I
Averaging is (X1 + X2) * 0.5
Size: a a a
2019 July 06
I
When you go from 44100 to 22050, you average each pair of samples and write it as a result sample
I
In cases when you go f.e. from 44100 to 32000 you must take into account that sample won't match for perfect averaging. In this situation you must also include neigbors multiplied by a weight that is effectively calculated as a "distance" to the sample on the time axis of a waveform
I
. . . .
..........
This is when they match with an integer factor. This is the easiest case.
I
Read each dot as a sample
I
0.4 0.9
0.4 0.64 0.74 0.9
I
2nd and 3rd are averaged from the source 2 samples
I
They get their values depending on how close do they get to original 0.4 or 0.9
I
AND THAT IS ALL ABOUT IT
I
They get their values depending on how close do they get to original 0.4 or 0.9
This is called interpolation
I
In simplest case you go for linear interpolation. In more advanced ones you might use Math.sin as interpolation
I
MG
so what actually determines the quality of the resampled audio
MG
besides the actual sample rate src and dest values
I
Matthew Good
so what actually determines the quality of the resampled audio
The answer is in your question. The source and the target sample rate.
I
The bit depth of the math you are using to resample it
I
If you calculate using signed bytes, you are doomed
MG
eg if from 44k to 48k is there such a thing as a
low quality and high quality resampling?I
The bit depth of the math you are using to resample it
.