In this lesson
Filters are shape machines
The slope of a filter decides how fast it cuts, and a normal EQ filter also shifts timing near its cutoff. Here is when that matters and when it does not.
Updated · 7 min read · Lesson 7 of 14 in Audio Science
In short
- Each pole of a filter adds 6 dB per octave of slope, so a high-pass setting is only complete with its slope.
- A minimum-phase filter also shifts the phase of frequencies well above its cutoff, where it no longer cuts at all.
- That shift matters when filtered and unfiltered versions of one sound are summed, so filter related tracks alike.
In this lesson (8)
The high-pass on every track
A common piece of mixing advice is to clean up the low end by high-passing everything except the kick and the bass. So you put a steep 24 or 48 dB per octave at 120 Hz on the vocal, the guitars and the synths. Each track looks tidy on the analyzer. Played together, the mix sounds thinner than before, and the drums feel less solid.
Some of those tracks had useful energy below 120 Hz, and the filters removed it. Every one of those filters also changed the timing of the frequencies around its cutoff, which matters as soon as filtered and unfiltered versions of a related sound meet.
Why it matters
Slope and cutoff decide what survives
A high-pass filter lets everything above its cutoff through and turns down what is below. The cutoff is usually the point where the level is 3 dB down. The slope says how fast the level falls after that, in decibels per octave. Each pole of the filter adds 6 dB per octave: a first-order filter falls 6 dB per octave, a second-order 12, a fourth-order 24 and an eighth-order 48.
So "a high-pass at 120 Hz" is only half a setting. At 12 dB per octave, 60 Hz is still there, 12 dB down. At 48 dB per octave it is effectively gone. On a vocal, a gentle filter set below the lowest notes of the part removes rumble and keeps the body. A steep one at the same frequency removes more than you meant to.
Listen
Shape a sound with a filter
Chords and soft noise, or a real mix, through one filter, with the spectrum drawn live. Sweep the cutoff and raise the resonance.
Start with your speakers or headphones low.
The science
Why ordinary filters move time
A normal analog or digital filter is minimum . In that kind of filter, changing the level of a band also changes its phase, and the two cannot be separated. Each pole contributes up to 90 degrees of phase shift. At the cutoff, a second-order high-pass has shifted the phase by 90 degrees and a fourth-order one by 180.
A phase shift that changes with frequency delays different frequencies by different amounts. The useful measure is , the delay of the energy around each frequency:
Here is the phase in radians. Where the phase curve bends, around the cutoff, the group delay rises. A fourth-order Butterworth high-pass at 40 Hz delays the content near 40 Hz by about 15 ms. At 80 Hz the delay is about 3 ms, and at 160 Hz it is under 1 ms. Steeper filters bend the phase harder and delay more.
On a single track this is usually hard to hear: the lowest part of a kick or bass gets slightly softer and longer. The trouble starts when the filtered track is added to something that carries the same sound without that filter: the inside and outside kick mics, a bass DI and its amp, a parallel bus filtered differently from the dry tracks. Near the cutoff the two versions no longer line up, and they partly cancel.
A filter avoids this by delaying every frequency by the same amount, so the waveform keeps its shape. The price is and pre-ringing: a steep linear-phase filter spreads a little of each 's energy before the hit, which can sound like a soft swell into a drum.
Try it
Hear what a filter changes above its cutoff
0 of 6 steps done
The minimum-phase filter moves the low end in time well above its cutoff, and the null test makes that obvious. On its own the filtered kick sounds nearly the same. Summed with the original, the difference is loud.
Common mistake
Filtering related tracks differently
You filter one of two related signals and not the other. A high-pass on the kick inside mic but not the outside mic, on the bass amp but not the DI, or on a parallel drum bus but not the dry drums shifts the phase in one path only. The low end of the sum comes out thinner than either track alone, and the usual response is to boost it, which does not fix a cancellation. Filter related tracks with the same settings, or check the sum with the switch and keep whichever setting gives the fullest low end.
The opposite mistake is reaching for linear phase everywhere. It adds latency, and on drums a steep linear-phase filter in the low end can smear energy before the hit. Use it where the phase shift is a real problem, such as on a bus that is summed with a dry copy.
Takeaway
Set slope and cutoff on purpose
Treat the slope as part of the decision. For general cleanup, a 12 dB per octave high-pass set below the lowest note of the part removes rumble with little phase shift. Save 24 and 48 dB per octave for real problems such as stand rumble or mains hum. When a filtered track shares a sound with another track, filter both the same way, or check the sum in mono with the polarity switch before you call the low end finished.
High-pass filter
A filter that removes energy below a cutoff frequency and lets everything above it through.
Phase
Where a wave is in its cycle at a given moment. Two copies of a sound that are out of phase partly cancel where one rises while the other falls.
Group delay
How long a filter holds back each frequency, in milliseconds. Steep high-pass filters delay the lowest frequencies the most.
Linear phase
A filter that delays every frequency by the same amount, so the waveform shape is kept. The cost is latency and a little ringing before sharp hits.
Latency
The delay between doing something, like pressing a key, and hearing the result.
Transient
The short, sharp start of a sound, like the crack of a snare or a pick hitting a string. It carries a lot of how hard a sound feels and how quickly the ear recognizes it.
Polarity
Whether a waveform is upright or flipped. Flipping polarity inverts the wave at every frequency at once. It is not the same as a time delay.
3 questions
Check yourself
Sources
Show all 2Hide sources
- Smith, J. O. (2007). Introduction to Digital Filters with Audio Applications. CCRMA, Stanford University. ccrma.stanford.edu/~jos/filters/
- MIT OpenCourseWare. 6.003 Signals and Systems, Fall 2011. ocw.mit.edu/courses/6-003-signals-and-systems-fall-2011/
Next lesson · 8 of 14
Resonance can sing or wreck the mixResonances give instruments their voice and turn harsh only when they ring too long or too loud. Learn what Q means and how to cut only what hurts.Previous: How to read a spectrum analyzer

Written by Virzy Guns
Songwriter and producer, founder of Virzy Guns Production. Now building HealingWave.