CNF Crashes with FFT setting of 16384 with stereo file and Keep Residue checked

[dbanani]

When I need to de-click a REALLY noisy record and no combination of filters does a good distortion free result, I revert back to using the noise removal and inversion method.
The first step in this process is to use VERY aggressive settings on the CNF.and Keep Residue to later invert phase and mix back into the original file.

The FFT setting of 16384 works perfectly for mono files
When a STEREO file is processed, DC10Forensics CRASHES EVERY TIME.
This is a carryover bug from DC8.5

[Craig Maier]

Ok - will look into it. What sample rate are you using?

Craig

[dbanani]

Hi Craig

Normally I use 96/24.
I have tried 44.1/16 and 22.1/8 but still get the crash every time with a stereo file. Never with a mono file.
Just tried again with a 44.1/16 file of 23 seconds duration and had the same result - crash after running part way through the file.

DC8.5 does the exact same for these sample rates running on a different computer.

I split and recombine to process stereo files but this takes twice the work and twice the time to run the various filters.

D

[Marc Hildebrant]

D...care to elaborate on your method for noisily songs ? Could be useful for some tunes I have.

Marc

[dbanani]

Of course Marc. I will post it in the General Audio Forum. May be a day or so because I am traveling right now.
Have you tried to duplicate the bug In this thread?
Just set your FFT in the CNF to 16384 and click Keep Residue then run on a small section of a stereo file or the whole file. Fast Edit mode is what I use.
It is possible that it is a computer related issue even though it is occurring on 2 separate computers the hardware is identical.
I have DC8.5 on one and DC10 on the other.
D

[Craig Maier]

What CNF Mode are you using? The CNF has a choice of Normal CNF, Spectral Subtraction, Auto Spectrum, and (in the Forensics version - AFDF).

Craig

[Craig Maier]

Wait a minute. It just clicked on me. You are trying to 'Run' the CNF in "Keep Residue" mode. Keep Reside mode is designed as a Preview Mode function only so the user can hear what they are removing before committing to a "Run" mode. It was never intended to be used as a "Run" mode. So, if you want to Run it that way for some reason, all bets are off because we did not design it to be used that way. If anything, the bug is that we do not block a "Run" when the "Keep Residue" mode is checked. That is probably what we need to look into - - - lock out Run when Keep Residue is checked.

Craig

[dbanani]

Hi Craig

This method using DiamondCut Software has been a "secret ingredient" in audio restoration for quite a while.
First mentioned in Aug 2000 at:

http://web.archive.org/web/200104100...ML/000796.html


In 2001 you helped bobk with the inversion process for this method:

http://www.diamondcut.com/vforum/for...remove-crackle.

That post above also lays out the basic process pretty well.
I've been using a variation of this method for about 15-16 years on extremely noisy records. With the proper settings, results are possible that exceed other filters in direct use mode. And exceed what other software is capable of also!


in Nov 2016 you referenced this method calling it an "an interesting technique involving a combination of the CNF and the Impulse Filter to remove impulse noise from a record:" used in the video:

http://www.diamondcut.com/vforum/for...e-filter-video

If you check the settings on the CNF at around the 4 minute mark in the video referenced above you can see the Keep Residue box checked.
Unfortunately, the video takes the user to other Audio Editing software WRONGLY stating that the process cannot be completed in DC7.
The entire process can be accomplished in DC8.5.


In 2007 a very positive article was posted about DiamondCut DC32 calling it the "Younglove Process" (same name as 2001 bobk post):

http://www.normanfield.com/decrackle.htm

There was a lot of talk about this method around the turn of the century on audio forums.
Some of the success DiamondCut experienced early on was probably a result of interest in this method. The comments about the utube video are a good indication of the excitement generated.
This process is what brought me to Diamond Cut Software originally. I wanted to check it out.

The process is also mentioned (along with a plug for DCART) at:

http://www.delback.co.uk/lp-cdr.htm


If you lock out RUN on the CNF with the Keep Residue box checked, then this method will no longer work and will drastically reduce the functionality of DC 8.5 and DC10.


D

[Craig Maier]

I have never used this method. I do not know the limits of constraint on it and since we did not design it to work that way, we are not sure how to approach any testing of this. Maybe try using smaller values of fft or different settings of attenuation or something like that.

Let's face it, there are ten trillion or more combinations of things that can be done with Diamond Cut software, especially considering the multifilter. We can not respond to every possibility, else our entire life would be spent looking at ten trillion corner conditions and trying to make each and every one work. With software, that is simply impossible.

Make sense?

Craig

[Craig Maier]

As an aside, if you use the CNF in spectral subtraction mode in non keep residue mode, that should yield the same result. Keep residue mode (as I recall, and it has been a long time ago now) simply subtracts the spectral subtracted signal from the original signal. So, it seems what you are trying to do (if I have it right) is redundant.

?

Craig

[dbanani]

Absolutely Craig!
I remember back with DC7 or maybe DC6 we had a Forum Topic about the different things the software was capable of doing and everyone gave their different uses. Pretty amazing.
Personally, I prefer to use the fantastic filters in DC products as they were designed. Only in extreme cases when nothing else gives satisfactory results do I try this method. It does take a LOT of time to get right and then the results are never perfect. I always prefer to buy or borrow a better copy of the record to work with. But sometimes there are no other copies.
D

[Craig Maier]

Rick increased the buffer size for the CNF and he thinks that it now works in keep residue mode.

Craig

[Marc Hildebrant]

Craig,

Which version of Dc products would have the increased buffer size ?

Marc

[Craig Maier]

It will probably be the DCArt10 version, or perhaps the DCForensics10.03. Not sure right now as we are still testing it.

Craig

[Craig Maier]

This fix is in the DCForesics10.03 version and has been tested and will be posted on our website at www.diamondcut.com. It may take a few days before it is posted. Just check it out now and then at www.diamondcut.com to see if it is there. Also, this fix is in DCArt10, beta 1 (for our beta testers). Ultimately, it will be in the replacement for DC8.5 which will be DCArt10.0x.

Craig