Loudness Redux, Again

New audio standards complicate issue
Publish date:
Image placeholder title

Los Angeles
What? Loudness again? The topic of loudness, particularly commercial loudness, has been with us as long as has commercial television.

In the 1960s, the FCC made a major effort to study loudness, pursuant to enacting some regulations. While all this produced some valuable technology in the form of the CBS Loudness Meter, for example, the FCC finally threw up their hands and abandoned this effort owing to a lack of agreement on the exact nature of the problem, let alone how it might be controlled.

Excessive loudness is frequently blamed on commercial producers. They are certainly not without culpability. I don’t have to explain to a television engineer how compressing the peak-to-average ratio of audio material makes it perceptibly louder, without making an instantaneous audio level meter indicate any higher value. But it isn’t just that simple.

While the dynamic range of commercials is compressed, and broadcast networks have long made efforts to mitigate this on the air, there are other problems, and other culprits. Other problems include the extremely jarring transitions that can occur when soft dialog or a soft musical ringout suddenly is smashed by a blaring commercial.

I well remember a commercial for some kind of designer water that ran a few years ago. It had 5.1 channel sound, and every channel including the subwoofer channel was constantly smashing against 0 dBFS. This raises another difficulty we confront today. In the 1960s, the FCC only had to deal with mono television audio. Today, we have stereo and 5.1 channel sound as well, and we have to deal with switching back and forth between the two.


Particularly in the early days of HDTV, we frequently had 5.1 channel program material intermixed with two-channel commercial breaks. In this situation, we cannot just watch the meters in order to successfully control the relative loudness between two-channel and 5.1 channel audio.

Presumably, in 5.1 channel mode, the dialog appearing in the center channel (and possibly in the left front and right front channels) is mixed at the correct level. When two-channel dialog takes over, in a commercial for instance, if the left and right channels are run at the same levels as they are run for 5.1 channel sound, phantom center channel buildup is going to make the two-channel dialog perceptibly louder, regardless of whether any compression is going on.

Finally, there are, as mentioned, culprits other than the commercial producers. Here where I live (hint: the second largest TV market in the U.S.), even the local PBS station consistently runs commercials (yes, PBS stations sell advertising, in case you’ve been living off-planet) and promos that blare in comparison to the program audio.

Today, we have some better tools to measure and possibly mitigate the loudness problem than was the case in earlier decades. A subcommittee of SG6 of the International Telecommunications Union Radio-communications Sector (ITU-R) has extensively investigated the loudness issue in recent years, and has produced some recommendations on how to measure television loudness.

The subcommittee undertook to evaluate a number of potential loudness measurement algorithms, submitted by participants, by correlating the results produced by these algorithms with subjective test data collected at five sites around the world.

This work turned up a surprising finding. The results produced by a simple frequency-weighted mean-square measurement algorithm, which was initially included to establish a performance baseline, correlated much better with the subjective test data than any of the results produced by any of the submitted algorithms, many of which were far more sophisticated.

The initial work was done using single-speaker mono material, but when the research was expanded to include two-channel mono (with phantom center channel), and multichannel material, the results held.


A metering algorithm is the method a particular meter uses to make its measurements. While in the past we have largely been concerned with the measurement of instantaneous peak or average audio levels, today we have some new algorithms and some new types of meters that measure loudness.

Loudness measurement algorithms are used to measure average loudness over some period of time that may range from minutes to hours. Two familiar loudness measurement algorithms are LeqA and LeqM. Leq denotes equivalent sound level, the mean-square average level integrated over some period of time T, while A and M are frequency-weighting curves. Mathematically, a weighted mean-square level is calculated thusly:

Image placeholder title

W represents the frequency weighting used; xW is the signal at the output of the weighting filter; xRef is some reference level; and T is the duration of the measured audio sequence.

Translated into English, this means that the value of the equivalent sound level of an audio sequence measured with weighting curve W is calculated by dividing the square of the signal at the output of the weighting filter by the square of the reference level, integrating this value over the duration of the audio sequence with respect to time, dividing the integrated value by the duration of the audio sequence, and then converting the result to dB by taking the log10 and multiplying by 10.

Frequency weighting of the loudness measurement is a critical factor to assure that the measurement corresponds to subjective perception. We have mentioned two weighting curves, M and A. M weighting is used for measurement of film audio at cinema sound levels, which is too loud for typical home television viewing. A weighting corresponds to the frequency response of the human ear at conversational levels. It is probably the most frequently used weighting curve, and it has been cited in the ATSC DTV standard. Neither weighting curve is ideal for the measurement of television sound loudness.

The loudness algorithm for television audio sound that emerged from this work is specified in ITU-R Rec BS.1770—Algorithms to Measure Audio Programme Loudness and True-Peak Audio Level, issued in 2006. It is implemented by passing each incoming audio channel (disregarding the subwoofer channel) through a pre-filter that takes account of the acoustic effects of the human head, then through a simple high-pass filter called the revised low-frequency B (RLB) curve, deriving the mean-square energy in each channel, weighting each channel according to the angle of arrival at the listener, and summing the channels to arrive at a composite loudness measurement. This measurement, called Leq(RLB) in BS.1770, has been re-named LKFS, where FS denotes full-scale reference level, and K is a previously unused designator letter to designate the RLB weighting curve. LKFS is the standard measurement algorithm for television audio loudness, and if it has not already done so, LKFS will replace LeqA in the ATSC digital television broadcasting standard.

The television audio loudness issue is about to come to a head. As the industry endeavors to establish a program of voluntary compliance to loudness control based on an imminent ATSC recommended practice for loudness measurement using the LKFS algorithm and a specific dialnorm number, a commercial loudness bill is moving through Congress. If passed into law, the Commercial Advertisement Loudness Mitigation (CALM) Act would give the FCC a year to adopt the ATSC RP, and would give the industry another year after that to purchase and install the equipment necessary to comply. The ATSC RP is expected to be released in this month. Stay tuned, but watch the volume!

Randy Hoffner, a veteran of the big three TV networks, is a senior consulting engineer with AZCAR. He can be reached through TV Technology.