One of the biggest hurdles when it comes to being proficient at recording is learning proper gain structure so your recordings are not too quiet or peaking and distorting due to being too loud. The reason we get these peaks is because of bit depth. Bit depth is the dynamic range of a digital recording; the difference in decibels (dB) between the quietest possible recording and loudest possible recording. Most engineers and hobbyists are aware of 16 and 24-bit depths and as you would expect 16 bit has a smaller dynamic range compared to 24-bit.
24 bit is the industry standard and has a dynamic range of 144 dB. This is plenty if you understand your gain staging but in certain situations even the best gain staging will fail to mitigate unwanted peaks in a performance or recording. This is where 32-bit float comes in.
32-bit float gives a whopping +1500 dB. This is an outrageous amount of dynamic range. Air itself can only handle up to 165 dB of sound before breaking up, once you reach 194 dB sound ceases to be sound and becomes pure force in the form of a shockwave and if you ever managed to make a sound at 1100 dB the result is a black hole that would have an event horizon larger than the known universe (so no more universe).
Obviously making a sound at 1100 dB would be impossible but what this means is that with 32-bit float you have essentially infinite headroom. You are effectively able to record extremely loud sounds and the data will be preserved so that when you begin post-production, even if the audio is clipping in your DAW, you can bring the gain down of that recording and all the sonic information will be preserved and free of any nasty digital clipping.
Relying on 32-bit float in lieu of proper gain staging is a no-no. Any engineer or producer worth their salt will have a good handle on gain staging, however, there are certain situations where 32-bit float would be extremely handy, essentially where you don’t have the luxury of doing multiple takes, such as recording audio for a documentary, live music recording or in situations where there is a large dynamic range which ends with an extremely loud peak.
Picture yourself trying to record a plane landing on an airstrip. When that plane is coming in, it’s going to be quiet due to the distance between you and it. Once it lands and gets closer, it’s going to be significantly louder. You could set your gain super low to anticipate this approach however you will lose the sound of the plane’s initial approach and will have a noisy recording due to the noise floor of your recorder of choice. If you were to record this situation with 32-bit float you would be able to capture the entire sequence while still preserving the audio, even if it started to peak the recording.
Speaking of noise floor, the other advantage of 32-bit float is that it lowers the noise floor. You could be recording something super quiet, bring the clip gain up in post-production and you would have your quiet sound nice and loud without any of the nasty hiss that is associated with the noise floor.
There are two things to consider when recording in 32-bit float. The first is that it only mitigates digital clipping, not analog. If your recorder has analog pre-amps and drives those too hard, that overdriven signal will remain when it gets converted to digital, and no amount of 32-bit float will preserve that recording due to the distortion coming from the analog side of the conversion.
32-bit float files are generally 33% larger than files recorded at 24-bit. In this day and age, this doesn’t really matter – computer storage sits in the realm of terabytes where most audio recordings barely exceed 100 megabytes making this a non-issue considering the trade-off being infinite headroom and an excellent safety net for recording.
As I mentioned above, 32-bit float should never be used as an alternative to learning proper gain staging however it is an extremely useful tool in any engineer’s kit and will no doubt save many recordings from being scrapped in the future.