- Author Profile John Moxey
- John Moxey Website
What Is an Audio Filter?
Before getting into specific filter types, it’s worth being clear about what a filter actually is and what it does.
You’ve probably come across the word “filter” in other contexts. A water filter removes unwanted particles from water passing through it. An image filter in Photoshop manipulates the colour and tonal composition of a photo. The core idea – selectively passing or removing something – applies to audio too.
An audio filter is a frequency-dependent circuit that processes a signal by boosting, attenuating, or passing specific frequency ranges within that signal. The audio frequency range runs from 0 Hz up to around 20 kHz, which roughly corresponds to the limits of human hearing. Everything that happens inside a filter happens within that range.
What makes audio filters slightly different from some other filter types is that they can do more than just remove content. As well as cutting frequencies, they can boost them. That makes them genuinely flexible shaping tools rather than just cleanup devices.
In practical terms, filters are used constantly in music production – in equalisers, synthesisers, guitar amplifiers, effects units, and mastering chains. Every time you roll off the low end of a vocal, brighten a snare, or sweep a synth patch, you’re using a filter. Understanding how they work means you’re making intentional decisions rather than just turning knobs until something sounds better.
Filter Order And Slope
One of the things that often goes unexplained in introductory articles is why filters cut the way they do – gradually rather than instantly. This is worth understanding before looking at individual filter types, because the concept applies to all of them.
In an ideal world, a filter would pass all frequencies on one side of a threshold and completely eliminate everything on the other. In reality, the gain changes gradually as you move away from the cutoff frequency. This gradual curve is described as the filter’s slope, and it’s measured in decibels per octave (dB/oct).
The slope is determined by the filter’s order – a value that describes the complexity of the underlying circuit or algorithm. The relationship is straightforward: each additional filter order adds 6 dB/oct of slope. So:
- A 1st-order filter has a slope of −6 dB/oct
- A 2nd-order filter has a slope of −12 dB/oct
- A 3rd-order filter has a slope of −18 dB/oct
- A 4th-order filter has a slope of −24 dB/oct
In a parametric EQ, you’ll typically see these expressed as −6, −12, −18, or −24 dB/oct settings on the high-pass and low-pass filters. Steeper slopes give you more aggressive, surgical cuts; gentler slopes sound more natural and musical.
The – 3 dB Point
Because filter circuits aren’t ideal, the gain doesn’t snap from 0 dB to full attenuation the moment a signal crosses the cutoff frequency. The cutoff (or threshold) frequency you dial in is the point at which the filter has already reduced the signal by 3 dB – sometimes called the – 3 dB point or half-power point. The actual attenuation continues to increase beyond that frequency according to the slope.
This is worth knowing because it means a filter is already working before it reaches the frequency you’ve set. If you set a high-pass filter to 80 Hz, it’s already cutting at 80 Hz – that’s where the gain is 3 dB down, not where cutting begins. The actual roll-off starts above that point and steepens as you go lower.
The cut-off isn’t where cutting starts
Phase Shift
Any filter introduces some degree of phase shift into the signal – a time offset between different frequencies that results from the way the filter affects the signal’s waveform. For most mixing applications, this isn’t a significant problem, but it becomes relevant in a few specific situations.
When you’re running parallel processing – for example, blending a filtered and unfiltered version of the same signal – phase differences between the two paths can cause comb filtering, where certain frequencies cancel out, and others reinforce. Similarly, using steep high-pass filters on multiple elements of a mix can shift the phase relationships between sounds enough to affect how they sit together.
Linear phase filter designs are specifically engineered to avoid this by applying equal delay across all frequencies. The trade-off is that they introduce pre-ringing – a slight artefact before a transient – and additional processing latency. Most DAW EQ plug-ins give you the choice between minimum phase (the default) and linear phase modes. For mixing work, minimum phase is usually the better choice. Linear phase makes more sense in mastering contexts or where phase coherence between parallel signals is critical.
