dBDigital audioReference

Audio numbers decoded: what 6 dB, 44.1 kHz, 24 bit, and Q actually mean

A parameter reference for recording engineers. Identify the reference and the conditions first, then decide whether the number matters in practice.

By · DADAO Acoustics Lab · DADAO AUDIO10 min read

Abstract comparison of four core audio parameters: amplitude, sample-rate waveforms, 24-bit quantization steps, and a bell-shaped Q curve.

Contents

What this article covers

6 chapters · About 10 minutes

Chapter01

Decibel

Ask for the reference before reading the dB value

A dB value on its own expresses a ratio between two quantities of the same kind; the suffix identifies the reference. In air, dB SPL usually refers to 20 μPa, dBV to 1 V, dBu to 0.775 V RMS, and dBFS sets digital full scale to 0.[1] Statements such as “1 dB is always just audible” or “10 dB always sounds twice as loud” are not fixed laws. Perception also changes with frequency, starting level, programme material, and the listening environment.

Figure 1 · Four common dB notations use different reference points; perceived differences still depend on frequency, level, and programme material.

Chapter02

Octave & Decade

Octaves and decades: two ways to state the same slope

An octave is a doubling or halving of frequency; a decade is a tenfold increase or decrease. The range from 20 Hz to 20 kHz spans about 10 octaves, or 3 decades. For pole-based filters, each order approaches a slope of about 6 dB/oct far beyond the cutoff, also written as 20 dB/dec; a second-order filter approaches about 12 dB/oct.[2] “Approaches” matters: the actual curve near cutoff still depends on filter type and Q, so a straight-line extension is not a substitute for measurement.

Figure 2 · High-frequency roll-off for first-, second-, and third-order filters beyond cutoff fc, shown on a logarithmic frequency axis.

Chapter03

Q & Bandwidth

The same Q value does not always produce the same curve

For a standard band-pass filter, Q equals centre frequency divided by the bandwidth between the two −3 dB points; higher Q means a narrower band.[3] Expressed in octaves, a one-octave bandwidth corresponds to Q≈1.414 and one-third octave to Q≈4.32. Parametric EQ implementations can still differ in peak gain, bandwidth definition, and interface conversion. When comparing plug-ins, do not copy the Q value alone. Match centre frequency and gain first, then compare the measured curves.

Figure 3 · Relationship between Q and octave bandwidth under the standard band-pass definition; verify the actual response of a parametric EQ.

Chapter04

Byte & Prefix

Byte, bit, MB, and MiB: the unit changes, not the file

One byte (B) contains 8 bits. SI units kB, MB, and GB use powers of 10; IEC assigns KiB, MiB, and GiB to powers of 2, so 1 MiB = 1,048,576 B.[4] A 1 TB drive (10¹² B) therefore contains about 931 GiB; no capacity has disappeared. Also distinguish sample precision from its storage container. WAVE can store 20 valid bits in a 24-bit container, so linear PCM precision is not limited to multiples of 8 bits.[5]

Figure 4 · SI decimal prefixes versus IEC binary prefixes: the difference grows from +2.4% at k to +12.6% at P.

Chapter05

Bit Depth

The practical value of 24 bit is lower quantization noise

SNR = 6.02 × N + 1.76 dB describes only the theoretical limit of an ideal N-bit converter with a full-scale sine input and quantization noise as the sole impairment.[6] Under that model, 16 bit yields about 98 dB and 24 bit about 146 dB; real converters are also limited by analogue noise, clocking, and circuitry. The additional 8 bits lower theoretical quantization noise by about 48 dB, but do not automatically create “24 dB more mix headroom.” The practical benefit of 24-bit recording is that conservative levels can still keep quantization noise very low. Record Safer Pro supports 16 bit, 24 bit, and 32-bit float; it is not fixed to a default 24-bit/192 kHz format.[9]

Figure 5 · Quantization SNR for ideal integer converters with a full-scale sine input; these bar values do not apply to 32-bit float.

Chapter06

Sample Rate

44.1, 48, or 96 kHz: start with delivery and processing

For a band-limited signal, sample rate must exceed twice the highest frequency; a practical system also needs transition space for its anti-aliasing filter.[6] 44.1 kHz inherited the time base of early video-based PCM recorders and later became the 16-bit CD sample rate, while 48 kHz is common in film, television, and broadcast workflows.[7][8] 96 or 192 kHz can give filters and some nonlinear processing more margin, but does not automatically improve the finished programme. One hour of uncompressed stereo occupies about 605.6 MiB at 44.1/16, 908.4 MiB at 44.1/24, or 3.86 GiB at 192/24; multiply again for additional recording channels.