For clarification Floating point and fixed point <-> float and integer
I have noticed that some of the audio encoders have strange options.
So i thought let's contribute a bit.
Normal ''cd'' audio is 16 bit. Almost all high end recording/Blu-ray audio and other forms of ''hi-res'' audio are all 24-bit. These are integer values.
Most audio edit software converts the 24-bit integer to 32-bit float to be more flexible during internal processing and audio editing. After Editing you always export to 24bit int again.
Currently a couple of audio encoders like flac have the option of 16 or 32 bit.
It would be more logical option to have 24bit instead of 32 bit. I cannot think of a single source or distribution that uses 32 bit integer audio. Its always 24bit and varies from 48khz up to 192khz.
Increasing the bit depth to 32 integer serves no purpose and will not increase audio quality only explode the file size.
There are even more obscure audio formats like 64-bit float and 39-bit integer but these are exclusive to high end Summing in Digital Mixers and DAW's so also totally not relevant.
Secondly a strange thing is that with AAC, EAC3, AC3 encoders it notes floating point format. (also mp3 has a float option)
This is completely incorrect, AAC, AC3, EAC3 and mp3 are fixed point encoders only and they ingest fixed point audio only. There is one exception in the form of floating AAC-LD(low delay) version for Realtime communication systems but this is completely removed from consumer or professional audio.
Could be that i do not properly understand the bitstream going true the software layers but from a audio engineering point of view it makes no sense 
Hope it helps 