This display, of the new ITU-R loudness meter developed at CRC, is the result of CRC's collaboration with CBC/Radio-Canada. CRC is assisting Canada's national public broadcaster in the integration of the new loudness meter into its operations.

The detective creeps forward, gun drawn, barely a shadow against the dilapidated wall. Above him the clouds break, sending down a shaft of moonlight. He flattens himself against the wall knowing, just as you do, that if he's seen, if he's caught, the outcome will be death. The music, quiet and intense, builds slowly in the background. Suddenly, in the darkness, a twig snaps behind him. He whirls around ... and the program cuts to a commercial that sends you shooting from the couch. You, and a million other viewers worldwide, dive for the remote to turn the volume down.

If this scene sounds familiar it's not surprising. Radio and TV audiences routinely complain about jarring changes in loudness between commercials and programs; when switching from one station to another; even between different programs broadcast on the same channel. As of this spring, though, thanks in part to a new loudness meter developed by the Advanced Audio Systems group of Communications Research Centre (CRC), lunging for the remote may soon be a thing of the past.

The Highs and Lows of Digital Conversion

Controlling loudness levels in broadcast media has always been a challenge, explains Anthony Caruso, Director of New Broadcasting Technology at the Canadian Broadcasting Corporation (CBC), but with the advent of digital technology broadcasters realized that a solution had to be found.

"When we began to migrate television from analog to digital we realized that the audio signal couldn't be carried separately from the video signal. It had to be embedded in the video stream."

This created a problem, says Caruso, because broadcasters get product - TV and radio programs as well as commercials - from producers around the world, and what is an acceptable loudness level in one country may be too loud or too soft in another. With the audio embedded in the video stream, however, the only way to "turn down the volume" is to decode or decompress the signal, readjust the audio portion, then recompress the whole thing. Not only is the process expensive, but it degrades the signal - especially the video signal - to an unacceptable level.

By the year 2000, broadcasters realized that the solution was to control loudness at the production stage, and the best way to do this was for the International Telecommunications Union (ITU) to set a standard - a sound level that all members would adhere to when producing broadcast material. But for all the members to adhere to a standard, they needed a uniform way to measure loudness and that, says CRC's Louis Thibault, Manager of the Advanced Audio Systems group, is not as straightforward as it first might seem. Take Batman, for example: a scene where he blows up a car is significantly louder than the romantic interlude where he holds his love-interest in his arms. Viewers don't leap to adjust the volume, however, so how exactly is the audience deciding if a program is too loud or too soft? What the ITU needed, says Thibault, was a loudness meter that accurately mimicked how people perceive loudness when watching TV or listening to the radio. While there were several meters on the market that purported to do just that, the ITU required solid scientific data to determine which one did the best job, and that task fell to CRC.

Perceived Loudness

"What we needed to do," says Thibault, "was to compare the loudness values calculated by the meters against the subjective values of loudness perceived by real listeners, so we needed a database on how people hear radio and TV."

The ITU asked five institutions around the globe - including CRC's Advanced Audio Systems group - to act as test sites to create such a database. Subjects were brought into a specially designed listening room and asked to match the loudness of a series of audio clips to the loudness of a reference clip. Test clips were taken from real broadcast material provided by ITU partners and ranged from newscasts, to classical music, to gun battles. When completed, the test data from all sites were then sent to CRC.

"The first thing we did," explains Thibault, "was compare the data to see if all the different sites got similar results. Once we knew that they had, we averaged across all the labs and calculated a subjective loudness gain or loss for each particular clip."

In other words, they figured out how much the subjects turned the volume up or down to match a particular clip to the reference-clip volume. This gave them a subjective measure of the perceived loudness for each of the clips against which they could compare the objective measure calculated by the meters. If a meter was good, it would score the clips in the same way the study's human subjects had.

Testing Meters

With the database established, Dr. Gilbert Soulodre, the CRC researcher then in charge of the study, set about to test the 10 loudness meters that had been submitted to the ITU. But in addition to the 10, he threw in two of his own. His extensive experience dealing with objective and subjective audio testing told him that these two very simple meters - computer-based algorithms - would provide results that were at least close to those of a human listener and, if nothing else, would act as a baseline or reference to ensure that all the equipment and test meters were functioning as they should.

"The study's results - the comparison of all the meters - were presented at an ITU meeting in Geneva," says Thibault. "Gilbert presented his findings and when the committee saw them it was really quite clear. One of Gilbert's meters had a correlation of 0.98 with the listener results. A perfect match would be a correlation of 1.0, so 0.98 is incredible. Gilbert's meter was the best of them all."

Thibault says that what made the finding so surprising was that Soulodre's meter was so simple. While some of the meters modeled the complex interactions between sound waves, the ear and the brain, Soulodre's algorithm filtered out low frequencies from the loudness calculation and averaged the power of what was left.

"What it means," says Thibault, "is that when you're adjusting the volume on the TV your ear acts as a high-pass filter: it's less sensitive to the low frequency sounds. Because your ear is more sensitive to the higher frequencies, especially those between 100 Hz and 8 kHz - the dominant frequencies in human speech - you base your volume calculation on the loudness of these higher frequencies."

Public Domain

CRC published the findings and placed them in the public domain. This allows audio manufacturers, broadcasters and production houses - even small companies in poorer countries - to use the loudness meter without royalty fees.

In April 2006, the ITU adopted CRC's loudness meter as the international standard to measure loudness in broadcast productions. Integration into operations is ongoing.

Having the meter, says CBC's Caruso, will have a big impact on their operations. "The fact that we can now measure loudness is a quantum leap. The next big step will be adopting a loudness level that everybody agrees on." And that, he is hoping, will happen in late April 2009 at the ITU meeting. As part of the agenda, delegates will vote on a standardized level - most likely -24 LKFS - for all broadcast production. While the number may not mean much to the listening audience, the result will. If that vote passes, your days of scrambling for the remote will soon be over. Stay tuned.

For more information contact Louis Thibault, Research Manager, Advanced Audio Systems, at 613-990-4349 or louis.thibault@crc.gc.ca.