The BAX EQ hardware measures flat within 0.2 dB from 1 Hz all the way up to 80 kHz, which is precisely why it needs cut filters. Wide bandwidth is a design decision, not an accident. It also means the box passes whatever you feed it, including a lot of material that is not music.
The unit answers that with seven switched positions on each filter, mounted on relays rather than a pot. Those switches are not a convenience feature.
Most people treat band-edge filtering as housekeeping. Clear the rumble, tidy the top, move on. The actual payoffs are larger and stranger than that.
That has a direct cost. Your limiter pulls gain down in response to energy nobody can hear. Every decibel it spends there is a decibel your snare does not get. Subsonic energy your monitors cannot reproduce is still competing for the loudest part of your record.
It arrives from more places than you would guess. Traffic and air handling through the walls, footsteps, mic stand bumps, and plosives that survived the vocal chain all deposit something down there. So do room modes and the decay tails of synthesized sub bass.
Woofers pay too. Excursion spent on inaudible content is excursion unavailable for the bass you actually wrote. That is why a monitor can sound strained without ever sounding loud.
The hardware's low filter starts at 12 Hz, and the stated targets are infrasonic rumble and DC offset. Those two problems are related more closely than the spec sheet suggests.
The consequences are lopsided. Positive peaks clip before negative ones ever get close, which means you lose headroom on one side of the waveform only. Level meters read the offset as signal, so your numbers stop meaning what you think they mean.
Editing gets worse as well. Cut into a track with an offset and the discontinuity at the splice point produces an audible click, no matter how carefully you placed the edit. Sustained offset can also stress loudspeakers and headphones.
Spotting it takes seconds. Zoom in on a quiet passage and look at where the waveform rests, or compare your positive and negative peak readings, because a healthy signal keeps them roughly level.
The fix is trivial once you know it is there. Offset lives at 0 Hz, so any high pass with a turnover below 20 Hz removes it. Three to five hertz is the usual choice. Faulty converters, analog stages, and the occasional badly behaved plugin all produce it.
So the playback argument is weak. The production argument is not. Every nonlinear process in your chain generates harmonics, and harmonics of a 30 kHz component land far above the Nyquist frequency of your session.
Those products do not vanish. They fold back down into the audible band as inharmonic aliasing, unless the process oversamples heavily enough to catch them. Feed a saturator or a limiter a pile of ultrasonic content and some of the mess reappears in the midrange, where you will absolutely hear it.
Oversampling in the offending plugin is half the answer. A low-pass filter ahead of it is the other half, and it is why the phase-coherent cut filters carried into software were modeled as part of the circuit.
The sample rate decides how much you should care. At 44.1 or 48 kHz, your converter's own anti-alias filter already removed everything above roughly 22 kHz. The high cut has little left to do. At 88.2 and 96 kHz, it has plenty.
The unit answers that with seven switched positions on each filter, mounted on relays rather than a pot. Those switches are not a convenience feature.
Most people treat band-edge filtering as housekeeping. Clear the rumble, tidy the top, move on. The actual payoffs are larger and stranger than that.
Everything below 20 Hz still counts against your peaks
Human hearing gives out somewhere around 20 Hz, and most monitors give up well before that. Your meters do not. A 10 Hz component sitting at a modest level still contributes to sample peak and to true peak.That has a direct cost. Your limiter pulls gain down in response to energy nobody can hear. Every decibel it spends there is a decibel your snare does not get. Subsonic energy your monitors cannot reproduce is still competing for the loudest part of your record.
It arrives from more places than you would guess. Traffic and air handling through the walls, footsteps, mic stand bumps, and plosives that survived the vocal chain all deposit something down there. So do room modes and the decay tails of synthesized sub bass.
Woofers pay too. Excursion spent on inaudible content is excursion unavailable for the bass you actually wrote. That is why a monitor can sound strained without ever sounding loud.
The hardware's low filter starts at 12 Hz, and the stated targets are infrasonic rumble and DC offset. Those two problems are related more closely than the spec sheet suggests.
DC offset clips your positive peaks first
DC offset is the same problem with the frequency taken all the way down to zero. The waveform sits off center, so instead of swinging from plus three volts to minus three, it might run from plus four to minus two.The consequences are lopsided. Positive peaks clip before negative ones ever get close, which means you lose headroom on one side of the waveform only. Level meters read the offset as signal, so your numbers stop meaning what you think they mean.
Editing gets worse as well. Cut into a track with an offset and the discontinuity at the splice point produces an audible click, no matter how carefully you placed the edit. Sustained offset can also stress loudspeakers and headphones.
Spotting it takes seconds. Zoom in on a quiet passage and look at where the waveform rests, or compare your positive and negative peak readings, because a healthy signal keeps them roughly level.
The fix is trivial once you know it is there. Offset lives at 0 Hz, so any high pass with a turnover below 20 Hz removes it. Three to five hertz is the usual choice. Faulty converters, analog stages, and the occasional badly behaved plugin all produce it.
Ultrasonic content matters where things go nonlinear
Be careful with the received wisdom here. Peer-reviewed listening tests found that intermodulation from ultrasonic signals is measurable on real loudspeakers. It only became just audible when nothing below 20 kHz was playing at all, and the authors concluded it is not pertinent to playback of current high-resolution recordings.So the playback argument is weak. The production argument is not. Every nonlinear process in your chain generates harmonics, and harmonics of a 30 kHz component land far above the Nyquist frequency of your session.
Those products do not vanish. They fold back down into the audible band as inharmonic aliasing, unless the process oversamples heavily enough to catch them. Feed a saturator or a limiter a pile of ultrasonic content and some of the mess reappears in the midrange, where you will absolutely hear it.
Oversampling in the offending plugin is half the answer. A low-pass filter ahead of it is the other half, and it is why the phase-coherent cut filters carried into software were modeled as part of the circuit.
The sample rate decides how much you should care. At 44.1 or 48 kHz, your converter's own anti-alias filter already removed everything above roughly 22 kHz. The high cut has little left to do. At 88.2 and 96 kHz, it has plenty.
