noise spectrum is saved without over-subtraction factor

[erazortt]

Currently the when using nfile or nstring dfttest also processes the clip. However this processing is different to the processing when the output file saved by nfile or nstring is then loaded using sfile. And the difference is due to the handling of the over-subtraction factor a.

The over-subtraction factor can be changed in the analysis phase using the nfile or nstring arguments. However the parameters saved in the output file are always scaled back to a factor of a=1, which is not even the default (that is 5 as far I can see) thus leading to the different processing of the clips when using the nfile or nstring argument in contrast to sfile. Since I see no other way of changing that factor later in the read-in phase using the argument sfile, I would think that the saved output file should have been saved including that factor.

Suggested code change, which I tested compiling dfttest with VS2019: In function getNoiseSpectrum almost the very last line (2276 in current master) is relevant for saving to the output file: outputSigmaFile(buf, dest, wscale / alpha, zmean);

There is the normalization with alpha. I would thus call for removing that normalization, so that the user gets that factor into the output file leading to a consistent behaviour when using sfile instead of nfile or nstring: outputSigmaFile(buf, dest, wscale, zmean);