Noise reduction to my cassette tapes

[patrik]

First of all, im a beginner. Now to the topic. I have transfered a casstette tape to my computer, there is a buzz-sound in the background like it usually is on a cassette tape. How is the best way to rid of it? Any tip on how to use noise reduction?

[DougDbug]

Noise reduction should work pretty well in this case (low-level, constant background noise). But, sometimes you can get artifacts (side effects) so sometimes you have to play around with the settings to minimize the artifacts, and sometimes you're better-off just leaving things as-is. Make sure to keep a back-up of your original file!

First, open your audio file and select a section of noise only. Click Copy. This will place a "noise fingerprint' into the clipboard.

Select Effect -> Filter -> Noise Reduction -> Use Clipboard -> OK. (There are other presets you can try, but the noiseprint method us usually best.)

You can also try a noise gate. A noise gate completely kills the audio whenever the level falls below a preset threshold (i.e. between songs).

There are noise gate presets under: Compressor/Expander -> Preset -> Noise Gate. You can try the presets and experiment with the settings from there. Again, you may end-up making things worse... It can be distracting if you can hear the noise cutting in-and-out. So, just experiment to see if you can make an improvement.

Good luck, and have fun!

[patrik]

First, open your audio file and select a section of noise only. Click Copy. This will place a "noise fingerprint' into the clipboard.

After this, should i select all?

And what is FFT size?

Thanks for the help!

[DougDbug]

Yes. Select all to apply noise reduction to the entire file. (But, you should start by experimenting on a short section.)

The FFT selection is explained better in the help file than I can explain it. If you don't get satisfactory results with the default settings, you can experiment. I don't think there is one "best" setting for everything... It depends on the audio & the noise you're trying to remove... More overlap will increase processing time.


FFT is a conversion from the "time domain" to the "frequency domain". In the normal time domain, you have (for example) 44,111 sample-points per second. In the frequency domain, you group a series of samples (maybe 1000 samples) and you figure-out the frequency content of this short section of sound. Each series of samples is called a "window". The windows need to overlap in order to get a smooth reconstruction of the original time domain format.

[mh]

Beware of overdoing noise reduction on a cassette source. In many cases the volume of the noise can equal or exceed the volume of the subtle high-frequency info in the track, meaning that post-NR you'll end up with a fairly dull and lifeless result.

If you're prepared to live with some high frequency hiss remaining in the track, one solution is to copy the fingerprint, Paste New it, apply a lowpass (experiment to find the best), then copy that again for use as the clipboard source.

[DewDude420]

noise reduction is just that...reduction, not elimination. your noiseprint is analyzed in fft and a profile is gathered. its simply a bunch of frequencies and amplitudes that represent the desired noise. this doesn't make it smart enough to know the difference between noise and sound it only knows when a block of frequencies are above or below the threshold in the profile. excessive reduction causes distortion and that boils down to the nature of the effect. youre basically removing frequencies and while in a perfect world, it'd work fine, were left with various leftovers. finding the balance is hard. FFT size relates to how many samples are going to be used to process. higher numbers give you better frequency response but your time resolution is reduced, resulting in frequency smearing. low numbers give you better time response but less accurate frequency response which contributes to the bubbly/pumping effect. sorry this lacks formatting, my cell phone doesn't handle forums well. good luck and have fun