Noise when exporting as WAV

[markosis]

Hello,

I'm using Windows 7 and Audacity 1.3 Beta.

When I export a recording as a WAV file, it comes out with added noise (background hiss). And if I export that wav file again, even more noise is added. Repeat this again and it adds more noise.

I also tested this by creating 10 seconds of silence, which is indeed silent. But when I export this as WAV, then listen to said WAV file, it has noise. If exported again, the noise increases. It increases more with each export as WAV.

I don't remember Audacity 1.2 doing this.

Is there a way to fix this?

I need a DAW which adds no noise whatsoever to WAV files. Does one exist?

[kozikowski]

You may be listening to Dither. You can turn dither off in Preferences > Quality. You should not resample if you do that, so set Audacity to 44100, 16-bit (not 32) and Stereo, or whatever your show is. If the digital specification is different at the Input, the Output or Audacity Internal, then Audacity will need to resample and either the noise or the distortion will go up.

Koz

[steve]

I need a DAW which adds no noise whatsoever to WAV files. Does one exist?

Use Audacity and keep everything in uncompressed 32-bit float format.

The trouble with 16 bit integer format is that the sample amplitude accuracy is too low to avoid noticeable quantize errors when processing- this is unavoidable. Dither helps to minimise the distortion effects caused by quantise errors, but the only way to completely avoid these errors is to use a much more accurate format such as 32-bit float (which is what Audacity uses by default).

The problem with 32-bit float is that few consumer grade audio applications support it, so at some point you will probably need to down-sample to 16 bit. This should be done after all processing is completed, and at this stage you have a choice between dither noise, or quantize error distortion. Dither noise is generally considered to be the less obtrusive option.

[markosis]

Thanks for the help. Dither was indeed causing the noise.

Does Dither create noise in all DAW's?

[kozikowski]

That's what dither is. It's intentional very low level noise added to the show to keep the digital errors from adding up.

Sub sampling is inexact. If you never changed sampling rates, it would be a perfect world. But we do. You can have music that fits perfectly inside the 32-floating environment and there are just no digital values at 16-bit that work. If you do enough of that, the distortion spikes and the show becomes damaged. If you add just a pinch of noise to the system, the errors never become periodic -- they never line up to create audible distortion. But then the noise becomes audible.

So the choices are put up with the inaccuracies of staying 16-bit the whole way -- no dither needed because there is no sampling changes, go up and down in the sampling environment and put up with either the noise or the distortion, or go up and stay there. There are systems that "know" what 32-floating is. If that's your world, then you win. Many us us have to deliver Music CDs and they, by definition have to present at 44100, 16-bit.

Koz

[PGA]

The problem with 32-bit float is that few consumer grade audio applications support it, so at some point you will probably need to down-sample to 16 bit.

That is not down-sampling. You are not changing the number of samples per second. All you are doing is changing the arithmetic representation of each sample. I accept that going from a 32-bit floating point arithmetic to a 16-bit integer arithmetic will result in loss of "arithmetical precision" which will manifest itself as unwanted audible artefacts and so you need some form of "smoothing" of the sound. But it ain't down-sampling!!

[steve]

That is not down-sampling.

Strictly speaking you are correct.
To be accurate I should have said:
"so at some point you will probably need to reduce the audio sample bit-depth to signed 16 bit integer format."