A Systems Approach to Implementing The CALM Act
The CALM Act is a
reality. So how do
you go about implementing
systems to ensure
“The most important
thing is applying
the golden rule—average
equal the dialnorm [metadata] value that
you transmit,” said Jim Starzynski, director
and principal audio engineer for NBC Universal
(NBCU) Advanced Engineering, and
chairperson of the Advanced Television
Systems Committee (ATSC) technical subgroup
S6-3 on digital television loudness.
Starzynski should know, and not only
through his work with the ATSC. For the
past few years, he has been architecting
and—with the help of numerous NBCU
engineering and operational teams—
implementing the various systems and
workflows throughout NBC Universal to
accurately measure loudness and make
corrections when necessary.
Starzynski shared some examples and
offered advice to help those networks,
groups and stations just starting to come
to grips with this issue.
ESTABLISH COMPANY STANDARD
Fig. 1: Content is delivered and played out at NBCU's standard target loudness or is loudness-normalized using automated workflows with file-based scaling or targeted processing. All achieve 24 LKFS and yield smooth loudness transitions between all types of content.
First, it’s important to have a good plan.
Establish a company loudness standard.
This is the average loudness for all mixed,
delivered and ingested content, also referred
to as a loudness target value. A station
may wish to conform to the loudness
standard of its affiliated network or networks,
or develop its own standard target
value. Either method will work as long as
the golden rule is applied, Starzynski said.
Average loudness must equal dialnorm.
In the first case, dialnorm metadata
management becomes much simpler, as
only a single value of dialnorm has to be
sent to or entered in the AC-3 audio encoder
for transmission for all content.
If the local standard target value is different,
the station must be able to switch
the correct dialnorm value (local or network)
for the AC-3 encoder. This adds
more complexity to systems design and
NBC Universal established –24 LKFS as
the company-wide standard.
According to ATSC A/85 “ATSC Recommended
Practice: Techniques for Establishing
and Maintaining Audio Loudness
for Digital Television,” LKFS means “loudness,
K-weighted, relative to full scale,
measured with equipment that implements
the algorithm specified by ITU-R
“A unit of LKFS is the same size as a
decibel, but is acquired with specific filtering
that models the human hearing system,”
A –24 LKFS is also the target loudness
value that A/85 recommends “for delivery
or exchange of content without metadata
(and where there is no prior arrangement
by the parties regarding loudness).”
A HIERARCHY OF SOLUTIONS
Analyze where content comes from
and in what form.
“Different types of content may require
different workflows,” Starzynski said.
ATSC document A/85 provides a hierarchy
of solutions. In addition, its quick reference
guides describe how to measure
loudness for different scenarios, and how
loudness levels can be corrected. (www.atsc.org
Workflow Scenario 1: Require content
providers to deliver
to the standard target
First on the hierarchical
providers to deliver
content that meets
the company standard
target value, and if it doesn’t, ask for
NBC Universal has issued technical
specifications, including loudness that it
requires content providers to adhere to.
These are openly available on the Internet
loudness spec is based on A/85 recommendations
for obtaining loudness measurements
for both long- and short-form
In particular, NBC requires that longform
content program providers deliver
an average of “–24 LKFS (±2 dB) dialog
loudness, not whispered or shouted,
for the soundtrack” by using an ITU-R.
BS.1770-based broadcast loudness meter.
This average value must be consistent for
all acts of the program.
For short-form content such as commercials,
the loudness measurement must
be taken using all soundtrack elements,
not just dialog, with all deliverables providing
the required average of –24 LKFS.
In either case, “minor variation of ±2
dB about the –24 LKFS loudness average
is anticipated and acceptable,” according
to the spec.
Primetime episodic programs are ingested
and reviewed in Edit 10 of NBC’s West
Coast facility. If it doesn’t meet NBC’s criteria,
the specific problem is discussed with
the supplier and a replacement is requested.
Starzynski said this long-form programming
is currently checked by a person who isolates
dialog in the show and measures its
average loudness. In the future this process
may become automated.
Workflow Scenario 2: Establish processes
to fix file-based content—Scaling Loudness
If short-form content is not delivered
to spec, then it’s vital to establish procedures
to fix the loudness level of the content.
Fig. 2: In addition to use of a loudness specification for deliverables and file-based scaling, loudness management for all content for NBC cable is ensured by use of real-time processing as documented in A/85.
One method is to scale the content to
the company standard target loudness,
which is relatively easy to implement in
a transcoder for file-based content. NBC
Universal applies this approach for the
NBC Television Network (Fig. 1), and all
NBCU cable properties at its Network
Operations Center (NOC) in Englewood
Cliffs, N.J., (Fig. 2).
File-based inbound short-form content
made up of commercials, promos, political
ads and PSAs (accounting for about
85 percent of the total) from commercial
aggregators and content providers including
in-house, are run through a serverbased
transcoder to convert files from
MPEG-2 to GFX-wrapped MPEG-2 used by
At the same time, “the transcoder takes
the original, delivered short-form file and
does an ITU-R.BS.1770 global [all elements
of the soundtrack, not just dialog)]
measurement of it head to toe, and compares
it to the target value of –24 LKFS,”
Starzynski said. “This is an automated process.
The transcoder determines the loudness
difference and applies an offset to
match the target. It’s a one-time volume
shift done in a linear fashion and doesn’t
affect dynamic range or quality.”
Workflow Scenario 3: Establish processes
to fix linear content—Real-time Target
Linear tape-based content is handled
in a different manner for NBC’s ownedand-
operated stations. For this scenario,
real-time target-based processing is used
at NBC Universal’s television station hub
The hub is the nucleus for shared content
among the 10 NBC O&Os and 15
Telemundo O&Os. Commercials and syndicated
programs for all the O&O stations
arrive and are managed at the hub, which
functions as a remote master control.
Once ingested and managed for loudness,
the hub provides all station short- and
long-form content that is not locally originated.
Taped-based content is digitized and
the loudness is normalized to the company
standard of –24 LKFS. To do so in real
time, a processor applies dynamic range
control to achieve the target loudness level.
While file-based scaling is transparent,
real-time target processing isn’t. “It does
alter the original content to some degree,”
Starzynski said. “It changes the dynamic
range, but the result is very, very good.”
In addition, all content for each NBC
Universal cable channel goes through a
real-time target-based processor in the
“In some cases it’s superfluous,” Starzynski
said. Commercials are normalized
as they come in, but other long-form program
material, especially from the archive,
isn’t yet, so a processor in the chain, just
prior to AC-3 encoding at the Englewood
Cliffs NOC, is used to manage that material.
Regardless of the workflow scenario,
Starzynski said that the end result in each
case creates an even playing field for all
suppliers using the practices in ATSC A/85
and by doing so, meets station and MVPD
compliance with the CALM Act.
Mary C. Gruszka is a systems design
engineer, project manager, consultant
and writer based in the New York metro
area. She can be reached via TV Technology.