How to Calculate and Choose the Bandwidth of a Filter for Real Circuits

The bandwidth of a filter is the specific range of frequencies between the upper and lower -3 dB (half-power) cutoff points where the circuit passes signals with minimal attenuation. In a real circuit, this parameter dictates exactly which data, audio, or RF signals reach your microcontroller's ADC or amplifier, while rejecting out-of-band noise. If you set it wrong, you either alias high-frequency switching noise into your digital readings or choke off the high-end of your audio signal.

The Math and a Worked Numeric Example

For a band-pass or band-stop filter, bandwidth ($BW$) is simply the difference between the upper cutoff frequency ($f_H$) and the lower cutoff frequency ($f_L$):

BW = f_H - f_L

For low-pass and high-pass filters, the bandwidth is technically from DC (0 Hz) to $f_c$, or from $f_c$ to infinity, but in practical engineering, we usually refer to the "passband width" up to the Nyquist limit or the physical limits of the op-amp.

Let’s run a worked numeric example on the bench. Suppose you are building an intercom system and need an active Multiple Feedback (MFB) Band-Pass Filter to isolate the human voice band while rejecting 60 Hz mains hum and high-frequency switching noise from a nearby buck converter. We want to pass frequencies from 300 Hz to 3400 Hz.

• Lower Cutoff ($f_L$): 300 Hz
• Upper Cutoff ($f_H$): 3400 Hz
• Calculated Bandwidth: $3400 - 300 = 3100$ Hz
• Center Frequency ($f_c$): $\sqrt{300 \times 3400} \approx 1010$ Hz
• Quality Factor ($Q$): $f_c / BW = 1010 / 3100 \approx 0.326$

Because the $Q$ is very low (wide bandwidth), a single MFB stage will be sloppy. A better approach is cascading a 2nd-order Sallen-Key High-Pass (set to 300 Hz) and a 2nd-order Sallen-Key Low-Pass (set to 3400 Hz). Using standard E12 component values and a dual TL072 op-amp:

1. High-Pass Stage (300 Hz): Use two 10 nF capacitors. The required resistor is $R = 1 / (2 \pi \times 300 \times 10nF) \approx 53 k\Omega$. We pick the nearest standard value: 51 kΩ.
2. Low-Pass Stage (3400 Hz): Use two 1 nF capacitors. The required resistor is $R = 1 / (2 \pi \times 3400 \times 1nF) \approx 46.8 k\Omega$. We pick the nearest standard value: 47 kΩ.

This cascaded design yields a flat 3100 Hz bandwidth with a steep 4th-order (24 dB/decade) roll-off outside the passband, perfectly tailored for voice.

What People Commonly Confuse With Bandwidth

When reading datasheets or designing in SPICE, it is easy to mix up bandwidth with adjacent concepts. Here is what trips up most hobbyists and junior engineers:

Bandwidth vs. Transition Band (Roll-off)

Bandwidth is the "flat" part of the road where your signal travels unimpeded. Think of the transition band like a highway speed trap zone—the speed limit doesn't drop from 65 to 0 instantly; it tapers off. A filter's roll-off (measured in dB/decade or dB/octave) is that taper. A 1st-order filter has a gentle 20 dB/decade taper, meaning signals just outside your bandwidth still leak through. A 4th-order Butterworth drops at 80 dB/decade, acting like a brick wall immediately after the -3 dB point.

Bandwidth vs. Center Frequency

Center frequency ($f_c$) is the geometric midpoint of a band-pass filter. Bandwidth is the total width of the passband. You can have two filters with the exact same 10 kHz bandwidth, but one is centered at 100 kHz (RF IF stage) and the other at 10 kHz (audio). The component values and op-amp Gain-Bandwidth Product (GBWP) requirements for these two circuits will be vastly different.

The -3 dB Point vs. Absolute Zero

A common mistake is assuming a low-pass filter set to a 1 kHz bandwidth completely blocks a 2 kHz signal. It doesn't. At 2 kHz (one octave past the cutoff of a 1st-order filter), the signal is only attenuated by about 7 dB. If your ADC is highly sensitive, that "blocked" signal will still show up in your digital readout. Always calculate the required attenuation at your specific noise frequency to determine the necessary filter order.

Where You Meet Filter Bandwidth in Practice

You will encounter bandwidth constraints across almost every mixed-signal and RF design. Here are the three most common jobsite and bench scenarios:

• Audio Crossovers: In bi-amplified speaker builds, the bandwidth of the low-pass filter feeding the woofer must perfectly mirror the high-pass filter feeding the tweeter. If the bandwidths overlap too much, you get a +3 dB acoustic bump at the crossover frequency; if they gap, you get a dead zone in the midrange.
• Superheterodyne RF Receivers: In amateur radio or SDR builds, the Intermediate Frequency (IF) stage relies on a highly specific bandwidth (e.g., 500 Hz for CW, 2.7 kHz for SSB) to reject adjacent channel interference. Using a 10 kHz bandwidth filter for an SSB signal will let in massive amounts of broadband noise, destroying your Signal-to-Noise Ratio (SNR).
• Switching Power Supply (SMPS) EMI: When filtering the output of a buck converter, the LC filter bandwidth must be set at least one decade below the switching frequency to adequately suppress the switching ripple, while remaining high enough to allow the control loop to respond to load transients.

Decision Path: Picking Your Filter Topology and Bandwidth

Use this decision tree to lock in your topology and select a concrete part number for your next build. Do not overcomplicate the design; pick the simplest topology that meets your roll-off requirements.

Default Recommendation: If you are building a general-purpose mixed-signal sensor interface on a breadboard or custom PCB and are unsure which to pick, default to a 2nd-order Sallen-Key Low-Pass filter using a MCP6002 or TLV2372 op-amp. It requires only two resistors, two capacitors, and provides a 40 dB/decade roll-off that is sufficient for 90% of hobbyist and commercial IoT sensor applications.

FAQ: Fine-Tuning Your Filter Design