High and Low Pass Filters
High-pass and low-pass filters are the most commonly used filter types in music production. They work by progressively cutting everything above or below a chosen frequency.
High Pass Filter (HPF)
A high-pass filter passes the frequencies above a set cutoff frequency and attenuates the frequencies below it. The name describes what the filter lets through – the highs pass; the lows are cut. You’ll also see it labelled as a low-cut filter, which describes what it removes.
In practical use, the HPF is one of the most useful tools in a mix. Rolling off the low end of instruments that don’t need it – acoustic guitars, room mics, overheads, even some synth pads – cleans up the low-frequency buildup that muddies a mix and makes space for the kick drum and bass to do their job without fighting everything else. On a vocal, a high-pass filter set around 80-100 Hz removes handling noise, proximity effect from close-miked performances, and low-frequency rumble without touching any of the musical content.
The slope matters here. A gentle -6 dB/oct roll-off sounds natural and transparent; useful when you want to tighten a source without it sounding filtered. Steeper slopes (-18 or -24 dB/oct) are more obvious and deliberate – good for problem-solving but easy to overdo.
One thing to watch: some HPF circuits introduce a small resonant peak just above the cutoff point, particularly at steeper slopes. In a synthesiser filter, this is an expressive feature. In an EQ, it can add an undesirable brightness or emphasis at the cutoff if you’re not careful.
The HPF you don’t notice is doing its job
Low Pass Filter
A low-pass filter is the mirror image of the HPF. It passes the frequencies below the cutoff and attenuates everything above. Also called a high-cut filter.
In mixing, the LPF is used less aggressively than the HPF but is just as useful. On a mix bus or individual tracks, rolling off extreme high frequencies can reduce harshness and digital grittiness. On a send going to a reverb or delay, a low-pass filter on the return tucks the effect further back in the mix, so it adds depth without cluttering the highs. In sound design and synthesis, the low-pass filter is the workhorse – sweeping a LPF over a sawtooth wave is the basis of most classic analogue synth sounds.
Shelving Filters
A shelving filter applies a fixed amount of boost or cut to all frequencies above or below a threshold frequency. Unlike a high-pass or low-pass filter, it doesn’t eliminate those frequencies – it raises or lowers them to a plateau and holds them there.
Think of it like a mixing desk fader for a specific frequency range. Boost the high shelf, and everything above the shelf frequency gets louder by however many dB you’ve dialled in. Cut the low shelf, and everything below the shelf frequency gets quieter by that amount. The transition between the unaffected region and the shelf is gradual rather than instant, which is what gives shelving filters their musical, broad-strokes character.
High Shelf Filter
A high-shelf filter affects all frequencies above the shelf frequency. Boosting adds brightness and air, which is useful for bringing out the upper harmonics of an acoustic instrument or lifting a vocal’s presence without surgical EQ. Cutting reduces harshness, controls sibilance in a broad way, or pulls back an instrument that’s competing in the top end.
The shelf frequency is typically set somewhere in the upper-mid to high range – 8 kHz or above for air and shimmer, lower for wider tonal adjustments. Because a shelf affects everything above the threshold equally, it’s a broad tool. It’s not the right choice for fixing a specific resonant frequency – that’s what bell filters are for – but it’s ideal for overall tonal balance.
Low Shelf Filter
A low-shelf filter affects all frequencies below the shelf frequency. Boosting adds body, warmth, and weight. Cutting thins out the low end, which is useful when an instrument sounds boxy or muddy, or when you need to make room for other elements without rolling frequencies off completely the way a high-pass filter would.
The distinction from an HPF is worth spelling out: a high-pass filter removes low frequencies progressively from the cutoff downwards, with full attenuation well below the threshold. A low shelf reduces (or boosts) low frequencies to a fixed level – they’re still there, just repositioned in level. If the low end is useful but too prominent, a low shelf is the more controlled approach.
Bell Filters (Peak Filters)
A bell filter, also called a peak filter or peaking EQ, boosts or cuts a range of frequencies centred around a specified point. The resulting shape on a frequency response graph looks like a bell curve, which is where the name comes from.
