Usually I like to work with two copies of the same album so I can avoid problems like this. But in this case, I've only got one.

There's a very loud "click" at the end of this track that I'd like to remove. I've tried interpolation, but it's still audible - mainly because the music is so faint. Is there any hope for it? Or will I just have to fade the track prematurely?

The click is easy enough to take out; perhaps it depends on what software one uses. This file has already had some seemingly strange stuff done to it, filtering out low frequencies and some rather heavy duty NR. This leave me uncertain of what I should hear. It would be easier to fix the click first, then do whatever else seems desirable.

The biggest problem you have is that the music at this point is a fading cymbal crash. This means that the usual strategies for hiding clicks this big don't work too well, because the nature of the sound is quite noticably different at the start and end of the click.

The click is very much quieter on the right channel, so the first thought was to replicate the right channel to the left in the brief section where the click is, which should then make fixing it somewhat easier. Unfortunately, the sound of the two channels at this point is radically different, so that didn't work. Frankly, I don't think you're going to achieve a seamless fix.

The best I could manage after a few minutes experimenting is here: Click to view attachment
You'll notice that the click is replaced by a much quieter "blip", caused by the faint "echo" of the cymbal splash due to an earlier block being spliced in. I've also tried to reduce the impact of the faint "plop" at around 2.25 sec. The fixes are a combination of manual block copying and spectral replacement. (I did try CoolEdit's "Fill Single Click Now" facility, which can sometimes work miracles, but in this case it didn't). My gut feeling is that additional effort is likely to be futile.

I see that Andy thinks the click should be easy to remove. I'd be very interested to hear how he would go about doing that, and hear his results, because frankly I don't think it's easy at all, due to the nature of the signal (as previously noted).

I did my normal manual declicking, then I used a highpass filter (2nd order Butterworth, cutoff at 150Hz) to remove the low frequency bump. I also removed the leftover bump from another click at 0:02.263 and removed three smaller clicks towards the end.

The major difficulty I see is the very heavy NR already done. This leaves a good bit of the noise around the click, especially following it. Because of that left over noise, at a high enough volume level, there is a noticeable change in the sound of the fade off (which already sounds peculiar became of the heavy NR), but the click is completely gone. It is not possible to tell what it would be like if the click were (properly) removed first. However, I can only hear this strangeness when listening at a level that is painful near the beginning of the clip. At a more ‘normal' sound level, it is inaudible, the fade is smooth.

I would be happy to post the result, but I have no web page, and I am unwilling to sell my soul to obtain access. If anyone can suggest an easy solution, I will post.

The only filter applied to the file was noise-image reduction using Pristine Sounds. I remove only the audible clicks individually with WavCorrector. Here are the same samples completely unfiltered and unedited.



I honestly can't hear where the click used to be. As for that "plop" at 2.25, you need to have the volume pretty loud to notice. It's still less audible than that click was when I applied the interpolation, so I won't worry about it.

Thank you.

Lucky for me, this is an EP and the other two tracks are available on CD. I don't have the patience to do this with an entire album.

You can just upload it as an attachment to a posting - that's how I posted mine.

Just deleting the waveform from sample 123645 to 124881 is enough. CEP automatically applies a 3ms cross fade to the cut, which probably helps.

The loudness of the cymbal does jump down at the cut, but unless I'm imagining it, it jumps down at the click anyway, implying some kind of AGC in circuit at some point during recording. Can't do a lot about this.

Cheers,
David.

I see that you got the answer to your problem, but I thihk that I can help with some more advice.

Never apply noise reduction (NR) before click/pop removal! In order to cope with transients, you have to use a smaller FFT size, but then you loose frequency precision in NR process. You must ensure that there are no audible clicks before running NR. You can use click/pop remover tools, manually edit critical samples (in cases when tool doesn't help), or cut/interpolate.

In this file it was enough to manually use Audition's Click/Pop Eliminator with settings:
FFT Size: 255
Pop Oversamples: 12
Run Size: 10

But I can't see attachment option to attach the fixed file

I use the David's method as the last resort, or when DC level shift caused by pop is to big.

BTW, I did a lot of old vinyl restoration for our National Library, so I encountered a lot of different cases.

I took a "silent" sample (a few seconds at the end of the album) that didn't contain any clicks/pops and used that for the noise print. I then applied it to the entire recording prior to the manual click/pop removal, since it causes some of the minor ones to become inaudible under my normal listening conditions. That way I would know which ones I still needed to remove.

It's my understanding that using such a noise print for noise reduction works the same whether the recording has click/pops or not.

You should declick before denoising. If you do it the wrong way around...

Stage 1: Denoise. Each click opens up the noise gate in each FFT bin of the denoiser, letting extra noise through.

Stage 2: Declick. The declicker will remove the click, but won't touch the noise which leaked through the denoiser in stage 1.

So you're left with little low level bursts of noise where the clicks used to be.


Declick first, and there's no such problem. The clicks are gone before the denoising, so the denoiser keeps the noise gates shut. No problem.


You'll have to figure out another way of deciding which clicks need to be tackled, and which ones ignored. Maybe work with two files: a denoised version, and a raw version which you declick, then denoise.

Cheers,
David.

There is nothing to add to David's explanation
In fact, I always do as he said. Keep the two versions of file for the reason he gave, and also to be able to fine tune NR settings. You don't want to apply to much NR, just the right amount. Otherwise, you get annoying artifacts and degrade audio instead of improving it.

Thanks for the tips.