer and an 8 Ω tweeter, treating both as resistances. 1. Woofer inductor, from \(f_x = \frac{R}{2\pi L}\): \(L = \frac{8}{2\pi(2500)} = 0.509\,\text{mH}\). 2. Tweeter capacitor, from \(f_x = \frac{1}{2\pi RC}\): \(C = \frac{1}{2\pi(2500)(8)} = 7.96\,\mu\text{F}\). 3. Practical parts: 0.50 mH and 8.2 µF move the corners to 2.55 kHz and 2.43 kHz, close enough. 4. Check the separation: at 10 kHz the woofer is 12.2 dB down while the tweeter loses only 0.25 dB, and at 625 Hz the tweeter is 12.1 dB down. The gentle slopes give each driver limited protection from the other's band.
Three-way systems and real loudspeakers
A three-way crossover adds a midrange path with a capacitor and an inductor in series with the driver, a pass-band filter: the capacitor sets the lower crossover \(f_1\) and the inductor the upper one \(f_2\), using \(C = \frac{1}{2\pi f_1 R}\) and \(L = \frac{R}{2\pi f_2}\). Many designs use steeper second- or third-order networks, falling 40 or 60 dB per decade (about 12 or 18 dB per octave), so that tweeters are spared bass energy that could damage them. Real drivers are not pure resistances: their impedance varies with frequency and peaks at their own resonance.
Common mistakes to avoid
First, using \(R_2\) alone for both corners of the limited low-pass filter: the lower corner uses \(R_1 + R_2\), the upper one \(R_2\). Second, using \(R_1\) for the upper corner of the limited high-pass filter: it uses \(R_1 \parallel R_2\), the resistance the capacitor sees. Third, forgetting that the shelf level is a voltage ratio, so it takes 20 log. Fourth, putting the woofer's inductor in parallel with the driver: series L makes a low-pass path, series C a high-pass path. Fifth, treating a loudspeaker's nominal 8 Ω as exact at all frequencies.
Key takeaways
A limited-attenuation filter falls 20 dB per decade between two corners, then levels off at a shelf of 20 log₁₀[R₂/(R₁ + R₂)]. Limited low-pass: f₁ = 1/[2π(R₁ + R₂)C] and f₂ = 1/(2πR₂C); output lags in between. Limited high-pass: f₁ = 1/(2πR₁C) and f₂ = 1/[2π(R₁ ∥ R₂)C]; output leads in between. A first-order crossover puts series L before the woofer and series C before the tweeter, with L = R/(2πf_x) and C = 1/(2πf_xR); at f_x each driver is 3 dB down. Real drivers are not pure resistances.
Flashcards
Flashcard
A filter whose gain levels off at a fixed value instead of falling to zero
Test your understanding
Question 1 of 5
A limited-attenuation low-pass filter has R₁ = 15 kΩ and R₂ = 1.5 kΩ. What is its shelf level at high frequencies?
Question 2 of 5
In the limited-attenuation low-pass filter (R₁ in series, then R₂ in series with C to ground), which resistance sets the lower corner f₁?
Question 3 of 5
A first-order crossover for 4 Ω drivers is to cross over at 3.0 kHz. What series capacitor should feed the tweeter?
Question 4 of 5
At the crossover frequency of an ideal first-order two-way crossover, what does each driver receive?
Question 5 of 5
In the limited high-pass filter (R₁ in parallel with C as the series element, R₂ to ground), what sets the upper corner f₂?
