As the FCC gets closer to unveiling the technical specifications that TV stations will have to meet under the CALM Act to eliminate dramatic shifts in volume between programs and commercials, equipment vendors are readying the needed monitoring, control and verification gear. Here's an overview of some of what's available now.
LOUDNESS, PART II
Vendors Turn Up Volume On Loudness Gear
Vendors feel the loudness buzz.
Since the CALM Act was signed late last year, “the phone has been ringing off the hook” for Lancaster, Pa.-based audio processing specialist Linear Acoustic, says field applications engineer Hal Buttermore. “If stations had been ignoring it, they’re not anymore,” he says.
The calls may become even more frequent and insistent. Last Friday, as expected, the FCC launched a proceeding to write the implementing rules for the law, which is aimed at eradicating once and for all the annoying phenomenon — intended or unintended — of the volume sometimes rising considerably when TV programs cut to commercials and promos.
Broadcasters and vendors will not know exactly what is needed to comply with the rules until the end of the year, by which time the rules are supposed to be written and adopted.
But they have a pretty good idea of what it will take. So broadcasters may not yet be buying the needed monitoring, control and verification gear, but as Buttermore points out they are shopping. And vendors like Linear Acoustics, Harris, Wohler, Miranda, Evertz, Ensemble Design and Volicon are ready with solutions, which they showcased at the NAB Show in April.
[For an overview of the CALM Act and the loudness problem, read Part I of this special report, “Despite Lack of Regs, TV Turns Down Volume.”]
BRAND CONNECTIONS
At NAB in April, Linear Acoustic introduced the fourth generation of its Aero.air transmission loudness manager, which fits in a two-rack unit chassis and has more processing power and features than previous models, such as offering Nielsen watermark audio encoding.
The top of the line $21,750 Aero.air 5.1 version is a 10-channel device that can process three audio streams and upmix to 5.1 audio from a stereo source. At NAB, Linear Acoustic also showed updated versions of its Aero.one loudness manager, which ranges from $7,000 to $13,000, and LQ-1000 loudness quality monitor, which uses the ITU BS.1770 spec and lists for $4950.The company also showed new versions of its Aero.qc and Aero.file file-based loudness controllers aimed at ingest applications.
With an eye to future compliance requirements, the LQ-1000 can log and store a week’s worth of loudness data and easily export logging data to external media.
The new LQ-1000 has an OLED multicolor display that gives an instantaneous readout of the loudness value or LKFS — blue if the loudness value is too low, yellow and red if it’s too high shown and green if it’s on target.
The unit can also display three different “integration times,” which is the length of the clip that is being measured for loudness. Increments can range from 10 seconds to 30 seconds to infinity.
“For different types of content, you want to measure different integration times,” says Buttermore. “For commercials, you want a quick integration time. But for sports or movies, long-form programming, you want a longer integration time to measure the average loudness of the program.”
Another vendor targeting the loudness market is Harris, which introduced at NAB the APM6803+ Multichannel Audio Processing Station, a multi-function product that uses loudness correction algorithms from audio specialist DTS Neural.
While Harris has sold loudness correction tools for years in the form of 1-rack-unit boxes, the APM6803+ is actually a complete audio processing system on a card, including Dolby encoding and decoding and upmixing/downmixing. It can be integrated into other Harris modular products such as its NetVX encoder and new Selenio “media convergence” platform. The software-based APM6803+ system can range from $2,500 to $10,000 depending on the number of program streams and audio channels supported.
“That’s the beauty of us moving everything over to a software-based platform,” says Stephan Gauthier, Harris product manager for modular signal processing. “We can process multiple streams on a single card, do the 5.1, a 2.0 and a 2.0 [stereo channels] and a mono channel, and can encode it on a single card.”
Gauthier is interested to see what loudness regulations the FCC comes up with, particularly the time period stations have to measure and log loudness.
While stations are free to run different loudness levels for different types of programming and don’t have to adhere to the -24 LKFS benchmark, Gauthier expects that most stations will adopt the -24 LKFS setting for all of their content as a no-brainer way to be compliant.
“The feedback I’m getting is that people want to set it and forget it,” he says. “If it means content has less dynamic range, so be it. Nobody is going to call and complain that the quiets aren’t quiet enough and that the louds aren’t loud enough.”
