The most well-known benefit of 32-bit float files is that they have vast headroom, such that signal levels in excess of 0 dBFS do not clip. However there is an important fidelity benefit for very low signal levels too.
We set up the following exaggerated scenario to highlight the differences between 32-bit float and 24-bit files. In normal operation, the difference would not be so stark. In a bag drop situation, extreme dynamic range applications like fighter jet flyovers, or with grossly incorrect gain settings, you can still have peace of mind recording in 32-bit float.
- Two MixPre-6 IIs, one set to 192 kHz, 24-bit, the other set to 192 kHz, 32-bit float.
- Two matched mics (Warm Audio WA14 set to cardioid pattern, -20dB pad), positioned coincidentally, each connected to mic input 1 on their respective MixPres.
- Both MixPre IIs set to mic input with lowest gain possible: trim gain set to minimum (6 dB), fader gain set to -50 dB, master left (L) gain set to -40 dB. This results in peak signal levels well below -100 dBFS, which are inaudible in the MixPre II’s headphones (even when set to max volume) and not high enough to be visible on its meters.
- Left (L) track recorded on both MixPre-6 IIs at the same time.
- Resulting 32-bit float file and 24-bit file imported into Adobe Audition for comparison.
- 24-bit file imported. Signal is low, appearing like silence.
2. 32-bit float file imported. Signal is low, appearing like silence.
3. 24-bit file normalized to -0.1 dBFS (>100 dB boost). Notice the excessive noise floor. Listen to the highlighted region (a piano tone).
4. 32-bit float file normalized to -0.1 dBFS (>100 dB boost). Notice the low noise floor. Listen to the highlighted region (a piano tone).