A high pass filter removes low-frequency content below a chosen cutoff point while letting everything above it through cleanly. You’d use one any time unwanted low-frequency energy is muddying your signal, whether that’s rumble in a vocal recording, wind noise in field audio, baseline drift in a medical sensor, or subsonic vibrations threatening your speakers. The specific cutoff frequency and filter steepness depend entirely on what you’re filtering and why.
How a High Pass Filter Works
Every high pass filter has a cutoff frequency. Below that point, the filter progressively reduces the signal’s strength. At the cutoff itself, the output drops to about 70.7% of the original signal level (a 3 dB reduction). The further a frequency falls below the cutoff, the more aggressively it gets attenuated.
How aggressively depends on the filter’s slope, measured in decibels per octave. A 6 dB per octave slope rolls off gently. A 12 dB slope is moderate. A 24 dB slope cuts sharply and is common in synthesizers and electronic music tools. To put that in practical terms: a 12 dB per octave high pass filter set at 100 Hz will reduce a 50 Hz tone by 12 dB. A 24 dB slope at the same cutoff would reduce that same 50 Hz tone by 24 dB, making it nearly inaudible.
Gentle slopes (6 or 12 dB) sound more natural and work well on vocals and acoustic instruments. Steep slopes (24 or 48 dB) give surgical precision, which is useful when you need a hard boundary between wanted and unwanted frequencies.
Cleaning Up Vocals in a Mix
One of the most common uses for a high pass filter is on vocal tracks. Frequencies below about 150 Hz rarely contribute to vocal clarity or intelligibility. Instead, they add muddiness that competes with the kick drum and bass in a mix. Nearly every professional mixing engineer puts a high pass filter on vocals as a default starting point.
The standard technique is to set the filter around 80 Hz and slowly sweep the cutoff upward. At some point, the vocal will start to sound thin, which means you’ve begun cutting into frequencies the voice actually needs. Back off slightly from that point, and you’ve found your sweet spot. For male vocals, this threshold typically lands between 80 and 150 Hz. Female vocals can often tolerate a slightly higher cutoff since the fundamental frequencies of a female voice sit higher. A gentle 6 or 12 dB slope usually works best here, keeping the transition natural so the vocal doesn’t sound hollowed out.
Filtering Other Instruments
Vocals aren’t the only tracks that benefit. In a full mix, stacking high pass filters across multiple tracks is one of the most effective ways to create clarity and free up headroom in the low end. Some common starting points:
- Acoustic kick drum: 20 to 50 Hz, removing sub-bass rumble while keeping the punch intact.
- 808 kick: 20 to 40 Hz with a steep 24 dB per octave slope to preserve the deep tone but eliminate subsonic energy that eats headroom.
- Cymbals and overheads: 100 to 200 Hz. Cymbals produce very little useful content in the low end, and any that bleeds through is just kick and snare leakage muddying the mix.
These are starting points, not rules. The right cutoff always depends on the specific recording, the arrangement, and how instruments interact with each other. Trust what you hear over any preset number.
Removing Wind Noise and Handling Rumble
If you’ve ever recorded audio outdoors, you know the low-frequency rumble that wind creates. It’s often well below the range of speech or music, sitting under 100 Hz, but it’s loud enough to dominate the waveform and distort everything above it. The same problem shows up with handling noise from microphone bumps and vibrations transmitted through a mic stand.
A high pass filter is the first line of defense. In video editing software like Premiere Pro, the built-in high pass filter acts as a low-cut tool that targets exactly this kind of rumble. Setting the cutoff below 200 Hz is generally safe for dialogue and won’t thin out the human voice noticeably. For pure wind rumble without much speech content to protect, you can push higher. Many field recorders and shotgun microphones have a hardware high pass switch (often labeled “low cut”) specifically for this purpose, typically set around 75 to 80 Hz.
Protecting Speakers From Subsonic Damage
Subwoofers and speakers face a real mechanical risk from frequencies below the audible range. Signals under 20 Hz, including DC offset (a constant voltage that shouldn’t be in an audio signal at all), cause the speaker cone to make large, slow excursions that generate no useful sound but stress the driver. Over time, this leads to overheating and mechanical failure.
A high pass filter set around 25 to 30 Hz before the signal reaches a subwoofer removes this subsonic content. In live sound and DJ setups, this is sometimes called a “subsonic filter” or “infrasonic filter,” but it’s just a high pass filter with a very low cutoff. DC-blocking filters in amplifiers serve the same function, stripping out DC offset and extremely low signals that would otherwise push the speaker cone to its physical limits without producing audible bass.
Sharpening Images in Photo Editing
High pass filtering isn’t limited to audio. In image processing, the same concept applies to spatial frequencies instead of sound frequencies. Low spatial frequencies are the broad, gradual tonal changes in an image (the overall lighting, large color gradients). High spatial frequencies are the fine details: edges, textures, and sharp transitions.
Photoshop’s High Pass filter works by taking the original image, applying a Gaussian blur to create a low-frequency version, then subtracting that blurred version from the original. What’s left is purely the high-frequency detail: edges and fine texture against a neutral gray background. When you blend this high pass result back onto the original image using an overlay or soft light blending mode, those edges get enhanced, producing a controlled sharpening effect. This is actually the same math behind the Unsharp Mask tool, which adds a scaled version of the high pass result back to the original. The High Pass method simply gives you more visual control over what’s being sharpened before you commit.
Medical Monitoring and Sensor Data
Heart monitors (ECG machines) pick up electrical signals from the heart that fall roughly between 0.1 and 100 Hz. The problem is that breathing and body movement create low-frequency noise called baseline wander, which causes the signal’s baseline to drift up and down, making it harder to read the heart’s actual rhythm.
A high pass filter set at around 0.5 Hz removes this drift while preserving the heart signal above it. The cutoff has to be very low because the diagnostically important parts of the ECG waveform, particularly the segment that doctors examine for signs of a heart attack, contain low-frequency components that would be distorted by a more aggressive filter. This is a case where the filter slope and cutoff need to be chosen with extreme precision, since cutting too aggressively could mask a life-threatening finding. More advanced approaches use adaptive algorithms rather than simple fixed filters, but the underlying principle is the same: strip out the slow drift, keep the heartbeat.
Choosing the Right Cutoff and Slope
The two decisions you always face with a high pass filter are where to set the cutoff frequency and how steep to make the slope. The cutoff determines what gets removed. The slope determines how cleanly the filter separates wanted from unwanted content.
For audio mixing, start low and sweep upward until you hear the signal start to thin, then back off. This works on any source. For speaker protection, set the cutoff just below the lowest frequency your system can actually reproduce. For noise removal in field recordings, keep the cutoff under 200 Hz when preserving dialogue. For image sharpening, the radius parameter in the High Pass filter controls how much detail is captured: a small radius catches only fine edges, while a larger radius includes broader tonal transitions.
In every case, the goal is the same: remove low-frequency content that adds nothing useful, while leaving everything above the cutoff intact. When in doubt, use a gentler slope and a lower cutoff. You can always filter more aggressively later, but you can’t restore detail that’s already been cut.