Cleaning CD and CDR audio files with the EZ Impulse Noise Filter

[dbanani]

The EZ Impulse filter with the indicated settings in DCForensics10 and DC8.5 works great for getting the occasional scratch or pop out of a file ripped from a CD or CDR. (The above settings removed 9 spikes in a 2 second span of audio as seen above in the Total Fixed)

Often I find that the noise is visible as spikes in the peak file and then I can zoom in on the few seconds where the file needs cleaning.

Because of the nature of the CD as a digital source, the crackle and narrow crackle filters are not usually necessary, thus the 0.0 settings for each.
In fact, I have never found a need for these to clean CD audio.

CAUTION: I do not run this filter on an entire file. Only the damaged sections. The Crackle and Narrow Crackle settings of 0.0 do not appear to turn this portion of the filter completely off (note the small blue bar at the bottom of Clicks/sec above) This appears to be making changes to the undamaged audio. The Clicks/sec bar on the Crackle side will move up and down and the peak file will show visible changes. This can be seen by clicking undo (Ctrl u) after running this filter and noting the resulting visible changes to the peak file.

[Craig Maier]

Hi,

I do not see an image on my machine, but I figure that all the work is done with the "Scratch" portion of the EZ Impulse Filter. What setting value did you use? Did you have the big click filter checkbox turned on?

Craig

[dbanani]

Hi Craig
The image did not stick the first time for some reason. I edited the original post and at least it is appearing on my screen as I am writing this.
The settings are:
Scratch 0.2
Crackle 0.0
Narrow Crackle 0.0
All check boxes are NOT checked.
D

[Craig Maier]

Hi,

Ok - thanks. I now see the filter dialog box. Interesting settings. How long in time duration were the dropouts (or whatever they were) that were fixed by this method?

Craig

[Marc Hildebrant]

I'm not sure that I understand the use of the EZImpulse filter in your application. Are you saying that if you use the CD "ripping" feature that some noise spikes are sometimes created ?
Or, are you saying that even CD audio can sometimes contain noise ?

Marc

[Craig Maier]

I think that dbanani is saying that some CDs and CDRs become damaged over time, leaving behind dropouts and/or spikes when ripped. But, let's see what he says about it.

Craig

[dbanani]

Or, are you saying that even CD audio can sometimes contain noise ?

Marc

Hi Marc
Absolutely. CDs can have distracting pops and clicks just as vinyl and also CDs can have annoying cutouts where the sound is not ripped for various lengths of time. This screen shot is a portion of a ripped file from a commercial Mamas and Papas CD with scratches on the surface of the CD.

These are visible and audible in the peak file. Just like a scratch on a record, scratches on a CD can cause this type of problem.

This is a segment of the file .0077 seconds in length with audible noise.

I have a box full of damaged CDs and CDRs, many of which have this problem.
I have used the DCForensics 10 and DC 8.5 EZ Impulse Filter with the settings noted in the original post to repair the vast majority of these.

As you can see in the screen shot, these are sometimes VERY short. But they ARE Audible.

It seems to me that individually one of these spikes does not always cause a problem (not audible)
But collectively, the spikes caused by the damaged frames on the CD with two or more scratches (or other imperfections) in the surface of the CD will be in the audible frequency range resulting in noise.

The frames being read each rotation when ripping a scratched CD may not read correctly (damaged in some way) and the next time around a similar problem occurs to another frame resulting in a series of spikes at a frequency which is audible sometimes (but not always).
This is just my theory. Not confirmed yet.

That is what I believe is occurring in the file shown above and causing the noise.

Unlike a scratched record, which results in a momentary errant wave being created in the stylus, the CD noise often appears as a series of spikes and not a wave proper. This lack of wave form can be clearly seen on the left channel in the screen shot above.
Using the Zoom In feature on DC8.5 or 10, you can see this difference if you compare the noise on a file from a CD to the noise on a file from a scratched record. The Window Tile Function in DC8.5 and 10 is very useful for this type comparison.

Thus the noise in the file from a CD is different than that from a scratched record...and much easier to get rid of using DC8.5 or 10.
One time over this song with the EZ Impulse Filter using the VERY LIGHT settings in the original post completely removed dozens of instances of this noise and left the resulting file completely clean of noise.

D

[dbanani]

How long in time duration were the dropouts (or whatever they were) that were fixed by this method?

Craig

Hi Craig
The length is hard to measure. As you can see from the post above, they are extremely short. DCForensics10 will not measure these individually but returns 0.0025 secs on the collected grouping of events (spikes) shown above.
I will send you a snippet of this portion of the audio so you can take a closer look.
D

[Craig Maier]

Hi,

I tried your recommended settings on the file sample that you sent to me, and it found and fixed 4 events in that 2 second segment. I could easily hear the impulses before running the filter. After running it, the sound was perfect from what I was able to determine.

