What is in this article?:
A/85 and R 128 differ significantly in philosophy and recommendations. Probably the most important difference, A/85 asserts that the loudness of a so-called “anchor element” (which is typically dialog except in programs emphasizing music, like live concert recordings) is most important, while R 128 asserts that the integrated loudness of the entire program is most important4 and, therefore, program loudness should be normalized based on an integrated BS.1770-2 measurement. The philosophy behind A/85 is similar to that of Dolby Laboratories, which, for many years, has asserted that dialog anchors most film and television programs, and that listeners set their volume controls to make dialog comfortably intelligible5. (Orban agrees more with A/85 than with R 128.)
The purpose of this paper is to present, using both meters, comparative measurements of the output of Orban’s current audio processors6 with the latest refinement of the CBS loudness controller technology.7
Test setup
A stereo recording of approximately 30 minutes of unprocessed audio from the output of the master control of a San Francisco network station was applied to the 2.0 processing chain of an Optimod-Surround 8685 processor, set for normal operation using its TV 5B general purpose preset. The digital output of the processor was applied to the digital input of an Orban 1101 soundcard, which was adjusted to pass the audio without further processing and to apply it to an Orban software-based loudness meter that simultaneously computes the BS.1770-2 Integrated loudness and CBS loudness. The first 750-second segment of the program material was a daytime drama with commercial and promotional breaks, while the remainder was local news, also with commercial and promotional breaks.
The BS.1770-2 meter was adjusted to produce a 10-second integration window in which, per the BS.1770 standard, all data are equally weighted. The CBS Loudness Gain control was set to -3.12dB. Data logged every 10 seconds and included the maximum meter indication produced by both the BS.1770 and CBS meters in each 10-second interval. (See Figure 1.) This produced 165 data points, which were imported into a scientific plotting application.8
Orban’s experimental long-term loudness measurement, based on the CBS meter and first published in 2008, was also included in the measurements and is also shown. This algorithm attempts to mimic a skilled operator’s mental integration of the peak swings of a meter with “VU-like” dynamics. The operator will concentrate most on the highest indications, but will tend to ignore a single high peak that is atypical of the others. This algorithm can be seen to share certain characteristics with the floating gate first introduced in EBU R 128 and later adopted in BS.1770-2.
The algorithm displays the average of the peak indications of the meter over a user-determined period: 10 seconds for these measurements. The average is performed before dB conversion. All peak indications within the period are weighted equally with the following exceptions:
- If the maximum peak in the window is more than 3dB higher than the second-highest peak, it is discarded.
- All peaks more than 6dB below the maximum (or second-to-maximum, if the maximum peak was discarded) are discarded.
Because the CBS long-term measurement discards a single peak if it is more than 3dB higher than the second-highest peak, the CBS long-term measurement tends to be biased about 3dB lower than a measurement that shows the maximum peak indication of the CBS meter in a 10-second period. The amount of bias depends on whether or not the loudness applied to the meter’s input is well controlled. This bias can be seen in the figures in this article. Because the Orban meter allows control of the level applied to the CBS algorithm via the “CBS Gain” control, setting it 3dB higher could better match the CBS long-term measurement to the BS.1770-2 Integrated measurement at the expense of moving the “maximum peak loudness” indication 3dB higher.
Results
To provide a baseline for discussion of the loudness-controlled results, we measured the unprocessed audio that was applied to the Optimod 8685’s input. The measurement results show the loudness of the unprocessed audio both as a function of time and as a histogram. The histogram sorts the meter outputs into 0.25dB or 0.25LK-wide slices9 and shows the number of measurements that fit into each of these slices. (See Figure 2.) Thus, the histogram portrays loudness consistency. When the histogram is clustered tightly within a few bins, the loudness is more consistent than it is when the histogram is spread out into more bins.
With all meters, the histogram of the unprocessed audio shows a wide spread. This is consistent with the EBU Loudness Range measurement for the entire clip, which was 16.5LK, while the LRA for the daytime drama alone was 19.2LK (including commercials). The BS.1770-2 Integrated loudness was -20LKFS, integrated over the entire measurement period, although the inconsistencies between the loudness of program material and commercials are large enough to make this 30-minute measurement essentially meaningless.
