Whenever I save a file through Audacity, the sound quality is slightly compromised, even if I select the highest possible settings. For example, if I try to save a .wav file and choose 96000Hz, it sounds worse than the original 1411kbs file. Is there a way to maintain the original sound quality?
Saving (File > Save Project…) is for Audacity Projects and is lossless.
You File > Export… WAV files, and “Save” is the last button when you do that.
1411 kbps was presumably 44100 Hz stereo 16-bit PCM WAV? What effects did you apply before you exported at 96000 Hz? Change Tempo and Change Pitch do sound worse if you make too extreme changes. You are trying to stretch sound un-naturally with those effects.
You can never improve the original by increasing the sample rate. You can theoretically make it slightly worse because of resampling. So save your disk space.
You don’t say in what player it sounds worse and how it sounds worse. If it’s a battery player it may struggle with the extra data (the 96000 Hz file will be more than twice as large as a 44100 Hz WAV).
Gale
I just played the files on my computer through WMP or VLC.
My preferences are:
Default Sample Rate 44100 Hz
Default Sample Format 32-bit float
Sample-Rate Converter High-Quality Sync Interpolation
Dither: None
Sample-Rate Converter High-Quality Sync Interpolation
Dither: Shaped (does this add slightly more bass?)
I think you should do double-blind tests to verify what you think you are hearing. If you still think there is a problem please give all the steps to reproduce it including the all the effects you are using.
Shaped dither adds low level noise which is concentrated in the highest frequencies. The low frequencies in the dither noise would not go above -60 dB even if you amplified the noise on its own to a peak of 0 dB.
Gale
Got it. Thanks, Gale!
I am not sure about the accuracy of the import and export of wav files.
As I could not import 2.4 GB wav files (about 8 hours 16 bits stereo at 2250 samples /sec) I decided to split then first.
Being alarmed with early results I used :
- Wave Agent to split the file in to mono files
- Bigasoft Audio converter to get the first 2 hours
- Direct Wav MP3 Splitter to get the first 2 hours
- Audacity to get the first 2 hours (the only thing I did was import, selecting the first to hours and export the selection)
Analysing the results in R it turned out that the first three method had the same result which was differed from the Audacity result.
More precise 80% of the samples had a different value although the average absolute difference was small (1.98)
To be sure I used Audacity again and in the same way to get the first hour of the 2 obtained with the Bigasoft Audio converter.
Analysing in R again, proved the 1 hour exported from Audacity was again different form the file imported.
Maybe I overlooked something. But is seems that just an import and export may change your wav files.
Steve,
Thanks for our reply. It might be the key for an explanation.
However, I didn’t change the sample rate.
But the section “Dither applied on Export - detailed explanation” in http://wiki.audacityteam.org/wiki/Dither starts with the statement
“Dither is only applied when converting from a higher bit depth to a lower bit depth”. And I didn’t change the anything.
I only imported a wav file, selected a range for export, and exported the selection.
Are there other issues involved?
Audacity works internally in 32 bit float format.
Normally when you export from Audacity you have done “something” with the audio (pretty pointless just importing a file then exporting it again without doing anything between 
). On that assumption, when Exporting, Audacity renders all non-muted tracks to a file. Unfortunately Audacity is not smart enough to know that you have done nothing, so it renders the audio (32 bit rendering) to your new file. If your file is in a format less than 32 bit, then dither is applied (unless you have switched it off in Preferences).
Steve,
Thank you very much. That solved my problem.
Jan.
PS: You right an import only to export is usually pointless. However in this case it helped to make my point clear. And moreover, as I could not import my over 2.5 GB wav files directly into R for statistical analysis, I used Audacity only to split them into manageable pieces.
Yes that is one of the few use cases where it is beneficial to temporarily turn off dither.
Personally I would like Audacity to handle bit format conversions more intelligently so that dither is only applied when it really needs to be applied (easier said than done )