Christian LeBlanc wrote:I think of it in terms of image resolution - like 16-bit will look as good as 48-bit in a thumbnail, but if you want to blow it up to poster-size, 16-bit is going to have lots of blocky artifacts and visual noise that you wouldn't otherwise notice. 24-bit is going to look a lot cleaner, though. The higher the number, the more "room" you have to tweak and make changes to your sounds without mucking things up too bad, which is why people will associate a higher bit rate with more headroom.
I also have a working understanding that a mix-down file will sound better in 24-bit if you were working in 32-bit, even though the song itself is only 24-bit.
What I understand of float, is that it's like a floating-point decimal, keeping high resolutions on what you're working in...when you mix down to a cd-quality file, you won't hear the 32-bit float quality, but you'll hear a difference than if you were working in 16-bit to begin with. Is everything above 24-bit automatically float?
You've sort of got it, but the picture analogies are throwing you off. First of all bit depth refers to dynamic range, not "quality", or resolution. Since all digital files have the same max level of 0dBFS, you don't actually gain any headroom when increasing bit depth, you actually gain footroom. But since we have more footroom before the noise floor, we can record at a lower level, thereby allowing ourselves more headroom, which is what this conversation is all about. 24 bit actually has 144dB of dynamic range which is approximately the range from dead silence to making your eardrums meet in the middle of your head. ie Way more than you need. But this allows us lots of freedom with levels, as long as we stay clear of 0.
Another misconception is that 32 bit will sound better than 24 bit which is totally false. 32 bit files are capable of a much greater dynamic range, but as stated 24 bit has more than enough dynamic range for humans, so as long as you maintain good gain structure and stay within that range there will be no difference between the 24 bit and 32 bit float files. If you do stray outside that range in 32 bit float, when you reduce down to 24 bit, you will be in for a world of hurt. 32 bit float files are actually just 24 bit files with an 8 bit modifier to fake their extended dynamic range. So basically you still only have 144 dB of dynamic range, but that range can be from 0-144, or 96-240, 144-288, or 576,254-576,398.
Everything above 24 bit is not float, for instance Pro Tools HD has a 48 bit fixed internal mixer, so it has a true dynamic range of 288dB, which is only useful when we starting getting into the math of signal processing, which I'm am woefully unprepared to discuss at any length. All of this is before we start talking about dither too, which is a whole nother kettle o' fish.