DeHum

About DeHum

DeHum 3 is designed to remove hum and buzz typically introduced by poorly grounded electrical equipment, but also other tonal noise sources like electrical motor noise. The dehum algorithm also supports adaptive hum reduction so that the algorithm adapts to fluctuations in the fundamental frequency of the hum signal.

Real life hum noise is likely to consist of a fundamental frequency and a set of harmonic frequencies. These are multiples of the fundamental frequency. DeHum allows you to set the number of harmonics to remove and also has the option to address only odd harmonics, since hum noise with only odd harmonics are frequently encountered.

There are three different operating modes available, explained below. By default and for most use cases, the Automatic option is a good starting point.

The three modes are mutually exclusive and are selected with the Automatic, Precision and Notch buttons. Automatic mode uses a machine learning model to detect and reduce the hum, so there is no fundamental frequency for you to set. In Precision mode, DeHum subtracts a hum signal reconstructed using a sinusoidal re-synthesis technique in order to minimize distortions of the wanted signal, and in Notch mode the hum reduction is performed using conventional notch filters instead.

Settings

The below settings are common in all modes:

• Operation Modes: Automatic, Precision, and Notch

• DeHum 3 provides three operating modes:

  • Automatic
    • In the new Automatic mode, DeHum 3 uses machine learning to automatically interpret and remove hum in the audio signal.
  • Precision
    • Precision mode allows the user to specify a fundamental frequency for the hum to be removed, and operates like the default mode in DeHum 2
  • Notch
    • Notch mode, hum is removed via conventional notch filters. This mode usually reduces hum effectively but is prone to artifacts. The CPU usage is lower in this mode, and it introduces no latency.

• Noise Floor (dB)

This slider, on the left side of the interface, sets the maximum amount of attenuation that DeHum 3 is able to use. If you hear artifacts after removing hum, try reducing the maximum amount of attenuation by raising this slider until artifacts are no longer audible.

• M/S mode

The Mid / Side (M/S) mode is only available when working with source material in stereo and it avoids drifting in the stereo image after processing. When enabled, the incoming audio is converted to a M/S signal prior to processing and decoded back to normal left / right stereo at the output.

The Mid / Side button is found along the bottom of the visualization graph in each mode.

• Solo Hum

Click this button if you wish to monitor the signal removed by the hum reduction algorithm.

The Solo Hum button is found along the bottom of the visualization graph in each mode.

Automatic Mode settings

• Reduction Curve

The Automatic mode provides a reduction curve with low shelf, mid bell, and high shelf filters available to shape the sensitivity of the reduction algorithm. The filters can be toggled by clicking the filter buttons in the bottom left of the Reduction Curve graph, beside the Mid Side button. In the below image, a mid bell filter is active.

We can also see a green circle at the peak of the bell filter, with corresponding arrows and handles at each side. Left click and drag the circle to change the frequency and gain of the filter. Click and drag the arrows to adjust the Q width of the filter. Click and drag the handles to change the slope shape of the filter.

Precision Mode settings

• Frequency (Hz)

The Frequency knob sets the fundamental frequency of the hum or buzz noise. If the hum originates from the power distribution net, the fundamental frequency should be set to either 60 Hz (American standard) or 50 Hz (European standard), depending on frequency of the AC power distribution in the country the recording was made. You can let DeHum 3 fine-tune this fundamental frequency automatically using the Scan button, indicated below with a red box.

• Sensitivity (%)

The sensitivity parameter is available in Precision mode only. Higher sensitivity values cause the hum reduction algorithm to classify more frequency components as hum.

• Adaptivity (Hz / s)

The adaptivity knob controls the maximum fluctuation of the fundamental frequency in number of Hertz per second that is allowed in the detection of the fundamental frequency. This value should be as low as possible while still detecting the fluctuations of the hum signal being removed.

• Harmonics

The number of harmonics to detect and remove. This should be as low as possible while still removing all the harmonics present in the hum noise.

• Odd harmonics

Click this toggle button to reduce only odd harmonics. Hum noise consisting of a fundamental frequency with only odd harmonics are frequently encountered in real life situations and are typically the result of sine wave signal with a symmetrical non-linear distortion.

Notch Mode settings

• Frequency (Hz)

The Frequency knob sets the fundamental frequency of the hum or buzz noise. If the hum originates from the power distribution net, the fundamental frequency should be set to either 60 Hz (American standard) or 50 Hz (European standard), depending on frequency of the AC power distribution in the country the recording was made. You can let DeHum 3 fine-tune this fundamental frequency automatically using the Scan button, indicated below with a red box.

• Adaptivity (Hz / s)

The adaptivity knob controls the maximum fluctuation of the fundamental frequency in number of Hertz per second that is allowed in the detection of the fundamental frequency. This value should be as low as possible while still detecting the fluctuations of the hum signal being removed.

• Harmonics

The number of harmonics to detect and remove. This should be as low as possible while still removing all the harmonics present in the hum noise.

• Odd harmonics

Click this toggle button to reduce only odd harmonics. Hum noise consisting of a fundamental frequency with only odd harmonics are frequently encountered in real life situations and are typically the result of sine wave signal with a symmetrical non-linear distortion.