But Gauthier does have a word of caution: broadcasters could actually use too much loudness correction and significantly affect the audio quality.
“If a network puts loudness correction on 24 hours a day, and the stations are also running correction on the feed, then you’ve got two people correcting loudness and you start really shrinking the dynamic range,” he says.
The quality impact from unnecessary loudness correction is similar to the cascading compression effect that video gets from being compressed and re-compressed through the transmission chain. With that in mind, Harris is developing hooks to its automation software so the loudness processing can be turned off when a station knows it has good-quality content coming into its plant.
In addition to the AMP6803+, Harris also sells a loudness logger and monitor, the Videotek LLM-1770, and a file-based loudness correction system, the Harris QuiC. Gauthier suggests that to ensure the best audio quality, stations should use file-based correction whenever possible.
“There’s a big difference between file-based loudness correction and real-time,” he says. “Because you have access to the entire clip, you’re able to change the loudness of that clip simply by applying a single gain correction to the entire clip, and you haven’t changed the dynamic range at all.”
Real-time loudness correction has some inherent limitations, says Jeff Riedmiller, director of the sound platform group for Dolby, which designed the DTV audio system to achieve a “Hollywood-to-the-home” audio experience.
“In a real-time loudness controller, you have a short memory,” says Riedmiller. “So the only approach would be to really hammer the daylights out of the dynamics, and constantly change those windows. Obviously, that would have a huge impact on the dynamics and spectral balance.”
Dolby is also concerned about the cascading correction problem. “Ultimately, to minimize that effect the best thing you can do right now is to try to use the processing as far upstream as you can get it,” says Riedmiller.
Dolby has sold loudness-focused hardware for years, including the LM100 loudness meter and the DP600 file-based loudness analyzer and correction tool. The LM100, which has been updated to support the BS.1770 spec, costs $3,000, while the DP600, which has already been used to correct the loudness of some three million cable spots, costs $17,500.
At NAB, Dolby announced it now will license its loudness metering and correction software to third-party manufacturers looking to make products to help broadcasters comply with the CALM Act. Most notable is Dolby Media Emulator, a loudness-estimation engine with Dolby’s Dialogue Intelligence technology.
Video and audio monitoring specialist Wohler introduced its own loudness measurement tool at NAB, the Pandora Loudness Analyzer, a compact device based on Apple’s iPod platform.
The two-pound desktop unit, which can also be rack-mounted and sells for around $2,000, uses an iPod Touch as its touch-screen interface that can be updated in the future through the iTunes store. It accepts 3G/HD/SD-SDI and AES inputs, and can measure up to eight channels instantly with adjustable over/under alarms.
Pandora uses the BS.1770 spec to measure loudness for user-defined period of time, ranging from 400 milliseconds to 60 minutes, and stores logs of each session for compliance purposes.
“That’s important,” says Jeff McNall, director of product line management for Wohler.“Someone goes to lunch and then you get a complaint; that’s a way to fight the ticket.”
Wohler’s broadcast customers aren’t rushing to invest in loudness tools yet, says McNall, but most want to make a decision on equipment this summer.
He tells customers that while not every channel may need loudness control, every one will still need loudness monitoring. He expects customers will buy measurement gear first to see how big a problem they have before investing in more expensive loudness-correction tools. Customers are looking for direction from the FCC.
“People are hesitant to make this investment when they don’t know what the final rules are going to be,” says McNall. “I don’t expect a purchasing rush to happen until two weeks before [the rules] have finally taken off and been mandated.”
Infrastructure, routing and monitoring supplier Miranda has its own loudness-logging tool under development, but is waiting on the FCC rules before finalizing it.
Guy Marquis, senior product manager for Miranda’s infrastructure group, thinks broadcasters are being prudent in holding off on buying logging products. “There are a lot of companies that offer logging, but all the manufacturers still don’t know what is required by the FCC,” he says. “Most of these things are software-upgradeable. But if you’re buying something today that will do loudness logging, the chance you will need to do an upgrade is very high.”
Miranda already sells loudness-correction tools with its line of real-time Automatic Loudness Control (ALC) products, which run on its Densite modular cards and are designed to perform loudness control on the broadcast signal as it goes to playout.
The ALC options include two higher-end models that use loudness-control algorithms from Linear Acoustic (multiband AEROMAX processing) and the German firm Junger Audio (Level Magic processing), which sell for around $4,000 for an 8-channel system, as well as a more cost-effective homegrown Miranda card with wideband processing that starts at $995 for a two-channel system.
“We take a wideband approach to loudness control,” says Marquis. “We don’t split the signal into five bands; we just apply loudness correction to the entire band.”
At NAB, Miranda also introduced a high-density loudness correction product, Axino, aimed at cable, satellite and IPTV operators and designed to be installed in a headend or uplink center. It is a powerful IP-based platform that is capable of grabbing more than 100 program feeds, and can fix commercial loudness problems as well as excessive channel-to-channel and program-to-program variation.
Like other vendors, Marquis points out that loudness correction doesn’t have to be performed if stations are willing to dynamically adjust the dialnorm value on their audio encoders to match the loudness of content. He says that one of Miranda’s Canadian broadcast customers, Tele-Quebec in Montreal, actually uses such “agile dialnorm management” on a regular basis. But such stories are few, and Marquis expects that most U.S. broadcasters will wind up settling on static dialnorm matching the target loudness of -24 LKFS.
“In theory it’s possible to use, in a broadcast environment, agile dialnorm management as opposed to static,” he says. “But there’s a high risk of losing the dialnorm value from the ingest point to the playout point. Servers might not support it, or perhaps somebody in production played with the levels for a good reason. But there’s always a reason why it could go wrong.”
Another vendor with an established loudness-correction product is infrastructure heavyweight Evertz, which has been selling its card-based Intelligain Loudness Processor for the past four years and has over 3,500 installs to date worldwide.
Like other Evertz products, Intelligain was engineered in-house with the company’s own algorithms, and it can now be incorporated in a range of gear including master control switchers, frame-synchronizers and the like.
“Dolby-E and Dolby AC-3, we OEM that. Everything else is organically done by the company,” says Tony Zare, Evertz product manager for modular products.
IntelliGain was designed around the BS.1770 spec from the beginning, says Zare, who has spent a lot of his time educating customers about loudness prevention. “Since 2007, I’ve been traveling the world talking about loudness,” he says. “A lot of people weren’t up to speed with what 1770 is about. But now it’s become more mainstream.”
The three-part IntelliGain system measures loudness first, and only performs a correction if the loudness doesn’t match a facility’s target, says Zare. “If the input loudness already matches the target, then IntelliGain just becomes a wire that the audio is passed through, so you can maintain the dynamic range and maintain fidelity.”
Zare believes that IntelliGain’s performance is competitively on par with other loudness-correction systems, but that Evertz’s advantage is in the “hook and handles” into other Evertz products and the ability to do integrated monitoring with Evertz’s existing VistaLINK network management software.
At NAB, Evertz released the IntelliGain Compliancy Logger, an optimized piece of software that will capture loudness data for proof of compliance. While the FCC has yet to finalize its compliancy rules, Zare is confident the new logger will be able to support the new requirements and says it can be easily integrated into existing plants. “Anybody using VistaLINK is already 95% of the way there,” he says.
Looking at the base of existing IntelliGain users, Zare estimates that 40% of loudness correction is performed at ingest. With that in mind, Evertz is integrating file-based products for loudness control that can work in the compressed domain.
Another infrastructure and signal processing supplier, Ensemble Designs, has updated its LevelTrack Automatic Gain Control product to support the BS.1770 spec in addition to traditional VU measurement. The product, which works with the company’s Avenue signal processing modules, runs around $3,000 per program stream. It applies gradual changes to the overall level of an audio signal.
With a nod to the CALM Act, Ensemble Designs has also developed Audio Compliance and Monitoring Software as an option that can be added for compliance verification and archiving. The data logged through the new compliance tool can be accessed remotely through Ensemble’s Avenue PC control software.
“Up to this point, stations have just had to deal with viewers and make sure they’re happy,” says Ensemble Designs marketing czar Cindy Zuelsdorf. “Now with the CALM Act, they have to get into compliance logging and maintaining the loudness data.”
Monitoring specialist Volicon is targeting that new requirement with Observer Loudness Monitoring, a new feature in its flagship Observer product that provides continuous measurements of program loudness (as per BS.1770) and true-peak signal levels.
Observer customers generally maintain a content archive ranging from 30 days to two years, along with all the associated metadata the Observer system captures, says Ed Hauber, Volicon director of sales. So the new loudness logging tool can easily pair metadata relating to loudness values and associated dialnorm numbers with an actual clip of the content itself, Hauber says, along with basic metadata regarding channel, date and time.
“When a user is told of a discrepancy, they can use this tool to go back to a specific daypart and time, see the values pinned to that media, and export a forensic copy of the media as it was aired,” Hauber says.
In Part I last Thursday, TVNewsCheck offered an overview of the loudness problem and the CALM Act. Read it here.
Comments (6)
len Kubas says:
June 2, 2011 at 1:12 pm
Excellent article, Glen, and including list prices is very helpful. LKFS, dialnorm and BT.1770 are the keys. By the way, the chips are starting to fall into place: the FCC’s NPRM last week, and 12 minutes before receiving the Tech Thursday email, my inbox received official notice that the ATSC has published the revised A/85 that will be used by the FCC to fashion regulations.
Scott McDaniel says:
June 3, 2011 at 10:41 am
In this current day mentality it is obvious that we have to bring in very sophisticated machines to keep our broadcasted material human for consumption. No body wants to assume the on line minute to minute responsibility of their output through the diligent use of common sense and human involvement and monitoring. There use to be individuals involved in the audio output just before the transmitter making sure that levels were even from source to source. Not today. They want one person half-assedly watching the pre-programmed computer automatedly A/V switching and rolling everything from every source. Don’t hire a human being, get another multi-level digital controller of yet another part of the spectrum to rearrange what has already been multiplexed beyond recognition as an audible decibel decern-a-ble to the human ear.
Tic-toc-tic-toc, forward into the past.
Peter Bright
len Kubas says:
June 3, 2011 at 2:14 pm
Peter, if you can find a way to manually adequately monitor dialnorm and the digital perceived volume levels in commercials that have yet to air, have at it. This has nothing to do with labor issues; humans cannot do it, and the solutions cost much less than a part-time minimum wage employee, and a fraction of the cost of having someone monitoring audio levels 24/7 (mostly while reading a book). What you describe is analog audio; it helps to understand the differences between digital audio and analog audio.
Lindsay Bold says:
June 6, 2011 at 9:56 pm
I am Chief Engineer of Orban, a company that has provided loudness controllers (under the “Optimod” and “CRL” brand names) to analog and digital television broadcasters since the early 1980s. Our current loudness controller products (Optimod 8685, 6300, and 1101) were overlooked in this article.
The 8685 is a processor for surround audio. While it measures well under the ITU BS.1770 loudness metering algorithm, the 8685 does not use BS.1770 technology for loudness control. Instead, it uses an Orban-refined, third-generation version of the CBS Loudness Controller technology that was first developed by CBS Laboratories in the mid-1960s, was later refined at CBS Technology Center in the early 1980s, and was further refined by Orban over the last few years. Thousands of processors using this technology have been sold in the last 35 years and they have processed literally millions of hours of on-air television programming. Interested readers can get more information here: http://www.orban.com/products/television/8685/
A comparison between the CBS and BS.1770 meters and a critique of the ungated version of BS.1770 can be found here:
http://www.orban.com/support/orban/techtopics/White_Paper-BS_1770_vs_CBS_meter.pdf
I have serious concerns about the current regulatory environment for television loudness control. Although the ITU released a new version of BS.1770 (BS.1770-2) in March of 2011, ATSC document A/85:2011 (“ATSC Recommended Practice: Techniques for Establishing and Maintaining Audio Loudness for Digital Television”), released literally weeks ago (25 May 2011) has not incorporated the -2 revision of BS.1770 (which specifies gating, addressing one of the two objections in my white paper referenced above). Instead, A/85:2011continues to recommend the following:
“Because the measurement of loudness per BS.1770 is an integrated measurement, quiet passages tend to lower the measured value. To avoid this, the integration may be paused during quiet passages. Automatic triggering, hold and resume, generally called “gating”, is being studied by some organizations including the ITU-R, and gating may be added to BS.1770 in the future. Some equipment may offer gating as a feature. As yet, there are no standards for gated loudness measurements. Users should utilize the current version of BS.1770 [which is, according to A/85:2011 “normative reference 3,” the obsolete BS.1770-1 (2007)] for measurements.”
It is widely anticipated that A/85:2011 will have the force of law because it will become the specified method for complying with the CALM act. If so, I believe that it is essential that the ATSC revise the document yet again to synchronize it with BS.1770-2. The Act takes effect in six months, which is already a very tight timetable for manufacturers to update loudness measurement and control products to BS.1770-2.
Apart from that, there is the issue of the accuracy of BS.1770 itself, even in gated form. In BS.1770-2, the ITU-R “further recommends that that consideration should be given to the possible need to update this Recommendation in the event that new loudness algorithms are shown to provide performance that is significantly improved over the algorithm specified in Annex 1.” There is good reason to be concerned about the accuracy of BS.1770. In figure 14 of the BS.1770-2 document, it can be seen that the subjective loudness (i.e., the loudness as determined by a panel of human listeners) of some pieces of test material disagreed with the meter by more than 6 dB in either direction. This is a lot. Particularly on the high side, it is significantly outside of A/85’s own specified +2/-5 dB “comfort zone” (A/85:2011 fig. E-1).
Then there is the problem of program material with substantial low bass energy, for which BS.1770 fails according to “Investigations on the Inclusion of the LFE Channel in the ITU-R BS.1770-1 Loudness Algorithm,” by Norcross and Lavoie (AES Convention Paper 7829, 127th AES Convention, New York 2009). Both A/85 and its European cousin (EBU R 128) recognize this issue and punt on it, suggesting that because BS.1770 cannot assess the LFE channel’s contribution to program loudness, perhaps the LFE channel should not be used at all in broadcast programming! In addition, the authors of both A/85 and EBU R 128 fail to fully understand the conclusions of the Norcross and Lavoie paper, which indicates that the problem is more than just the Dolby-specified 10 dB of boost that the LFE channel receives upon playback on loudspeakers. In fact, the problem may be with the K-frequency-weighting curve used in BS.1770.
In two AES papers, authors from the Australian Broadcasting Corporation and the University of Sydney address the accuracy of BS.1770’s K-weighting curve by using octave-band noise in subjective listening tests. The authors state:
“Comparison of the test results with an image of the filter curve currently specified in ITU-R Recommendation BS.1770 (Figure 13) shows good agreement at 250 Hz and above 500 Hz, reasonable agreement at 500 Hz, but marked difference in the bottom two octaves. The relatively good performance of the BS.1770 algorithm in ITU trials suggests that, in partial loudness terms, there was probably not much test content in the 125 Hz band or below. While the existing BS.1770 filter curve is probably a good choice in applications where the program is dominated by speech, and it is certainly an improvement on the A and B curves in that application, it is likely to give significant errors in measuring the loudness of other programs with more partial loudness in the lower frequencies, such as movie soundtracks and popular music. It is therefore desirable to improve on this filter for more general measurement of program loudness.” [Cabrera, Dash & Miranda, “Multichannel Loudness Listening Test,” AES Convention Paper 7451, 124th AES Convention, Amsterdam 2008.]
The issues described above indicate that there are significant potholes on the road to loudness nirvana. Moreover, the U.S. recommendation (ATSC R/85:2011) disagrees with both the European recommendation (R 128) and the international recommendation (BS.1770-2) regarding use of gating in the BS.1770 measurement, while the European and international recommendations disagree with each other regarding the details of how the gating should be implemented. And none of the recommendations can accurately handle loudness contributed by the LFE channel or any by low, strong bass for that matter.
Both A/85:2011 and R 128 recognize this in an indirect way by stating that the ultimate loudness arbiter is a human listener. A/85 states, “Because loudness is a subjective phenomenon, human hearing is the best judge of loudness. When combined with a known mixing environment, such as that described in Section 10 of this RP, experienced audio mixers using their sense of hearing can produce well-balanced program sound (content) with remarkably consistent loudness.” Meanwhile, EUR Tech 3343 (“Practical guidelines for Production & Implementation in accordance with R 128”) states, “Use and trust your ears! They are the best loudness meters. Smile when you watch your colleagues working with the loudness paradigm as if nothing else ever existed.”
This is excellent advice from both sides of the Atlantic, recognizing that BS.1770 is a somewhat rough measure. However, if A/85:2011 takes on the force of law, this advice is likely to be ignored in favor of “meter reading,” where “ears” are ignored in favor of a meter whose carefully logged readings can be presented to the “loudness cops” as evidence that a broadcaster was in constant compliance with the CALM Act.
The night before I wrote this, I heard an excellent example of the pitfalls that await those who substitute BS.1770 for human judgment. I was watching network television where the program material was clearly being disciplined by BS.1770. Despite this, I often felt that a human operator would have made better choices than the automated loudness control method in use. A particularly egregious example was a commercial for an action movie containing lots explosions and other high-octane effects. Substantial dynamic compression had clearly been applied to the commercial’s soundtrack, which brought up its integrated RMS level. The result of processing to the BS.1770 meter? The perceived loudness of the explosions and other dramatic soundtrack elements was substantially below the loudness of the unadorned dialog contained in the preceding program, which was a medical drama. Not only did this juxtaposition sound ridiculous, but if I were the person paying for that spot to run in network prime time, I would have been furious that its impact has been ruined by subjective loudness in a big action scene that was probably 4-5 dB below the quietly spoken dialog in the preceding program material.
I will conclude with a commercial of my own, wrapped in some philosophy. The third-generation CBS Loudness Controller, as implemented in my company’s products, does not make mistakes like this because (among other things) it pays much more attention to short-term loudness peaks (i.e., peaks whose duration is a few hundred milliseconds) than does BS.1770. It also takes the LFE channel into account. In addition to controlling subjective loudness, the 8685 can also automatically re-equalize dialog and adjust its balance with respect to other program elements, which is often needed to improve the intelligibility of less-than-ideal source programming. It does so without causing the dialog to audibly modulate the loudness of the other program elements, which can be a problem with controllers that take an active approach to conditioning the audio.
In the example above, the CBS controller would have constrained the subjective loudness of the action-movie commercial to approximately the same as the preceding dialog. While some might argue that even this is not ideal, this philosophy produces audio that is always comfortably listenable and intelligible without short-term blasts of loudness that disturb family members and neighbors who are not watching the same program. This technology means that viewers do not have to watch TV with the remote control in their hand to control correct volume levels. Because the CBS controller is aware of and controls short-term loudness peaks, compared to BS.1770 it is much harder to “game” the CBS controller by inserting short, high-loudness peaks in a commercial that also has substantial periods of low-RMS material such that the integrated power over the entire commercial is relatively low. The history of television broadcasting has shown that satisfying the mass audience requires audio with well-controlled short-term loudness because otherwise short-term high-loudness events, like gunshots, explosions, and screeching tires, will simply become noise pollution to those who are not watching the program.
Conclusions? While BS.1770 is certainly an improvement over the totally uncontrolled loudness variations typical of DTV before the advent of A/85 and R 128, it is unwise to blindly log the reading of a BS.1770 meter and assume that compliance ensures that there are no loudness-related irritants in the audio. As we have seen, the meter may disagree with a human listener by up to 6 dB, cannot be trusted with material having heavy bass below 100 Hz, and can easily miss short-term, highly annoying loudness peaks.
Bob Orban
Chief Engineer
Orban
Lindsay Bold says:
June 6, 2011 at 9:59 pm
Since the posting software seems to have eaten the paragraph breaks in my comment above^^^^, let me try this again:
I am Chief Engineer of Orban, a company that has provided loudness controllers (under the “Optimod” and “CRL” brand names) to analog and digital television broadcasters since the early 1980s. Our current loudness controller products (Optimod 8685, 6300, and 1101) were overlooked in this article.
The 8685 is a processor for surround audio. While it measures well under the ITU BS.1770 loudness metering algorithm, the 8685 does not use BS.1770 technology for loudness control. Instead, it uses an Orban-refined, third-generation version of the CBS Loudness Controller technology that was first developed by CBS Laboratories in the mid-1960s, was later refined at CBS Technology Center in the early 1980s, and was further refined by Orban over the last few years. Thousands of processors using this technology have been sold in the last 35 years and they have processed literally millions of hours of on-air television programming. Interested readers can get more information here: http://www.orban.com/products/television/8685/
A comparison between the CBS and BS.1770 meters and a critique of the ungated version of BS.1770 can be found here:
http://www.orban.com/support/orban/techtopics/White_Paper-BS_1770_vs_CBS_meter.pdf
I have serious concerns about the current regulatory environment for television loudness control. Although the ITU released a new version of BS.1770 (BS.1770-2) in March of 2011, ATSC document A/85:2011 (“ATSC Recommended Practice: Techniques for Establishing and Maintaining Audio Loudness for Digital Television”), released literally weeks ago (25 May 2011) has not incorporated the -2 revision of BS.1770 (which specifies gating, addressing one of the two objections in my white paper referenced above). Instead, A/85:2011continues to recommend the following:
“Because the measurement of loudness per BS.1770 is an integrated measurement, quiet passages tend to lower the measured value. To avoid this, the integration may be paused during quiet passages. Automatic triggering, hold and resume, generally called “gating”, is being studied by some organizations including the ITU-R, and gating may be added to BS.1770 in the future. Some equipment may offer gating as a feature. As yet, there are no standards for gated loudness measurements. Users should utilize the current version of BS.1770 [which is, according to A/85:2011 “normative reference 3,” the obsolete BS.1770-1 (2007)] for measurements.”
It is widely anticipated that A/85:2011 will have the force of law because it will become the specified method for complying with the CALM act. If so, I believe that it is essential that the ATSC revise the document yet again to synchronize it with BS.1770-2. The Act takes effect in six months, which is already a very tight timetable for manufacturers to update loudness measurement and control products to BS.1770-2.
Apart from that, there is the issue of the accuracy of BS.1770 itself, even in gated form. In BS.1770-2, the ITU-R “further recommends that that consideration should be given to the possible need to update this Recommendation in the event that new loudness algorithms are shown to provide performance that is significantly improved over the algorithm specified in Annex 1.” There is good reason to be concerned about the accuracy of BS.1770. In figure 14 of the BS.1770-2 document, it can be seen that the subjective loudness (i.e., the loudness as determined by a panel of human listeners) of some pieces of test material disagreed with the meter by more than 6 dB in either direction. This is a lot. Particularly on the high side, it is significantly outside of A/85’s own specified +2/-5 dB “comfort zone” (A/85:2011 fig. E-1).
Then there is the problem of program material with substantial low bass energy, for which BS.1770 fails according to “Investigations on the Inclusion of the LFE Channel in the ITU-R BS.1770-1 Loudness Algorithm,” by Norcross and Lavoie (AES Convention Paper 7829, 127th AES Convention, New York 2009). Both A/85 and its European cousin (EBU R 128) recognize this issue and punt on it, suggesting that because BS.1770 cannot assess the LFE channel’s contribution to program loudness, perhaps the LFE channel should not be used at all in broadcast programming! In addition, the authors of both A/85 and EBU R 128 fail to fully understand the conclusions of the Norcross and Lavoie paper, which indicates that the problem is more than just the Dolby-specified 10 dB of boost that the LFE channel receives upon playback on loudspeakers. In fact, the problem may be with the K-frequency-weighting curve used in BS.1770.
In two AES papers, authors from the Australian Broadcasting Corporation and the University of Sydney address the accuracy of BS.1770’s K-weighting curve by using octave-band noise in subjective listening tests. The authors state:
“Comparison of the test results with an image of the filter curve currently specified in ITU-R Recommendation BS.1770 (Figure 13) shows good agreement at 250 Hz and above 500 Hz, reasonable agreement at 500 Hz, but marked difference in the bottom two octaves. The relatively good performance of the BS.1770 algorithm in ITU trials suggests that, in partial loudness terms, there was probably not much test content in the 125 Hz band or below. While the existing BS.1770 filter curve is probably a good choice in applications where the program is dominated by speech, and it is certainly an improvement on the A and B curves in that application, it is likely to give significant errors in measuring the loudness of other programs with more partial loudness in the lower frequencies, such as movie soundtracks and popular music. It is therefore desirable to improve on this filter for more general measurement of program loudness.” [Cabrera, Dash & Miranda, “Multichannel Loudness Listening Test,” AES Convention Paper 7451, 124th AES Convention, Amsterdam 2008.]
The issues described above indicate that there are significant potholes on the road to loudness nirvana. Moreover, the U.S. recommendation (ATSC R/85:2011) disagrees with both the European recommendation (R 128) and the international recommendation (BS.1770-2) regarding use of gating in the BS.1770 measurement, while the European and international recommendations disagree with each other regarding the details of how the gating should be implemented. And none of the recommendations can accurately handle loudness contributed by the LFE channel or any by low, strong bass for that matter.