It's pretty interesting to me since I had not seen this before. It is good to know that there is a remedy for the problem! Thanks for the research and your conclusions. I guess as time goes on, we are going to be seeing more of this type of issue.

Craig

[Marc Hildebrant]

D,

Well....thanks for the info on the CD spikes. I guess that I had always thought that they should be clean and have no distortion if generated from master recordings.
I'll check some of my CD's !

Marc

[dbanani]

I agree Marc, that new CDs are almost always without problems.

The trouble is that the CDs do deterioate over time. Some also have manufacturing flaws. Some older ones have the layers begining to separate causing a form of Disc Rot. I see this sometimes on comercial CDs

A good article about this is on wikipedia:
https://en.wikipedia.org/wiki/Disc_rot

Also, with CDs and CDRs which have seen a lot of use, scratching, scuffs, abrasion, sun damage, and bending from handling can cause problems with readability.

I recently restored the audio from 87 music CDRs which were originally home recorded as cassette tapes in the 1970s - 90s. The cassettes were transfered to CDRs years later to "Preserve the Music forever".

While most were completely free of problems, several had Disc Rot to the point of being almost unreadable.
Portions of all of these were recovered. Some tracks were beyond recovery.

In one instance, the writing side completely separated from the CD when placed in an Ultrasonic CD cleaner. Fortunately, I had ripped this one and all the other's first before attempting any cleaning. There were gaps but some of the audio was salvaged.

Cleaning before ripping is essential to insuring the best rip possible. Evan a finger print can cause the audio to be distorted. I first encountered this with DC 7, I believe, when I was testing for jitter on/off issues. There was an audible click in the resulting wav file with jitter correction turned on. Turned out to be a dirty disk with several finger prints. A quick cleaning produced a perfect rip.

D

[Craig Maier]

I think that we are going to add that preset to the next release of the product which will be DCArt10. I already put it in there under the name "CD Scratch Pulse Remover". Thanks for the idea and I think many will benefit from it. It also makes sense to put it back into the DCForensics version too since many archives are CD/r based.

Craig

[sus4chord]

Very cool. I had never thought of that, and my CD collection is getting older as each day passes.

[dbanani]

Yup, just this morning I pulled out a copy of a cd by folk singer, Ralph Miller, who I saw live in Mississippi in 2007. I bought three different CDs from him that night. Ripped them with DC5 or DC6 the day after I bought them. Then put them on a shelf in my office. These CDs are no longer available anywhere. Very few were made and none are being offered for sale.

I re-ripped it today with DCForensics 10 along with some other CDs I bought. (easier to rip again than to dig the old hard drive out of storage) When I did, one of the songs had several audible dropouts and pops/ticks. I only found this problem because I listened to the wave files of the entire cd on DCTunes after I re-ripped.

Went back in my archives and found the Rip I did in 2007. Luckily I still had that old drive (actually a backup HD since the original had crashed a few years ago) and the wave files are still intact.

There were NO dropouts in the wav file ripped in 2007. I zoomed in close in DC10 and saw no evidence of any problems with the 2007 rip at the exact same location as the dropouts in the rip today. (Or anywhere else for that matter) The 9 year old rip sounds perfect. All tracks!

Looking closely at the CD It is actually a Memorex CDR with a printed label.

The CD has been in the original case in my home office since I ripped it so there are no environmental issues. Never once removed from the case until today. I always listened to the ripped file.

When I examined the CD, the damage was visible on the music side of the disc. Black spots under the smooth clear outer plastic layer corresponding to the area that that song would occupy. No Physical damage or scratches. Some of the printing on the label side also corresponds with the area of the dropouts so it appears the ink bled through the label and damaged the music layer.

I guess this means that anything we put on a CDR over the years needs to be checked for these type problems and a backup made to a hard drive, SSD, cloud or vinyl.

If it is really important audio, the only backup option I would consider is vinyl. I have many records over 100 years old and NONE have dropouts. The music and voices on vinyl still sounds as good as it did when the record was made unless it was scratched or otherwise mishandled.

[Craig Maier]

Hi - Interesting observations. I have heard from other customers that CDRs that use paper labels become damaged over time. The exact mechanism for this is not known. It could be the paper label glue and/or uneven expansion and contraction of the paper label compared to the CDR itself causing deformities in the recorded surface. What is not clear is whether or not your CDR in question had a paper label or some other type.

Craig