Continuous Noise Filter - Keep Residue

[Dan McDonald]

I process a lot of Edison cylinders and very early 78s, where the noise tends to be very high. I have had good luck with the following:

After de-clicking, I select a portion of the record where there is no signal, and sample the noise, then zoom the file out to the complete song and preview it with the "keep residue" filter on.

I then adjust the filter until I cannot hear enough of the source to tell if there is music or not (that is, it sounds like pure noise. Then I run the file through the CNF using with Keep Residue still on.

I then use file conversion to turn this into the right channel of a stereo file, and use file conversion to turn the original file into the left channel of the same stereo file (using paste mix). Then I convert the file into a mono file again, using the L-R file conversion.

This works incredibly well for some songs. Occasionally it introduces some artifacts, but I have found that when that happens, I just need to go back and redo the original keep residue function so that the signal is less audible.

Anyway, just thought I'd pass this on for those who are working on the old mono recordings.

Dan McDonald

[Rick]

Re: Continuous Noise Filter - Keep Residue

Dan
Have you compared the output of your method vs what you get out of the CNF directly with the same settings? It would be interesting to hear the difference.

Rick Carlson

[Dan McDonald]

Re: Continuous Noise Filter - Keep Residue

I have compared them when I do them as two different "start from scratch" procedures. The problem for me, I think, is that it is hard to tell when you're introducing artifacts into the signal, but fairly easy to tell when you're introducing music or speech into pure noise, so it's easier for me to get good results this way.

Also, since you have control over the left and right channels in the file recombination, you can adjust the noise reduction a bit to "fine tune" again before recombining the file. I find that when I recombine, I get the cleanest sound with the "noise" channel at -.3, and the music channel at around 0 or a slight positive value.

I don't know if what I'm doing is mathematically the same as just running the CNF filter, but it gives me better results this way.

The other day I did it with a cylinder recording I had processed as well as I thought I could a few months ago, and now it sounds ten times better than it did!

Dan

[Craig Maier]

Re: Continuous Noise Filter - Keep Residue

Pretty interesting. I will have to give it a try. Thanks for the idea.