Bell filters are controlled by three parameters:
Centre frequency: the frequency at the peak of the boost or dip.
Gain: how much boost or cut is applied, measured in dB. Positive values boost; negative values cut.
Q (Quality factor): determines the width of the bell curve. A low Q gives a broad, wide curve that affects a large range of frequencies around the centre. A high Q gives a narrow, precise curve that affects only a small band of frequencies. In some EQ plug-ins, you’ll see bandwidth expressed in octaves rather than Q, but they’re describing the same thing.
The versatility of bell filters makes them probably the most used filter type in day-to-day EQ work. A broad, gentle boost (low Q, modest gain) can add warmth, presence, or air across a natural-sounding frequency range. A narrow, steep cut (high Q, significant negative gain) is the tool for fixing a specific resonant problem – a honky boxiness in a guitar recording, a ring in a snare drum, a harsh frequency in a vocal.
In practice, cuts and boosts require different Q settings to sound musical. When cutting, a relatively high Q is generally appropriate – you’re targeting a specific problem and don’t want to pull down more than necessary. When boosting, a lower Q usually sounds more natural, as a very narrow boost sounds phasey and artificial. A useful rule of thumb: cut narrow, boost wide.
Cut narrow, boost wide – until it isn’t
Band-Pass Filters
A band-pass filter does the opposite of what the name might initially suggest to a newcomer – rather than passing a band of frequencies and leaving everything else alone in the usual EQ sense, it passes only the frequencies within a chosen band and cuts everything outside it. Both above and below the passband are attenuated.
You can think of a band-pass filter as a high-pass filter and a low-pass filter placed in series – the HPF removes everything below the band, the LPF removes everything above, and what’s left is the band in the middle.
The width of the passband is controlled by Q, just as with a bell filter. A high Q gives a narrow passband; a low Q gives a wide one. When Q is high enough, the result is essentially a highly selective filter that isolates a very specific frequency range.
Band-pass filters are less common in standard mix EQ work, but they have clear applications. The classic telephone effect – that thin, mid-frequency sound of a voice through a phone – is created by a band-pass filter centred somewhere around 1-3 kHz, removing the lows and highs completely. Band-pass filters also appear in crossover networks, where different frequency bands are sent to different speakers, and in synthesiser architectures. In mixing, they can also be useful on reverb sends to limit the frequency content feeding the reverb and control how it sits in the mix.
Band-Stop Filter
A band-stop filter is the inverse of a band-pass filter. Where a band-pass filter passes only the frequencies within a chosen band and cuts everything else, a band-stop filter cuts the frequencies within a chosen band and passes everything outside it.
The band that’s attenuated is called the stopband. As with band-pass filters, Q determines its width – a high Q produces a narrow stopband, a low Q produces a wide one. A narrow-Q band-stop filter targeted at a very specific frequency is what most engineers know as a notch filter, which is covered in the next section. A wide-Q band-stop filter is used when you need to reduce a broader range of frequencies without completely removing them.
In mixing, a wide band-stop is occasionally useful when two instruments are competing in the same frequency area, and you want to carve out some space without resorting to a narrow surgical cut. In live sound, where feedback is a constant concern, band-stop filters help tame the frequencies most likely to ring out in a specific venue.
Notch Filter
A notch filter is a band-stop filter with a very high Q – a very narrow stopband. Rather than cutting a broad range of frequencies, it targets an extremely specific frequency and attenuates it steeply, leaving frequencies on either side largely unaffected.
The practical application of notch filters in music production is mostly corrective. The most common use is eliminating mains hum: 50 Hz in the UK and Europe, 60 Hz in the US and Canada. A precisely placed notch at exactly the right frequency removes the hum with minimal effect on anything else in the signal. The same approach works for electrical interference, ground loop noise, and specific resonances – a ringing snare drum, for example, often responds well to a precise notch at the problem frequency rather than a broader bell-curve cut that takes too much else with it.
