Recalcitrant (and invisible) noises

[calberga]

I have some LPs which I am attempting to clean up, and they contain noises which are somewhere between a click and a small thump. The impulse filter doesn't touch them, and when I isolate them there is nothing visible in the wave file to indicate the cause. I have managed to pin down the locations of some, using repeated "Play from here" trials, and finding where the noise is heard and where it isn't, but there is still nothing to see.

At least some of the cases seem to occur at points where the music changes pitch, for what that is worth (if anything).

In some cases I have managed to eliminate the noise by judicious cut-and-pasteover, but in some even this fails. I can totally replace the suspect area and still have the noise!

Has anyone ANY suggestion for dealing with such cases? Would anyone be willing to look at a couple, if I were to post small sections of wave files on alt.binaries.test (or the location of your choice. I have a small ISP provided area where I can put HTML and such like, but they don't allow wave files or mp3s. Maybe zip files would work, I haven't tried them. Or I could rename the files as jpg files.

Anyway, any help would be welcome.

[Malcolm]

Re: Recalcitrant (and invisible) noises

Hi calberga: I have similar problems. Refer to my Posting under DC 5 "Elusive Pops". Your description could almost fit my problem. Something new, however, since these defects did not appear whilst monitoring the recording to hard drive, I tried re-recording at a sampling rate of 96K and performed my restoration under that condition, then finally resampled back to 44.1K and found that all but one defect were gone. I still think the problem is somehow how the software deals with a high frequency transient burst (even though my problems were in low modulated areas). I tried all the isolating and cutting tricks until I almost went out of my mind. The cuts always left a similar noise even though I carefully matched up the crossover points and and interpolated the cut points. Try that, I hope it helps you
Malcolm

[ 01-21-2003, 09:07 AM: Message edited by: Malcolm ]

[calberga]

Re: Recalcitrant (and invisible) noises

Thanks for the response. Unfortunately, my digital I/O card (MAudio DI/O 2448) doesn't handle 96khz even though my A to D does -- penny wise pound foolish?

Examining the record I suspect the "thuds" correspond to either a mild scratch, or more likely some little anti-dimples, bumps in the vinyl. It is a record from the 70's, and I know there was a lot of bad (recycled) vinyl around then.

I've just received advice about cleaning these up using (hush!) CoolEdit. I haven't tried the procedures, or figured out if they would translate to DC-Millennium operations yet. I'll post back on success or failure.

I must admit that CoolEdit (I've only been playing with it for less than a day) does some things more straightforewardly than Mil., but I REALLY don't like much of the interface, finding Mil. much easier to use. Of course I've been using Mil. for several years, but I think on a number of counts my feelings are objective. I may wind up having to move back and forth between them to get everything done, a daunting prospect.

I did, by the way, post a sample, 14.7 meg., on alt.binaries.test, under the unlikely subject line:

HWV 359a et al, Scans and Text, Sample CE.wav

(I know, I know, sloppiness -- I didn't update the subject line when I posted using PP2000.)

The sample has NOTHING to do with Handel.

If someone would be kind enough to look at it, perhaps they can come up with a DC method of fixing this.

[calberga]

Re: Recalcitrant (and invisible) noises

I warned that I would report back.

Some kind soul on alt.binaries.sounds.mp3.d looked at my sample (as I said above) and sent me his comments and fixup.

The bad spots are almost always a spot where the amplitude changes abruptly for a cycle or so and then subsides to the current normal. He referred to it as a "canyon". The Cooledit "Fix one click now" apparently will handle these fairly well (if they aren't too big, bigger than patch-interpolate though). However I haven't figured out how to isolate them there, while I can fairly easily in Millennium, using successive play-from-here trials to bracket the spot causing the noise, zooming in and repeating, usually right down to one or two cycles. What you will find is an area that either has a few peaks that are smaller or larger than the surroundings ones. Not the jagged shapes of the usual click, rather normal, just with too much or too little amplitude.

I tried a dome shaped gain change, but couldn't seem to get it right. Copy and Paste-over has worked in all but one case so far -- the failure is very a very indistinct case, and I am having trouble isolating it.

So -- is there an easier way to do all this? Will the magic of version 5 clear away all the fog? Will it warm up tomorrow? (Almost certainly no to the last.)

Is anyone other than Malcolm listening?

[JackH]

Re: Recalcitrant (and invisible) noises

I'm listening... Actually, I was waiting to see what sort of responses you got, because I've encountered the same sort of thing you describe on a few occasions.

I've had success a time or two by trial-and-error -- find the approximate location of the offending sound, interpolate a little bit, play back and see if it helped, and if not, un-do and try an interpolate just past the previous attempt.

Another method is to create a very brief "dip" in the level at the suspected point, and go through the same process. This, if nothing else, can help find the location of the noise.

Thern there are those cases where I can't find anything that works, and I have to live with it -- or find another copy of the recording in question.

[calberga]

Re: Recalcitrant (and invisible) noises

One interesting (infuriating?) point. If you compare the offending section to the original wave file (pre any processing) it often appears that the problem is caused by click removal. Not always, but often you will find a large click in the pre-impulse-filter file which corresponds to the "thump".

The conjecture I've seen is that the click removal gets the spikes, but doesn't take care of the reverberation that usually follows a big spike.