Spectral Difference Question

[Dan McDonald]

I have an old, very hissy, multi-generation tape. I have another performance by this group on a different night, with some of the same songs, but in very clean condition.
I tried the spectral difference filter on it using the clean copy as the source, and it really makes a huge difference in the hissy tape - the instruments are much clearer, and overall the sound is very good, except the tape hiss increased dramatically. I read in the manual that you should do all the restoration work before applying the spectral difference filter, and I do that, but still get a very hissy result.

Is there any particular de-hissing that might work better in this situation? There's very little high frequency sound (music) on the tape I'm restoring above 5kHz , but the clean copy goes up to about 16khz, so it's really emphasizing the high end. I am having trouble cleaning up the hiss without also losing the signal.

Now that I phrased it that way, I guess that's really not any different from any other time we do hiss reduction - but the difference is the amplification afterwards.

Would it make sense to run the CNF before the spectral filter and run it again after to try to clean it up?

Dan

[Craig Maier]

Dan,

hmmm - - - I am a little confused by the terminology here. There is a spectral subtraction mode in the CNF and there is also a a Spectral Filter that has a mode called Spectral Difference. Am I right in assuming that you are using the Spectral Filter here in Spectral Difference mode? If so, I would expect that the noise would go through the roof when using that mode with the clean tape as a reference. If that is correct, I think that I would then go back and apply another instance of the CNF in Normal mode to the final result to kill the hiss above 5 kHz.

But, maybe I do not have the situation correct here - - - put me on track.

Craig

[Dan McDonald]

Hey Craig -
Yes, you're correct. I meant spectral difference, not spectral subtraction - I need to start re-reading my posts before I push submit.

Anyway, it really works wonders with the tape in terms of clarity of the music when I'm using the very small fft values, but as you say, the noise level is through the roof.

If I add harmonics, etc. to the tape before I use the spect. difference, then you really hear some odd, sort of grating, harmonics, because they become really loud.

Dan

[Craig Maier]

Hi Dan,

I understand. So, as I said, try taking the result of the Spectral Difference and then feed that through the CNF and drag the inflection points above 5 kHz all the way up to the top. You may have to add a point or two to do that, but that is easily done with your mouse.

Alternatively, divide the last process into two pieces. First, run a Low Pass Filter set for around 5 kHz at the steepest slope (24 dB/Oct) or use the Brick Wall Filter. Then, run the CNF on that file.

That should help. Let us know if it does.

Craig

[Dan McDonald]

OK - will do, Craig. thanks for the help.