SOUND ON SOUND

[TheLegit]

Just wondering if when rendering files to audio (say a MIDI track running a VST) , do you render to 24 44.1 or 32 Bit Float as that option is in there?

[James Perrett]

If the levels are sensible then I'd just render to a 24 bit file. You won't lose any information by doing this and you'll save a bit of disc space.

In fact I often use .Flac files these days which can halve your disc space requirements with no loss of quality.

[Martin Walker]

I do still render my Reaper projects as 32-bit float, but then they immediately get tweaked i Har-Bal and pre-mastered in Wavelab 8, where they rapidly get halved in size after having their LUFS adjusted and 16-bit dither added.


Martin

[RichardT]

24 bit for me. The dynamic range of 24 bits is something like 144dB so you’ll be fine even with relatively low level signal levels.

[TheLegit]

Ok thanks all

[Martin Walker]

Yep, thinking on, I'm overdoing my renders - I just seem to remember the internal Reaper format was 32-bit float, so I defaulted to exporting in that exact same format before my final sonic tweaks.

The dynamic range of my tracks is already well under 144dB, so I think in future I'll switch to 24-bit rendering too :headbang:


Martin

[CS70]

Yes, 32 or 64 bit floating point is good for calculations, such as they happen inside the mixing engine of the DAW. It allows for stuff like exceeding 0dbFS on a bus, going into another bus and bringing the gain down to below 0dbFS there, without loss of information. A 24 bit integer mixing engine would simply not be able to represent the intermediate product, which is why mixing engines use the larger floating point numbers.

But so long your final product, i.e what is on the master bus, is below 0dBFS (as it should), you don't need floating point, because there's no further calculation to do: the definition of being below 0dB 24 bit Full Scale is exactly that all the samples fit in 24 bits.

In short: if you aren't exceeding 0dBFS on the master bus, a stream of 24 bit integers is equivalent to a stream of 32 bits floating point numbers, so there's no need to save in that format nor to dither.