There is one subtle difference between writing 24-bit audio with audiowrite vs. wavwrite.
When writing int32 as 24-bit audio in wavwrite, the data range in y is interpreted as:
–2^23 ≤ y ≤ 2^23–1
When writing an int32 as 24-bit audio in audiowrite, the data range in y is interpreted as:
-2^32 ≤ y ≤ 2^32–1
So your data is still being written but slightly quieter. If your 'signal' variable is in the 24-bit range you will need to rescale it to the 32-bit range to get the same results as using wavwrite. If 'signal' is just a normalized double (-1 ≤ y ≤ 1), the you can simply supply the double value and audiowrite will scale it for you properly:
audiowrite(['test.wav'],signal,Fs,'BitsPerSample',24)
