At 9/6/14 11:10 PM, Bit wrote:
Nonsense
Or, better yet, I just scienced up a chart.
Collected file sizes are in MB. Tested sample rates are in kHz. Green, Yellow, and Red represent relatively how awful the resulting audio is -- Green: best, Yellow: passable, Red: my ears are bleeding. Black spaces show untested or untestable data.
All test files are children of the 192kHz 24-bit FLAC file which I used as a reference. File contains a rip of one side of a vinyl album ~20 minutes of music.
Results were interesting.
Converted files show a nearly linear change in file size depending on the sample rate selected, all seeming to converge at what looks like a minimum possible 4-5MB. The smallest file I created was MP3 VBR-9 at 8KHz which was a miniscule 5MB. At higher bitrates, halving the sample rate seemed to halve the resulting file size as well. At lower qualities, the effect became less pronounced. Unfortunately, all of the 8KHz files I created were not pleasing at all to listen to.
Keeping the 48KHz sample rate and stepping down to a lower bitrate seemed to be a more viable solution. MP3 VBR-0 48KHz to MP3 VBR-5 48KHz nearly halved the file size. MP3 VBR-5 48KHz to MP3 VBR-9 48KHz was another significant drop in file size. Thankfully all of the 48KHz recordings were passable even when the bitrate was reduced.
Conclusion:
Ruthlessly cutting the sample rate to 8KHz will reduce the file size to its minimum, but shouldn't be necessary when nearly the same effect occurs when using a low bitrate such as VBR-9. Solutions such as MP3 VBR-9 48KHz seem to be the best compromise of audio quality and file size.