UNDERSTANDING
BROADCAST
METERING SPECS
Let me qualify this before you
stop reading: I don’t think loudness
metering is inappropriate for games.
What I think is inappropriate is the
specification most people keep
talking about: EBU-R128. I’ll come
back to why I think this is wrong for
games, though, because I want to
talk first about the specification you
should be focusing on instead.
ITU-R BS.1770 is a metering
specification embedded with three
important qualities. First, it employs
a scale that emulates human
perception of the frequency spectrum
(thanks to the K equalization
curve). Second, that measurement
takes place over a period of time,
because the duration of a sonic
event affects the way we perceive
it. Third, it specifies a need for “true
peak” metering, a way to calculate
intersample peaking and to correct
for failures of Peak Program Meter
(PPM)/Quasi-PPM integration time.
These are the elements that are most
important to us as listeners. Any
meter that is ITU-R BS.1770-compliant
will meet these requirements.
The awesome thing about
BS.1770 is its emulation of human
perception. This method comes closer
to measuring audio in the way we
hear than any previous meter does.
Let’s say you had two sound files
that measured - 30 LKFS (that’s
“Loudness, K-weighted, relative to
Full Scale”; for those who are more
familiar with the nomenclature LUFS,
the two are analogous). One file has no
frequency content above 300Hz, and
the other is broadband and noiselike.
That equal loudness measurement
means that when played through our
speakers, the noise is perceived as
being equal in volume. Those same
two sounds adjusted to the same
meter reading in another scale, dBFS
or LEQ-A for example, will likely not
be perceived as the same volume.
This measurement scale provides
subjective predictability, regardless of
spectral content, which is a good thing.
WHY BROADCAST
MODELS AREN’T
GOOD FOR GAMES
Now let’s talk about EBU-R128
(and lump in ATSC-RP A/85, for
good measure). These broadcast
specifications are implementations
of ITU-R BS.1770. They do not specify
metering criteria, other than the
fact that BS.1770 is required. They
specify measurement criteria. What
they state is that a specific loudness
measurement, over “infinite”
duration, must be met by broadcast
materials. (“Infinite” simply means
that the measurement must be
averaged over the length of the
program or advertisement.)
NORMALIZING
LOUDNESS IN YOUR
ASSETS
Remember why BS.1770 is so
awesome? That predictability of
perceived loudness could be the
goose that lays a golden egg of
workflow efficiency. Let’s say that
your game’s sonic assets fall into
the following categories: ambient
sounds, small weapons, character/
vehicle movement, music, and user
interface. Start by defining a target
loudness measurement for assets
in each category. If you can predict
that all of your ambience assets
are the same perceived volume,
wouldn’t it make it easier to mix and
transition from one ambient area to
another? What if, months after your
initial implementation of weapon
sounds, you had to create three
new weapons? Or perhaps you’ve
been working with placeholder
music and have just received the
final versions. If you knew that all
of your existing weapons measured
at - 27 LKFS, you could easily create
new sounds for implementation into
the existing mix engine.
“SHORT-TERM”
MEASUREMENTS AND
LOUDNESS RANGES
PROTOT YPE 2 audio director Rob
Bridgett recently shared some
interesting numbers from his work.
Here are some “long-term” loudness
measurements from specific states
of the final mixed game: cutscenes
- 23, stealth and dialogue-heavy
gameplay - 23, action - 19, “insane”
action - 13. These states were mixed
artistically for maximum impact and
experience, which is important for
the artistic nature of games. Picking
one infinite loudness measurement
to suit this game would not only be
a difficult task, it would lessen the
audio team’s creative contributions.
One solution is to work in smaller
measurement windows and to
utilize a minimum and maximum
measurement range—a pair of
loudness thresholds.
When I say smaller
measurement window, I’m referring
to somewhere in the neighborhood
of five seconds—maybe as short as
two. In this manner, the loudness
measurement will more accurately
reflect the in-game events as they
happen. Because I’m arguing for the
use of dynamic range in games, a
game developer | November 2012 24
