Bass Master - You are correct. Preserving your audio at 32bit when moving between programs reduces the number of times the file is dithered/truncated and then padded.
It is important to understand, however, that when you're recording at 32bit the software is just making up up the least significant 8 bits (e.g. probably just padding with 0's). They don't become important until DSP is applied to the file.
An example. You record a 24bit file in your software, and you then apply some VST effect to it destructively (e.g. you change the file). All VST effects expect 32bit floating point input streams (normalized between -1 and 1 FYI), so when applying the effect, your software converts the 24bit samples it reads from the file to 32bit samples, sends those 32bit samples off to be processed, then dithers or truncates the 32bit output back down to 24bit and saves the new values to the file. Whatever minute details that were contained in those last 8 bits are lost. However, if you were to start with a 32bit file in this example, the dither/truncate stage would not occur, and you would preserve those little 8 bits that may prove to be important.
Now does that mean everybody should start recording 32bit waves? Not necessarily. If you record at 24bit and mix a project totally in realtime (e.g. all effects are applied in realtime), then you are sacrificing absolutely nothing. The software takes all of the 24bit samples from your files, converts them to 32bit samples, processes everything at 32bit, then dithers/truncates back to 24bit for playback. Yes there is that last dither step, but it is necessary regardless because your converters are only 24bit! In this case, the only time you'd need a 32bit file might be when you mixdown. Mixing down at 32bit will allow you to preserve some fidelity as you move your mixdown between various mastering applications.
Slackmaster 2000
