ractical broadcast streams interrupt programs with commercial advertisements, which themselves must be independently measured. Real-time monitoring and logging requires independent assessment of program material and of the commercials that separate or interrupt it
This year broadcasters must begin measuring and controlling the loudness of commercials, both in absolute terms and relative to the program. Richard Cabot, PhD is CTO of Qualis Audio contends that while managing loudness is important, it’s only part of what's required to deliver quality audio to today’s viewers.
The shift to surround sound from stereo brings additional complexity to the tasks of audio monitoring and quality assurance. Unfortunately, this comes at a time of tightening budgets and shrinking pools of skilled personnel. These changes are forcing many people unfamiliar with the subtleties of audio into roles where they are responsible for ensuring its quality.
Traditional audio measurement equipment is often based on what was technically simple to measure (eg. Level, THD, inter-channel phase, etc.), rather than what correlated best with human perception.
The Qualis Audio Sentinel was designed to be an electronic listener, modeling important perceptual parameters rather than basic technical parameters. It measures loudness, downmix compatibility, intelligibility, levels, balance, hum, metadata and other relevant parameters. Although it can display results graphically in a standard browser window, it doesn’t need to, as all measurements create numeric results that are tested and generate errors if outside user selected boundaries. It can act on its QoE judgments and issue alarms based on one or more of these errors via contact closures, email, SNMP traps, audible or visual indicators. This allows the Sentinel to function equally well as an unattended monitor, reducing personnel requirements, or as an assist to a skilled operator, allowing attention to be put on other tasks besides policing common audio issues.
Although DTV brought surround sound to the picture, most viewers still listen in stereo or mono, making compatibility of the surround signal with these formats crucial. Previous commercial attempts to detect downmix compatibility problems, however, required subjective interpretation of a complex visual display. Essentially a multidimensional version of the scope display long used to assess stereo-to-mono compatibility, they require a skilled operator’s attention. Such displays cannot be distilled down to a pass/fail result, making true unattended monitoring impossible.
Qualis Audio’s patent-pending technology compares what a human would hear when listening to the surround program with what would be heard listening to a downmixed version. As shown in the figure, it downmixes the surround to stereo and mono versions. The spectrum of the original surround channels, along with the stereo and mono downmixes, is measured to 1/30th octave resolution. The surround channel spectra are combined and compared to the spectra of the downmixes. Differences represent content which has been lost and are expressed in dB as a function of frequency. This provides an intuitive understanding of the problem and allows simple alarm thresholds to be defined. The user simply selects how much degradation is acceptable when the surround mix is converted to stereo. The detection algorithms provide duration thresholds as well, ensuring that transient problems are ignored and only sustained problems generate errors.
Loudness and other audio parameters are defined and measured from beginning to end of a programming segment. Practical broadcast streams interrupt programs with commercial advertisements, which themselves must be independently measured. Real-time monitoring and logging requires independent assessment of program material and of the commercials that separate or interrupt it. This normally requires multiple measurement meters or convoluted and error-prone manipulation of the data provided by a single meter. The Sentinel solves this problem by maintaining two independent meters for each of the two audio streams it measures.
A little known fact about the internationally standardized method of measuring loudness is that for some material, the loudness results vary with the reproduction format. Although a piece of content might match the -24 (+/-2) LU target specified by the FCC when measured in surround, the loudness can fail these limits when reproduced in stereo. The instrument addresses this by measuring all surround content in both its original format and after downmixing to stereo. If a particular program or commercial differs significantly after downmixing a warning is issued so corrective action can be taken.
Another important point to take into consideration is the accuracy of the audio meter being used. There are many loudness meters on the market, but broadcasters should not be complacent and assume that their meter is compliant with EBU R128, ITU 1770 and ATSC A/85; some are not! Thanks to a unique collaboration between Qualis Audio and Prism Sound, a new loudness meter verification suite has been developed that allows broadcasters to determine whether every aspect of their meter design complies with the latest international loudness standards. This programmable, easy-to-use test suite runs on Prism Sound’s dScope Series III audio analysis platform, and is free to download from www.prismsound.com/loudness1770, along with a set of fixed wave files which replicate the dScope Series III test suite.
It is important when mixing surround programs to keep the dialog louder than other content so it remains intelligible. As a guide to accomplishing this, professional sound mixers are often instructed to maintain a minimum level difference between the center front and the left and right front channels. The levels of these three front channels are displayed on meters of virtually all mixing consoles. They are usually in close proximity to each other, typically arranged in the order LF, RF, CF. This makes comparisons relatively easy and so the visual comparison technique has become a common practice.
The left and right front levels are often, but not always, representative of the overall program level. A common exception is when mixing live sports and crowd noise occurs in the surrounds. In this situation a better guideline would be to compare the center front level to each of the other surround channels in the program. However, this would be far more difficult because of the larger number of meters involved, their larger physical separation and the presence of the LFE channel meter, usually next to the center front meter, which would not be involved in the comparison.
Even when performing the simpler task of comparing center front level to the left and right front levels continuous attention is required. If the user is not looking at the meters intelligibility may inadvertently drop to an unacceptable level.
Although level can be a useful proxy for loudness, it is frequently inaccurate. The user is attempting to judge the intelligibility of the dialog in the program being monitored. Any inaccuracy of the loudness assessment results in inaccuracy of the intelligibility estimate. A fundamental error stems from the reliance on wideband, or non-frequency selective, assessments. These wideband measures do not consider many aspects of the human auditory system.
The ratio of dialog to remaining program is computed using 400ms momentary loudness computations. The loudness of the remaining audio (in LKFS) is subtracted from the loudness of the center front channel (in LKFS) and the resulting values are processed with a running 3 second average.
Since there are frequent periods of no dialog it is important to compute the indication only when dialog is present. Otherwise the ratio of dialog to remaining content will be reduced by the fraction of time that dialog is not present. A threshold is set which must be exceeded by the dialog or the computation is not performed. If the computation has not been recently performed the display color is changed to grey to indicate that the reading is “stale”.
Signals from a playout server (via GPI lines or Ethernet) let the Sentinel know when the stream is a program versus a commercial, and when a new program or commercial begins. The instrument handles everything else, creating a time-stamped list of programs and commercials with their loudness and other measured audio-quality parameters.
Sometimes it is impractical to issue this information in real time from the playout server. In such cases, Qualis Audio offers a solution which merges the measurement log data with as-run logs, automatically creating a clear and itemized report of measured loudness (and other measures if desired) for each item broadcast. Content that falls outside legal or user-defined limits is flagged in the report. The results can be automatically emailed to a specified distribution list, eliminating the need for labor-intensive manual tracking systems.
Some customers employ the Sentinel as a “forensic” tool, making measurements on content after a question arises about its suitability. Its extensive analysis ability allows rapid assessment of problems, and the storage of all measurement data eliminates the need to re-measure content. Unlike other instruments that log measurements, the user does not have to review confusing text files of measurement data. Log data is viewed through the same intuitive graphical UI that displays real-time measurements.
With CALM Act requirements becoming effective at the end of the year, much attention has been given to loudness. Although fully compliant with all international loudness standards, the Sentinel goes much further. Besides characterizing loudness more thoroughly than anything else available, it assesses a wide range of other parameters necessary to deliver quality audio to viewers. By applying psycho-acoustically-based measurement algorithms and intelligent alarm generation, it delivers improved audio quality while decreasing costs.