Notch filters are also used in feedback rejection on stage, where the system’s EQ is tuned by placing notches at the frequencies most likely to ring in that room.
In audio engineering texts, notch filters go by a number of names: T-notch filter, band-limit filter, band-elimination filter, and band-reject filter are all in circulation. They all describe the same thing.
One thing to bear in mind with notch filters: because the Q is very high and the attenuation can be steep, they introduce a more pronounced phase shift than broader filter types. In most corrective applications, this doesn’t matter. Where it might – for example, if you’re notching a frequency in a signal that’s also being processed in parallel – it’s worth being aware of.
Hunting hum by ear
How Filter Types Relate to Each Other
It’s useful to see these filters not just as individual tools but as a connected system. A few relationships are worth knowing:
A high-pass + low-pass combination in series produces a band-pass effect – only the frequencies between the two cutoff points survive.
A band-stop filter is the inverse of a band-pass – it removes what the band-pass would pass, and passes what the band-pass would remove.
A notch is simply a band-stop filter taken to a high Q extreme.
A shelving filter is related to the HPF and LPF, but instead of progressively attenuating to silence, it attenuates to a plateau – a fixed gain level rather than full elimination.
Understanding these relationships helps when you’re reaching for a tool and need to decide between options. If you’re using a high-pass filter but want to preserve some of the low-end character rather than remove it, a low shelf cut might be the better choice. If you need to isolate a frequency band completely for an effect, a band-pass makes more sense than trying to approximate it with two separate EQ moves.
Filters in EQ vs. Filters in Synthesisers
Throughout this article, the focus has been on filters as used in EQ, which is where most recording engineers and producers encounter them most often. It’s worth acknowledging briefly that the same filter types appear in synthesisers, sometimes with different names and with controls designed around musical expressiveness rather than precision correction.
In a synthesiser, the filter (usually a low-pass filter, though many synths offer multiple modes) is a core part of sound design rather than a corrective tool. The key difference is resonance – also called Q or emphasis – which in a synth filter can be pushed to the point of self-oscillation, where the filter itself generates a pitched tone at the cutoff frequency. On an EQ, excessive Q in a bell filter sounds phasey and unmusical; on a synthesiser, that same resonance peak is an expressive feature.
Many classic synthesiser filters – the Moog ladder filter, the Roland 303’s diode ladder, the Oberheim SEM state-variable filter – have distinctive sonic characters that come from the specific design of their filter circuits. These aren’t perfectly clean or textbook-accurate. Their imperfections are part of the sound, which is why software emulations of those circuits continue to be popular despite the availability of technically superior designs.
When you’re using a synthesiser and sweeping a filter or adjusting resonance, you’re working with the same underlying principles as EQ – cutoff frequency, slope, Q – just in a context where those parameters are being used for creative expression rather than spectral management.
Same knob, different job
Quick Reference
| Filter Type | What It Does | Typical Use |
|---|---|---|
| High-Pass (HPF) | Cuts below the cutoff frequency | Removing low-end rumble, clearing mix space for bass |
| Low-Pass (LPF) | Cuts above the cutoff frequency | Removing hiss and high-frequency noise; taming harshness |
| High Shelf | Boosts or cuts all frequencies above threshold | Adding air; reducing harshness broadly |
| Low Shelf | Boosts or cuts all frequencies below threshold | Adding body; controlling low-end without a hard cut |
| Bell / Peak | Boosts or cuts around a centre frequency | Corrective cuts; broad tonal shaping |
| Band-Pass | Passes only frequencies within a band | Telephone effects; creative sound design; crossovers |
| Band-Stop | Cuts frequencies within a band | Reducing competing frequencies; feedback control |
| Notch | Narrow band-stop with very high Q | Removing mains hum; eliminating resonant rings |
Discuss this article in our Music Forum.
Related Articles
Do you want to read more gear reviews? If so, you can find articles and tutorials on our Music Product Reviews page.
You might find the following selection of recording and music production articles by John Moxey useful:
