View Single Post
Old 24th October 2009
Gear Guru
UnderTow's Avatar

Verified Member
Originally Posted by bob katz View Post
No, this is incorrect. 0 dBFS is the same level for all wordlengths for all intents and purposes, within thousandths of a dB. The longer the wordlength, the lower the lowest signal which can be encoded.

Hey Bob,

I suspect the original poster is referring to 32 bit float and 64 bit float in which case they do not clip at 0 dB FS.


I am feeling a bit lazy to do the maths now but if my memory serves me right, 32 bit float has 1536 dB of dynamic range. About 800 of which is above 0 dB FS. 64 bit float has even more. You could say it is close to unclippable.

The thing to remember is that this is only valid while remaining in floating point! If you need to get a signal out to you speakers for instance, it will clip if you go above 0 dB FS for the simple reason that all converters use fixed point formats.

On the other hand, you can have a bus or a track inside a floating point DAW go above 0 dB FS without clipping as long as the signal is reduced in level before it gets converted to fixed point (or hits a plugin that can't handle signals above 0 dB FS for whatever reason).

PS: 24 bit fixed point format also clips at 0 dB FS. There is no extra headroom.
PPS: Another thing to remember is that although floating point signals don't inherently clip above 0 dB FS, you can not be sure that any plugin fed with such a signal will behave correctly. Staying below 0 dB FS just makes life easier. There are enough things to worry about without having to worry whether plugin developers took bad gain staging into account. :-)