Is the Click Removal effect a sharp-cornered high-pass filter that eliminates all sharp edges? If so, it might roll off the fast rising edges of everything (such as cymbal crashes) so maybe it would be best to manually select a window around each click before using Click Removal.
Or does it specificly look for isolated high-frequency pulses? In this case the entire track can be selected before removal.
Hopefully it is the latter because that would make the process much easier.
One trick to avoid false-positives is to reverse the audio before processing for click removal and then reverse it back again afterwards.
Personally I use Brian Davies’ excellent ClickRepair software fo my click removal (it costs a little but is well wort it IMO). It has a setting to do the “reverse processing” automatically. See this sticky thread: https://forum.audacityteam.org/t/click-pop-removal-clickrepair-software/1933/1
WC
What does reversing the data stream do that helps click removal? I expanded around several clicks and some have sharp leading edges and some have sharp trailing edges.
Does anyone have any thoughts about my other questions? Is click removal a high-pass filter or does it specifically look for short pulses? Is it better to use click removal on the entire recording or manually search for clicks and then select a small area around the click for removal?
Also, are the default settings of 200 (threshhold) and 20 (max spike width) good choices? What are the units for these values (dB, usec, ???)
It helps to remove false detections.
“Natural” or “musical” sounds may have an abrupt attack (for example a snare drum hit) but are very unlikely to have such an abrupt decay as there is always a tendency for mechanical sounds to ring. Thus, when reversed, natural sounds have a slow attack, and so are ignored by the click detection.
It is not just a high pass filter. It looks for sudden peaks in the context of the surrounding audio. It does not work well if the selection is too short.
I think so, though I am not an expert on this effect.
I don’t think there are any units.
This is probably not the best Click Removal effect in the world, but having worked on developing other click removal effects I am not being critical. Click removal (without damaging sounds that are supposed to be there) is extraordinarily difficult to do. The Audacity Click Removal effect tends to err on the safe side, that is, it is more likely to miss (not correct) clicks than to damage the audio that is supposed to be there.
Now that I have a bit more experience there is one last question concerning click removal. I have been manually locating clicks, selecting a 0.1sec window around each click to minimize collatoral damage to the surrounding audio and then removing the clicks. This has worked well but is very time-consuming. I have not noticed any change in the audio surrounding the click when looking at the waveform close-up implying that the Audacity click removal tool works very well.
If the audio is reversed, click removal applied to the entire track and then un-reversed will that remove the clicks and also preserve the integrity of the audio or is my manual method a better choice? The full-track method is surely be much easier.
You can detect clicks visually using the spectrogram display …
http://wiki.audacityteam.org/wiki/Click_removal_using_the_Spectrogram_view
And can fix them with “Repair” in the “Effect” menu.
I don’t think that reversing the audio is actually necessary as there are probably not many (if any) false detections.
Do ensure that the track is 32 bit float format before using Click Removal (or any other processing).
You can check to find exactly what is removed by Click Removal using the following steps:
The new track will contain the difference between the processed original track and the copy, so this shows exactly what has changed. The track should be silent except for where clicks have been removed.
I just tried Steve’s suggestion of comparing the pre and post click removed tracks from an old Glenn Miller LP. The difference channel was predominantly flat (as expected) but there were areas with large differences. When listening to these sections of pre and post click removal there was a very noticable difference. One section included hand clapping and expanding the original waveform showed the hand claps to look exactly like clicks so they weres removed (or at least very muted).
It appears that depending on the material selecting the entire track and applying click removal may not be a good idea. For this type of material manually listening for clicks and then selecting a small window around them (approx 0.1sec) for click removal might be better.
I’ve not used Click Removal very much, but it seems to only attack clicks when they are a lot louder than the average level in that part of the track. So for music that is not very “percussive” it is probably safe to run Click repair on the whole track, but if there are sharp percussive sounds in the music it will be better to tackle them more manually. Anyway, you now know a way to compare the “before and after”.