Both A/85:2011 and R 128 recognize this in an indirect way by stating that the ultimate loudness arbiter is a human listener. A/85 states, “Because loudness is a subjective phenomenon, human hearing is the best judge of loudness. When combined with a known mixing environment, such as that described in Section 10 of this RP, experienced audio mixers using their sense of hearing can produce well-balanced program sound (content) with remarkably consistent loudness.” Meanwhile, EUR Tech 3343 (“Practical guidelines for Production & Implementation in accordance with R 128”) states, “Use and trust your ears! They are the best loudness meters. Smile when you watch your colleagues working with the loudness paradigm as if nothing else ever existed.”
This is excellent advice from both sides of the Atlantic, recognizing that BS.1770 is a somewhat rough measure. However, if A/85:2011 takes on the force of law, this advice is likely to be ignored in favor of “meter reading,” where “ears” are ignored in favor of a meter whose carefully logged readings can be presented to the “loudness cops” as evidence that a broadcaster was in constant compliance with the CALM Act.
The night before I wrote this, I heard an excellent example of the pitfalls that await those who substitute BS.1770 for human judgment. I was watching network television where the program material was clearly being disciplined by BS.1770. Despite this, I often felt that a human operator would have made better choices than the automated loudness control method in use. A particularly egregious example was a commercial for an action movie containing lots explosions and other high-octane effects. Substantial dynamic compression had clearly been applied to the commercial’s soundtrack, which brought up its integrated RMS level. The result of processing to the BS.1770 meter? The perceived loudness of the explosions and other dramatic soundtrack elements was substantially below the loudness of the unadorned dialog contained in the preceding program, which was a medical drama. Not only did this juxtaposition sound ridiculous, but if I were the person paying for that spot to run in network prime time, I would have been furious that its impact has been ruined by subjective loudness in a big action scene that was probably 4-5 dB below the quietly spoken dialog in the preceding program material.
I will conclude with a commercial of my own, wrapped in some philosophy. The third-generation CBS Loudness Controller, as implemented in my company’s products, does not make mistakes like this because (among other things) it pays much more attention to short-term loudness peaks (i.e., peaks whose duration is a few hundred milliseconds) than does BS.1770. It also takes the LFE channel into account. In addition to controlling subjective loudness, the 8685 can also automatically re-equalize dialog and adjust its balance with respect to other program elements, which is often needed to improve the intelligibility of less-than-ideal source programming. It does so without causing the dialog to audibly modulate the loudness of the other program elements, which can be a problem with controllers that take an active approach to conditioning the audio.
In the example above, the CBS controller would have constrained the subjective loudness of the action-movie commercial to approximately the same as the preceding dialog. While some might argue that even this is not ideal, this philosophy produces audio that is always comfortably listenable and intelligible without short-term blasts of loudness that disturb family members and neighbors who are not watching the same program. This technology means that viewers do not have to watch TV with the remote control in their hand to control correct volume levels. Because the CBS controller is aware of and controls short-term loudness peaks, compared to BS.1770 it is much harder to “game” the CBS controller by inserting short, high-loudness peaks in a commercial that also has substantial periods of low-RMS material such that the integrated power over the entire commercial is relatively low. The history of television broadcasting has shown that satisfying the mass audience requires audio with well-controlled short-term loudness because otherwise short-term high-loudness events, like gunshots, explosions, and screeching tires, will simply become noise pollution to those who are not watching the program.
Conclusions? While BS.1770 is certainly an improvement over the totally uncontrolled loudness variations typical of DTV before the advent of A/85 and R 128, it is unwise to blindly log the reading of a BS.1770 meter and assume that compliance ensures that there are no loudness-related irritants in the audio. As we have seen, the meter may disagree with a human listener by up to 6 dB, cannot be trusted with material having heavy bass below 100 Hz, and can easily miss short-term, highly annoying loudness peaks.
Bob Orban
Chief Engineer
Orban
Lindsay Bold says:
June 6, 2011 at 10:02 pm
OK…I give up regarding paragraph breaks 🙁 Trust me — the original text had them.