Orban
8585

The 8585 and Dolby Digital® (AC3) Metadata

There are three important pieces of metadata in the AC3 bitstream.

When used correctly, these can help address the problem of inconsistent loudness between different sources while allowing viewers to individually choose the amount of dynamic compression they hear. However, experience so far has shown that the metadata implementation in the broadcast chain has often been too haphazard to prevent audience irritation.

Orban believes that the most realistic approach to handling AC3 dialog normalization is a hybrid approach. It is important to consider carefully what program material will truly benefit from the ability to be heard with unprocessed dynamic range. Prime-time dramatic shows, newer feature films, and classical music concerts all use dynamic range for dramatic impact and are therefore candidates for full exploitation of the AC3 DRC metadata. Material that airs with full Dynamic Range Control implemented should be refined in production so that it sounds polished and consistent without further processing. Each show, film, and concert must have a dialog normalization value pre-assigned to it, derived from a long-term loudness meter or by human audition. It is probably impractical to pass through, without review, dialog normalization values created by program and commercial providers because some commercial providers will inevitably try to game the system to make their commercials excessively loud. Instead, if dialog normalization is to be actively used in transmission, the broadcaster must strip its existing value from the program and then must preview each piece of program material, replacing the value with one that will ensure consistency from one piece of program material to the next.

Even program segments whose Dialnorm value is set automatically according a long-term loudness measurement like ITU BS.1770 may still have short-term loudness peaks that are extremely annoying. Any program material that will not benefit from being heard with full dynamic range should be processed with the 8585 so that viewers can hear the audio comfortably. They should not be blasted by loud effects or commercials or being forced to strain to understand dialog. Most program material, including commercials, live news, sports, most documentaries, game shows, talk shows, soap operas, and pop music videos and concerts, can receive 8585 processing. The 8585 controls subjective loudness very well, so a single dialog normalization value can be applied to all program material whenever the 8585 is online. The advantage of this strategy is that the 8585 will guarantee that all of this material is comfortably listenable and that commercials are not excessively loud. With the possible exception of sport and some concerts, this program material does not rely on extreme dynamic range to make its point, so it is unlikely that compression will damage the artistic integrity of this programming.

No one needs more dynamic range on talk shows or on the local news! The 8585 can smoothly activate and defeat its dynamics processing on-air via GPI triggers or other remote control, so it is easy to implement this strategy.

Another important 8585 feature is "automatic re-equalization" of program material. The 8585's multiband compressor can automatically re-equalize program material towards a preset target spectral balance by applying more gain reduction to frequency bands containing more power. The 8585's compressor's band coupling controls determine the maximum amount of re-equalization permitted.

In the two-band compressor, re-equalization tames excessive bass, which can otherwise cause muddy balances. The five-band compressor can perform more detailed automatic re-equalization, which can be particularly useful for program material such as live news and for any material where dialog intelligibility is a problem. Approximately 60% of digital OPTIMOD-TV users use two-band compression; the rest use five-band compression.

Dynamic range compression in Dolby Digital (using DRC metadata to achieve compression at the receiver) is a simple dynamic gain adjustment performed over the entire audio bandwidth; it does not do automatic re-equalization. The level detector determining the amount of DRC compression can be frequency-contoured to mimic the equal-loudness curves of the ear and has the ability to "look ahead" at upcoming program level changes. This is sufficient for many applications, but may be improved with the addition of a multiband device like the 8585 to handle certain programming that may not get sufficient treatment from a single-band device like that in DRC.

The 8585 has RS485 serial connections that can accept and emit AC3 metadata. In v1 software, all metadata from the input is passed to the output except for dialnorm and DRC. The 8585 applies its active, user-specified dialnorm value to the output metadata stream, which also instructs the downstream Dolby Digital encoder to re-author the line-mode and RF-mode DRC metadata to complement the active dialnorm value using the compression profile in the input metadata.

About Surround Synthesis

After substantial research, we chose not to include surround synthesis (more formally call "blind upmixing") in the 8585. Here's why:

  1. Most TV sets are either mono or stereo. As specified in the Dolby Digital® standard, when the transmission is flagged as 5.1 channel these sets reproduce a downmix of the five main surround channels. (The .1 channel is discarded.) If you transmit stereo, stereo TVs will accurately reproduce the original mix. This is not true if you synthesize an arbitrary surround mix from the original stereo and expect the TV to extract stereo from it through a downmix. Unlike the old Orban stereo synthesizers (which were completely mono-compatible), 5.1 upmixes are not stereo-compatible and can cause comb filtering and other coloration.
  2. Synthesized upmixes are particularly hard for the Dolby AC3 codec to pass to the receiver without adding without adding audible artifacts. The codec is much more transparent to stereo and to normally mixed 5.1 material.
  3. If you want viewers with surround setups to hear an upmix, just set the Dolby ProLogic flag in the Dolby Digital bitstream you transmit to viewers and let the Dolby ProLogic or ProLogic II decoder in every surround receiver create the surround.
  4. Even this may be a bad idea. In AES Convention Paper 6915 ("Perceptual Evaluation of Algorithms for Blind Up-mix"), Spore, Walther, Liebetrau, Bube, Fabris, Hohberger and Kohler described double-blind listening tests comparing stereo music with blind upmixes of it using a variety of upmix algorithms. All major algorithms from USA, Germany, Switzerland and Denmark were included in the test.

For listeners in the center stereo "sweet spot," none of the upmixing schemes was rated significantly better than the original 2-channel stereo input and most of the schemes were rated worse. Upmixes were only preferred if listeners were located away from the sweet spot. Because the program material was music, the results may not apply to typical television programming. Nevertheless, we have heard some extremely weird and unpleasant-sounding upmixes on the air. Certainly, dialog emerging from what sounds like 16-foot wide heads wrapped halfway around the living room should concern any broadcaster who does not want to give viewers an excuse to hit the "mute" button or fast-forward the program material.

return to top