Understanding Sample Rate, Bit Depth and Dither

Should every session be recorded at 192 kHz? Does 32-bit floating point eliminate clipping? These questions are easier to answer when sample rate and word length are separated from marketing claims.

Sample rate and bandwidth

Under the sampling theorem, a sufficiently band-limited signal can be reconstructed when sampled above twice its highest frequency. Converter filters and real hardware behavior matter; higher rates also increase CPU load and storage.

Bit depth and noise

For ideal uniform PCM quantization, more bits increase theoretical dynamic range. Recording with modern 24-bit converters provides substantial usable headroom; real analog electronics generally set practical noise limits first.

What floating point can and cannot do

32-bit float files preserve large numerical headroom in suitable workflows, but cannot undo clipping at an analog microphone, preamp or converter front end. An ordinary 24-bit file recorded with sensible gain is often entirely adequate.

Why dither exists

Appropriately applied dither decorrelates quantization distortion when reducing word length. It is normally applied once at the final fixed-point conversion stage, not repeatedly throughout an unchanged floating-point mix.

Choosing a recording format

24-bit PCM at 44.1 or 48 kHz is a practical baseline for many music projects. Higher rates increase data throughput and may be appropriate for specific production workflows, but do not automatically improve audible quality. One minute of mono, uncompressed 24-bit/48 kHz PCM takes approximately 8.64 MB, excluding file overhead. 32-bit float files cannot restore overload that occurred in the analog input stage. Apply appropriate dither when reducing word length for a final fixed-point delivery format.