Click Detection

[knighte]

Hi Derek,
Allow me to say that you have made an old Vietnam Vet extremely happy! I have Lp's from that era which are very dear to me, and now I can make recordings which are better than the first time I played the disc.
1. Am I mistaken or is the program more likely to miss a large click (looking at the waveform) if the surrounding area is relatively quiet? Can allowances be made for this condition?
2. Is it possible to have the software search for the extreme changes in amplitude relative to time and then make the parameters flexible? If one still heard clicks and crackles in a certain area of the music he could then find these changes and decide if they are music by looking at the waveform. Nailing down the exact location of offending noise would then be less time consuming.

[citguy]

Hi Larry. Before Derek replies, have you tried using "go to maximum sample" under the 'Wave Form' menu? If the offending 'click' is above the acoustic level of the recording it should zero in on it and you can then correct it manually.

Stan

[knighte]

Hi Stan, I thought the same thing one day and gave it a try. Trouble is the clicks which are missed, are in areas where the pre and post levels are much less, even though the click is quite obvious. However, the level of the click can be quite a bit less than the maximum levels of the recording and still be heard. I have been playing around with the noise profile trying to use one of these "areas" to see if the software is more aware of these conditions. I'll let you know if it helps. By the way, if a click is buried amongst loud passages, I use "cut and splice" to zero in on it. Once you realize it is gone, (keep the selection as small as possible) undo the "cut" and it's amazing how obvious the click appears against the white background. The background is black in the selection area. Thanks for the try. Love this program!

[Glenn]

Hi Larry;
I've noticed the same problem as you have, that being the obvious clicks here and there that Wave Corrector misses. I haven't a clue as to why this is, but I console myself knowing that the programme catches a great many that are very well concealed.
Your idea for running the de-noiser is a good one, but it isn't likely to work as you'd expect. The conventional wisdom is that click removal should always be done first. There is something that might help the situation however, though I haven't tried it yet personally. The idea is to boost the treble before you declick and denoise, then reduce the treble again afterwards by an equal amount. It's the same idea as the analogue noise reduction methods that have been used in the past, with additional digital noise reduction added in between. This is supposed to work very well with the denoiser, but I'm not sure how much benefit would be gained with Wave Corrector's click detection. It may help to catch those easy ones, but it may also throw a lot of false positives into the mix.
Glenn

[knighte]

Hi ALL!
This is getting exciting. Stan, your idea of using "max sample" needs to be reconsidered. If we could do a "max sample" on a selection block, it would solve my concerns. Then one could just continue to sample until the largest missed clicks were gone. Stop at any stage, right? Also, thanks to Glen, I realized that the noise profile is primarily for hiss. When he mentioned boosting treble I knew he was referring to Dolby noise reduction. I tried the boost Glen, with no success. However, I have noticed that changing the level of the recording by using normalization, the program is able to locate more clicks. For instance, knocking the level down -20db enables the software to find small clicks in a quiet passage that it missed originally. Couple this with the rumble filter, save the file, reload, change levels, and rescan. In fact, if this were done first, and then the "max sample" on a block selection, I doubt there would be much left. Am I nuts or what? A word of caution - changing levels and rescaning till you're blue in the face will cause distortion in the music file. Better off to select the problem areas. One other thing, someone on another post mentioned "Audacity" http://audacity.sourceforge.net/download/windows
This program gives a person the ability to null the noise between tracks and quite a few other things that Wave Corrector is not designed for.

[Derek]

Sorry for the delay in replying but I've been rather busy and you've raised a number of issues that needed a considered reply.

First, the recommended way to deal with missed corrections is to mark a block around the area where they've occurred and to superscan one or more times to 'mop up' the clicks missed in the initial pass.

Sometimes, the characteristics of a click are such that it will not be detected. This is usually because it has slow rising and trailing edges (it may sound more like a thump). In this case, it is necessary to use a manual correction or to use cut & splice to remove it.

Now to some of the issues raised by this thread.

To go back to your original questions:

1. Am I mistaken or is the program more likely to miss a large click (looking at the waveform) if the surrounding area is relatively quiet? Can allowances be made for this condition?

