Default Technical Specifications for Digital Audio Files

This document describes default digital audio file specifications to use if there are no other considerations or
requests from collaborators.

Once finished mixing, at least two audio files are created: the ​mix​ and the ​delivery​. The mix is intended to be
retained in your own file system as an archive of your work. The mix is saved at a high quality so that
multiple delivery audio files can be made with varying specifications. The intention of the mix is to retain as
many options as possible when creating delivery audio files. The intention of a delivery is to create an audio
files for a specific person and usage. At the mix stage you are considering your own needs. At the delivery
stage you are considering the needs of another person.

Once finished mixing in a multitrack digital audio workstation:

● Export a high quality mix audio file.
● Save the mix for your own records.
● Process a copy of the mix to meet a standard of loudness and peak levels.
● Export delivery​ ​audio files to deliver or distribute.

Mix:
A high-quality, stereo, uncompressed audio file. This file is to be saved on your hard drive as an archive of
your work.

Peak level:​ -4 dBFS (Maximum)

Loudness:​ -12 LUFS to -24 LUFS

Metering:​ K-14 and mix to average “0.”

Filetype:​ Stereo, interleaved, broadcast WAV files

Sample Rate:​ 48,000 Hz(if will be sync’d to video) 44,100 Hz (Just music)

Wordlength(Bit Depth)​: ​24

Dither: ​triangular(flat)

Filename: ​A​ ​thorough description of the contents of the file. The premaster is for your records, it
must be easily found in the future. Include the class name, assignment name, song name, version,
and anything else you would need to find it easily in the future.

CW191_Homework1_BlueBossa_Mix.wav

No spaces, underscore dividing components, dot only used once before the file extension,
“camel case” within each part of description.

Normalization: ​Export without normalization to retain this level in the mix audio file.
Delivery(master):
A stereo audio file created by processing and exporting the mix. When created to be delivered to another
person, consider that person’s needs then process and export a file that has a suitable loudness, type, and
name.

Peak level:​ Generally do not go above -1 dB FS TP (Full Scale True Peak). In practice there is much
variation. Analyze reference tracks to determine comparable standards and process the mix
accordingly.

Integrated Loudness:​ -8 LUFS to -15 LUFS

Length: ​Start immediately and end naturally. Use short fades to avoid audible clicks at the beginning
and end.

Filetype:​ Stereo, interleaved, broadcast wav files.

Sample Rate:​ 48,000 Hz is required when synchronizing to picture. 44,100 Hz is often used for
music not being synchronized to picture.

Wordlength(Bit Depth)​: 24 bit

Dither: ​Apply triangular(Powr 1(no noise shaping) dither when exporting after any digital signal
processing has been applied.

MP3:​ After all the necessary processing and editing is completed, if a lower filesize is needed mp3
files may be used.

192 kbps, VBR off, best encoding, HP filter off, normal stereo mode.

An mp3 should NEVER be used in a project again, always work with original uncompressed
files. MP3 is a lossy compression suitable only for listening.

File names:​ When creating a file name consider how the recipient will use the file. File names must
convey all the information that the recipient needs to do his or her job.

For instance: Consider your professor, he or she needs to know many things about the
audio file: your name, the assignment title, project name, version number, description of
file(score rough mix, pre-master, master).

class name_assignment name_song name_description of file_student name.file

extension

CW191_Homework1_BlueBossa_delivery_LStearns.wav

No spaces, underscore dividing components, dot only used once before the file
extension, “camel case” within each part of description.
Peak level and loudness Discussion
The following section is a longer discussion about peak levels and loudness, and an explanation of all the
terms in the level and loudness suggestions.

At the very basic, we are talking about an average level and a peak level.
● How do these levels relate to each other and to digital maximum?
● How do we calculate and monitor these levels?
● How do we talk about these levels?
● What guidelines should we follow during recording, mixing, and mastering?

Levels in digital audio, from loud to quiet:

● Digital max: 0dBFS
● True Peak of program(often greater than digital max)
● Sample Peak of program
● Average of program(RMS, LUFS, K-weighted, VU, etc…)
● Noise Floor of program
● Dither applied to the rendered audio
● Digital minimum

Bit Depth and dynamic range:

● Human hearing has a 120 dB dynamic range.
● The dynamic range of a digital system is 6 times the wordlength :
○ 16 bit * 6 = 96 dB dynamic range
○ 24 bit * 6 = 144 dB dynamic range
● There is a 48 dB difference in dynamic range between 16 and 24 bit digital audio.
● In a 24 bit workflow, recording with peaks at -48dBFS is the 16 bit equivalent of peaking at
0dBFS.
● In a 24 bit workflow, one could peak at -24dB FS and have an equivalent dynamic range of
human hearing.
● 24-bit is 48 dB better than 16-bit

