Automatically learning noise and hiss remover.
Some radio stations have a problem with constant sounds, such as a 50/60 Hz hum from a bad cable (which can be hard to find), a high pitch tone from a fan or airconditioner, a constantly present hiss etc. PNR Noise & Hiss can learn what the disturbing sound sounds like, and then remove it.
For constant tones, the removal is nearly perfect, with nearly no side effects, even if a tone is as loud as the actual programming. For non-constant tones, including hiss, PNR can reduce it by a few dB without causing noticeable artifacts.
See this YouTube video for an example of how to set it up and what it does:
How it works:
The average value and sigma (variance) are used to determine how much more audio should be removed than the minimum. If this is increased too much, artifacts will become more audible.
Main PNR settings.
Enables PNR Noise & Hum removal.
Perform Dehummer after the Declipper.
If your material is clipped first and then noise/hum is added, for example if you play clipped CD's on a system that has a hum, the hum needs to be removed first, and declipper should be done afterwards. In that case, leave this setting off.
But, if for some reason the clipping happened after the noise/hum was added, then you can enable this setting to declip first, and remove the noise and hum afterwards.
- Minimum Multiplier
Multiplier for the minimum amount of audio measured.
For completely constant tones, using 100% here will completely remove the sound, lower values will always reduce less. Normally 100% is a good value here.
- AVG Multiplier
Multiplier for the average level that was measured.
- Sigma (variance multiplier)
Multiplier for the sigma (variance) level.
- Sigma Steepness
Steepness when going from Minimum to Sigma filtering behavior.
ANALYSIS STEP panel
The settings here are [b]only[/b] used during analysis!
Training parameters panel
Parameters that affect training - must be set before training panel
Dehummer filter strength panel
Dehummer active filtering settings panel
Minimum value panel
Average and Sigma values panel
- Ignore lowest measurements
Percentage of lowest measurements to ignore.
- Ignore highest measurements
Percentage of highest measurements to ignore.
- Train with current input
Enables collection of PNR data.
Enable this on moments of silence (with the disturbing sounds present), so the filter can learn what to remove. You might need to click Clear training data to remove data that was collected earlier. This step needs to be performed at the correct Sample rate, Processing latency, Link error '264' and Gain setting - when any of these settings are changed, the learned information becomes unusable, and the filter is disabled until either the settings are restored or a new learning stage has been performed.
- Train with current input, add to existing training data
- Clear training data
Clears all the learned information.
Always press this when a new Train with current input step will be performed with changed audio (different disturbing sounds).
Continuous Learning panel
Dynamically adjust the Dehummer behavior based on the input.
Note: This is not necessarily a good idea for good audio quality, it's more intended at forensics uses. For example, if you want to filter out the sound of a driving car, which slowly changes in sound, it can help to make the Dehummer learn continuously.
On music this will sound bad, so don't use it on that.
- Continuous learning (EXTREME CPU)
Enables continuous learning.
As said above, this should only be used for forensics purposes, not for radio.
- Learn from the past (EXTREME CPU LOAD)
Sets the time (history) to be considered for continuous learning.
Higher values can cause the CPU load to increase dramatically.