DeHum 3 in Acoustica 8 has three operating modes, and only Automatic uses the new machine-learning model to detect hum without a set fundamental frequency. Precision and Notch stay available because electrical noise is not one tidy problem.
The Acoustica 8 machine-learning restoration update makes Automatic the obvious first stop when a recording carries steady hum or buzz. Acon recommends it for most cases, but the other modes give you tighter control when the interference has a known frequency or the session cannot tolerate added latency.
Electrical hum often arrives as a fundamental plus a ladder of harmonics rather than one isolated tone. Grounding problems, power supplies, motors, and other equipment can leave different patterns behind, so choosing a mode by habit is less useful than choosing it by what the spectrum and your ears reveal.
You still control how the detector behaves. Automatic includes a Reduction Curve with low-shelf, mid-bell, and high-shelf filters, letting you alter sensitivity across the spectrum instead of accepting one global decision.
A bass-heavy recording can contain wanted energy close to the hum region, while a brighter electrical buzz may spread well above the fundamental through harmonics. Shaping reduction sensitivity lets you protect one area while asking the detector to work harder somewhere else.
Noise Floor sets the maximum attenuation available to the algorithm. If the cleaned signal starts sounding pinched, hollow, or unstable, raising it reduces how deeply DeHum can dig, which is often smarter than changing modes immediately.
The processing method is different from a bank of notches. Precision reconstructs the hum with sinusoidal resynthesis and subtracts it from the wanted signal, an approach designed to reduce distortion compared with simply cutting narrow frequency bands from everything passing through the processor.
Sensitivity controls how readily components are classified as hum. Adaptivity sets how much the detected fundamental may drift per second, while Harmonics tells DeHum how many multiples of that fundamental it should chase. You can also restrict processing to odd harmonics when the interference follows that pattern.
Research on adaptive cancellation of periodic interference helps explain why frequency tracking matters. Periodic electrical contamination can move slightly rather than staying nailed to one frequency, so a processor that follows it has an advantage over a rigid cut when the source is unstable.
Restraint matters here. Keep Adaptivity as low as possible while still following the drift, and remove only as many harmonics as the recording needs. Every extra target creates another opportunity to disturb wanted tonal material.
The controls remain familiar. You set or scan the fundamental, choose the permitted frequency drift, decide how many harmonics to remove, and can limit processing to odd harmonics. The removal method changes, not the diagnosis you performed beforehand.
Notch makes sense when latency matters more than surgical transparency, or when simple narrow cuts solve the hum cleanly. Precision is the better place to turn when those cuts start taking audible pieces of bass, voice, or sustained instruments with them.
Solo Hum gives all three modes a useful reality check because it lets you hear only what the processor removes. Wanted notes, vocal body, or stereo detail becoming obvious in that monitor signal means the repair is reaching beyond the defect.
Stereo material gets another safeguard through M/S mode. DeHum converts left and right into Mid and Side before processing, then decodes them afterward, specifically to reduce stereo-image drift during hum removal.
Automatic is the fast diagnostic choice, Precision is the controlled subtraction choice, and Notch is the low-latency filter choice. The right mode depends less on hum level than on whether its frequency is known, whether it drifts, how much wanted material shares its harmonics, and how much processing cost the session can tolerate.
The Acoustica 8 machine-learning restoration update makes Automatic the obvious first stop when a recording carries steady hum or buzz. Acon recommends it for most cases, but the other modes give you tighter control when the interference has a known frequency or the session cannot tolerate added latency.
Electrical hum often arrives as a fundamental plus a ladder of harmonics rather than one isolated tone. Grounding problems, power supplies, motors, and other equipment can leave different patterns behind, so choosing a mode by habit is less useful than choosing it by what the spectrum and your ears reveal.
Automatic mode handles the uncertain cases first
Automatic does not ask you to enter 50 Hz, 60 Hz, or another fundamental. Its machine-learning model interprets the signal and decides what belongs to the hum, which helps when the interference is easy to hear but awkward to measure or changes enough to make a fixed setting troublesome.You still control how the detector behaves. Automatic includes a Reduction Curve with low-shelf, mid-bell, and high-shelf filters, letting you alter sensitivity across the spectrum instead of accepting one global decision.
A bass-heavy recording can contain wanted energy close to the hum region, while a brighter electrical buzz may spread well above the fundamental through harmonics. Shaping reduction sensitivity lets you protect one area while asking the detector to work harder somewhere else.
Noise Floor sets the maximum attenuation available to the algorithm. If the cleaned signal starts sounding pinched, hollow, or unstable, raising it reduces how deeply DeHum can dig, which is often smarter than changing modes immediately.
Precision mode suits hum you can actually identify
Precision asks you to specify the fundamental frequency, either directly or with Scan. For ordinary mains-related interference, 50 Hz and 60 Hz are common starting points, but Scan can fine-tune the detected frequency instead of assuming the recording sits exactly on the nominal value.The processing method is different from a bank of notches. Precision reconstructs the hum with sinusoidal resynthesis and subtracts it from the wanted signal, an approach designed to reduce distortion compared with simply cutting narrow frequency bands from everything passing through the processor.
Sensitivity controls how readily components are classified as hum. Adaptivity sets how much the detected fundamental may drift per second, while Harmonics tells DeHum how many multiples of that fundamental it should chase. You can also restrict processing to odd harmonics when the interference follows that pattern.
Research on adaptive cancellation of periodic interference helps explain why frequency tracking matters. Periodic electrical contamination can move slightly rather than staying nailed to one frequency, so a processor that follows it has an advantage over a rigid cut when the source is unstable.
Restraint matters here. Keep Adaptivity as low as possible while still following the drift, and remove only as many harmonics as the recording needs. Every extra target creates another opportunity to disturb wanted tonal material.
Notch mode trades finesse for speed and zero latency
Notch mode uses conventional notch filters at the fundamental and selected harmonics. Acon describes it as effective but more prone to artifacts than Precision, while giving it two practical advantages for live or heavily loaded sessions. CPU use is lower, and the mode introduces no latency.The controls remain familiar. You set or scan the fundamental, choose the permitted frequency drift, decide how many harmonics to remove, and can limit processing to odd harmonics. The removal method changes, not the diagnosis you performed beforehand.
Notch makes sense when latency matters more than surgical transparency, or when simple narrow cuts solve the hum cleanly. Precision is the better place to turn when those cuts start taking audible pieces of bass, voice, or sustained instruments with them.
Solo Hum gives all three modes a useful reality check because it lets you hear only what the processor removes. Wanted notes, vocal body, or stereo detail becoming obvious in that monitor signal means the repair is reaching beyond the defect.
Stereo material gets another safeguard through M/S mode. DeHum converts left and right into Mid and Side before processing, then decodes them afterward, specifically to reduce stereo-image drift during hum removal.
Automatic is the fast diagnostic choice, Precision is the controlled subtraction choice, and Notch is the low-latency filter choice. The right mode depends less on hum level than on whether its frequency is known, whether it drifts, how much wanted material shares its harmonics, and how much processing cost the session can tolerate.