Defining Loudness:
● “Loudness” is a human perception: something ​is​ a specific amplitude, something ​seems​ to
be loud.
● The perception of loudness changes based on amplitude, duration, and frequency.
● A very short transient of a high amplitude will not seem loud when compared to a longer
duration tone at the same amplitude.
● A tone of equal amplitude at 100 HZ and 3000HZ will have very different
loudness(3000HZ being louder). Research “equal loudness contours.”
● Dynamic effects like limiting and compression can increase loudness while keeping peak
level the same. This is accomplished by manipulating the duration of sounds.
● A calculation of loudness takes into consideration amplitude, frequency(pre-filtering) and
duration(averaging).
● Loudness within a digital system is measured in “Loudness Units Full Scale”(LUFS).
 ● A 1 LUFS gain change is equal to a 1 dB gain change. LUFS can often be treated as dB, but
we must know that LUFS is referring to human perception.
● Since duration is critical in our perception of loudness, the calculation of LUFS must be an
average over time. There are three time ranges used in this calculation:
○ Integrated Loudness(Program Loudness)- Loudness calculated over the entire
duration of the file.
○ Short term Loudness - averaged over a 3 second interval
○ Momentary Loudness - averaged loudness over a .4 second interval.
● Loudness range(LRA)- distance between the maximum and minimum momentary loudness
over a range of time, ignoring moments of silence: Contemporary pop mixes have an LRA of
2-7 LU. Classical music an LRA of 8-23 LU.
● There are many “averaging” meters: VU, RMS, K-scale, LU.

Peak Level and loudness suggestions

Recording(24 bit)
● Peak level: -6 dBFS (Maximum)
● Loudness: -14 LUFS to -30 LUFS Momentary (depending on what is being recorded and
how it will be used in the final mix).
● Metering: K-20 and record to average “0.”
This may seem low, but because of the wide dynamic range of 24 bit audio, this is still accessing a much
wider range than a 16Bit recording can.

Mix
● Peak level: -4 dBFS (Maximum)
● Loudness: -12 LUFS to -24 LUFS
● Metering: K-14 and mix to average “0.”

Delivery
● Peak level: -1dBFS True Peak (Maximum)
● Loudness: -8 LUFS to -15 LUFS

Suggestions and Clarifications

True Peak Full Scale:
True peak meters use a higher internal sample rate when calculating peak level. This accurately
represents how digital to analog converters will react to sampled audio. Positive true peak full scale
levels are moments where the digital to analog converter will go above digital maximum and distort
even when the sample peak level is below digital maximum(it is impossible for the sample peak to be
above digital maximum).

Peak to Loudness Ratio(PLR):

 The distance in dB between the true peak and integrated loudness is a reasonable measure of the
micro-dynamics of a program and gives a clear indication to the amount of audio compression
applied.

In mastered commercial releases the PLR falls between 14dB(uncompressed, acoustic, “natural”)
and 7dB(highly compressed “loud”).

A mastered track with a true peak of -1dBFS, would require an integrated loudness between -8
LUFS and -15 LUFS. With a different true peak, the loudness changes proportionally. This is
confirmed by analysis of commercial releases, for loud mixes we will often see a true peak over
0dBFS with an integrated loudness of -7 LUFS(Adele, “Rumor Has it” has a 7 dB PLR).

Metering:
These guidelines require metering that can display loudness units(LU). Loudness meters often
require the configuration of a “target” to be adjusted based on your point in the process and the
aesthetic goals. A “target” of 0 will make all the meters correspond to LUFS (Loudness Units Full
Scale). A target of -14 LUFS on a loudness meter is the same as the common K-14 standard
metering.

In Protools: setting the meters to K-14 will display loudness units where -14LUFS = 0 on the K-14
meter. The idea is to have a loudness “target” in mind for your task, say -20 LUFS when recording,
then recording so that the meters read near 0. For recording a meter configuration of K-20 is
appropriate(metering pre-fader), change to K-14 for mixing(metering post-fader).

Dither:
● Contemporary DAWs have a 32 bit floating point bus, exporting to 24 bit represents a huge
reduction in wordlength and necessitates the application of dither.
● Dither avoids quantization distortion in a 24 bit render. A -144 dB sine wave test tone
rendered to 24 bit without dither produces a high number of harmonic and inharmnonic
partials(quantization distortion). When rendered with dither no additional partials are
present.
● Dither increases the dynamic range of a 24 bit export. A -150 dB test tone is audible when
exporting 24 bit with triangular dither, the same tone renders as silence when rendered
without dither.
● Triangular “flat” dither is best for most circumstances.

Calibrating monitors:
Configure a test tone of -20 LUFS mono pink noise within the computer then adjust the monitor
level to be 83 dB SPL C-weighted Slow at the listening position. This is the “0” point for the
monitors. The lower the monitoring level the more compressed the music will need to be to achieve
a comfortable(ideal) listening loudness. To master pop music, set the monitors 15dB below the
calibrated “0” reference, to master jazz, -6.

Loudness Normalization:
Streaming services like iTunes, Pandora, Spotify and Youtube all perform “Loudness Normalization.”
This involves first analyzing a track to find the integrated loudness, then adjusting the playback
level to meet a standard. Because of this process, highly compressed “loud” music is automatically
reduced in level before playback. Loudness differences between songs is automatically
compensated for by the streaming service, so there is little benefit for trying to be “louder” than the
other tracks, the more the music is compressed, the more the playback system turns the level down.
This process is also done on European radio, and will be followed in American radio soon.