Quote:
Originally Posted by deanrichard
I'm not sure you understand bit depth. The number of bits does not increase or decrease levels. it only increases signal detail and raises the S/N ration. You can overload a signal just as easily in 16-bit as higher bit depths.
If a player distorts with a 32-bit version of an MP3, and does not distort with the 16-bit version of that same file, it is broken.
|
The OPs post makes sense to me. Fixed point audio, regardless of whether it is 16,24 or 32 bit will adhere to a limit of 0dbfs where it inevitably clips and information is lost forever. A 32 bit floating point audio file (like used in PT and most modern DAWs) is different. It has a 24bit mantissa that essentially carries the audio sample information, and an 8 bit exponent that effectively scales the range of audio up and down across a huge dynamic range.
If you create a 32bit floating point file with peaks well beyond 0dbFS, you can still play it through a 16bit fixed point converter without clipping. All you need to do is turn down the gain digitally so that the peaks are below 0dbfs. According to the OP, this is possible even with common apps like VLC on OSX.
I never considered 32bit floating point as a delivery standard. It's an interesting thought. How do you compress and brick wall limit to 0dbFS if there effectively is t one?