Actually the program is more sensitive during quiet sections. The click detector works on the relative amplitude of the click compared to the surrounding waveform. Therefore, during quiet sections, the detector is more likely to fire. You can exaggerate this effect with the AGC setting in the advanced options. However, note that as you increase this setting, it makes the detector less likely to fire during loud sections. Therefore, to get the effect you want, you would increase the main sensitivity control to say 4 or 5, and prevent too many false positives during loud sections by backing off the AGC setting to say 2 or 3.

2. Is it possible to have the software search for the extreme changes in amplitude relative to time and then make the parameters flexible? If one still heard clicks and crackles in a certain area of the music he could then find these changes and decide if they are music by looking at the waveform. Nailing down the exact location of offending noise would then be less time consuming.

This is really what the program does. The procedure is that it makes corrections and then you review them to see if they're acceptable or whether they need adjustment or removal. The magnitude value in the corrections list gives an indication of how severe the click is likely to be and hence how significant the correction is likely to be.

By the way, if a click is buried amongst loud passages, I use "cut and splice" to zero in on it. Once you realize it is gone, (keep the selection as small as possible) undo the "cut" and it's amazing how obvious the click appears against the white background. The background is black in the selection area.

Yes, this is a very effective technique. However, you don't need to explicitly use the cut & splice' command. When a block is marked, you can use the 'Audition Corrected' toolbar button to simulate a 'cut & splice'. You can use this feature to home in on an unwanted click.

Reference Noise Profile: This has nothing whatsoever to do with click detection. It is only used by the hiss and hum filters.

Reference applying treble lift: This is unlikely to have any significant effect as the lift will be applied to both the wanted music and the unwanted clicks. Therefore the relation between the two will be unchanged.

[knighte]

Hi Derek,
Thanks for the tip on using a marked block.

"Actually the program is more sensitive during quiet sections. The click detector works on the relative amplitude of the click compared to the surrounding waveform. Therefore, during quiet sections, the detector is more likely to fire. You can exaggerate this effect with the AGC setting in the advanced options. However, note that as you increase this setting, it makes the detector less likely to fire during loud sections. Therefore, to get the effect you want, you would increase the main sensitivity control to say 4 or 5, and prevent too many false positives during loud sections by backing off the AGC setting to say 2 or 3.

I've been experimenting with a couple of missed clicks that were super scanned at level 5 and the AGC set to Zero. Bumping up the AGC seems to have no effect on these. Please don't get me wrong, I don't expect a complete rewrite of an excellent piece of software. I'm wondering if it would be too difficult to implement Stan's idea of having the program do a "max sample", but have the capability to do this within a marked block. I have zeroed in on the few missed clicks that just won't go away, (an LP with more defects than the face of the moon) and found that once the resolution is increased, they are the "max sample". Perhaps they are just the "max sample" on the waveform, not of the music passage. I think that marking off one of these quieter passages and doing the sample, would enable one to find the rouge clicks more quickly. Thanks much for your earlier reply.

[Ray Bell]

Hi all

From the discussion it seems that adding the functionallity to go to the maximum sample within a marked block could be useful in tracking down some of the occasional clicks which Wave Corrector misses. This seems to me (I'm sure Derek will tell me if I am wrong) relatively straightforward to add as it appears to build on functionallity which is already present i.e. selecting a block and finding the maximum sample in the entire file. The location for the control also seems fairly obvious as it could be an additional item in the block menu.

However these clicks are relatively easy to spot by the conventional process of repeatedly playing the section of music around the click and progressively zooming in on it .
The clicks I find hardest to track down are those which are buried in the music and appear for example as a sharp step in an otherwise more sinusoidal waveform. I find that the slow playback is the most useful tool for pinning these down but if anyone has a better technique I'd be interested to know.

Ray

[Glenn]

Hi Ray,

I find that the slow playback is the most useful tool for pinning these down but if anyone has a better technique I'd be interested to know

This hasn't worked well for me, as the transient nature of the click disappears into the background noise with the slower speed. What I do is scan through a section visually using the arrow keys to sweep the waveform past the screen. At the proper resolution, the waveforms undulate past in a repeated fashion. The defect stands out as an aberration against the uniformity of this repetition. It takes a cerain state of mind to be able to see the disruption, but with this approach I've been able to find clicks that were literally nothing more than a broadened wave crest against the surrounding norm. The trick is to find the correct resolution and sweep in such a way that the wave undulations appear to stand still. I believe this is a visual analogue of the way we hear clicks in the first place; repetitious patterns that are suddenly (and unmusically) interrupted.
Glenn