Does “Normalizing” this waveform with the default settings (Normalize to 0 dB with DC offset correction) produce DC offset?
What is the “correct” way to Normalize this waveform to 0 dB?
HP filter with 1Hz cutoff, then use Normalize?
I think it reveals DC offset rather than producing it.
I’m guessing that the VLF component in the clip is confusing Normalize, because it sees it as a constantly changing DC offset.
POL
Tricky. As Irish noted, the original waveform has a VLF component that is all below zero. You can see this by applying a 10 Hz LPF to the track. Normalize with DC offset removal appears to shift that VLF component up so the average is zero. This is probably correct behaviour. DC offset is not being created, although it looks that way if you just examine the start of the waveform. As Irish also noted, this is a case where a HPF works better.
– Bill
Leaving aside the proposition that “Normalize” and “Amplify” should be one tool and DC Processing should be another, what is the goal? As I harp frequently, Audio doesn’t go down to DC and shows that do have no place in Audacity or any other sound program.
If you load a show with significant DC, a warning should immediately present “[X] Remove DC?”. Unless you are a professor experimenting with special low frequency effects, anything you do to that show except Remove DC will result in a severely damaged audio presentation.
And Remove DC – the separate tool – should include averaging to get rid of pops at the front and back and then HPF to get rid of subsonic damage during the show. Exactly what would happen if you ran your show through an audio transformer.
Koz
I don’t quite understand the “pops at front and back” comment. And if you’re going to run a HPF on the track then you don’t need the “other” DC removal, do you?
– Bill
This was a go-around much earlier when it was proposed that 10Hz High Pass Filter (for example) was good for everything. It’s not. That won’t remove the error in the first and last portion of the show, since at that exact instant, the “DC Level” has energy above 10 Hz and will produce a click or pop.
It was advanced that it should be possible to “fake out” the current DC Removal program, but nobody could come up with a way to do it outside of non-audio, artificial waveform constructs. Really reaching for it. For well into the high-99% of sound work presented, it works perfectly fine.
In the case of a badly damaged sound board or other broken equipment, you might need both. There were several postings of extraordinarily badly damaged sound clips which were ‘audio’ only because the poster insisted. They had wandering DC offset, and IMHO, if the show duration had been longer, would have failed completely.
As far as Normalize and Amplify being the same tool, I’m not making that up. That’s how Cool Edit does it.
[X] Link tracks?
Koz
The example in this topic is at zero (-inf dB) at the ends, but there is an offset in the middle portion.
The DC offset removal tool fails on this example because it moves the entire waveform so that the average positive and average negative are equal, which in this case moves the ends away from zero (thus causing a pop at beginning and end.
The high-pass filter method will fail to zero the beginning of an offset track, but in this case the beginning is already at zero, so the high-pass filter method is the correct way to remove the offset.
A 10 Hz high-pass filter is not good for everything, but the point of this exercise is that neither is averaging positive and negative sample values. There is no one “fix-all” tool, it’s horses for courses. Cool Edit Pro offered both of these types of DC offset correction.