Achieving consistent loudness measurements Part 2
Mar 25, 2011 12:35 PM, By Jack Kontney
The key elements of accuracy in loudness measurement, as detailed by Karsten Hansen, CEO of DK-Technologies in Denmark.
As the force behind metering and monitoring specialist DK-Technologies, CEO Karsten Hansen has made a career of finding insightful, accurate solutions to the monitoring needs of broadcasters. With the passage of the CALM Act in the United States and ongoing regulatory changes in Europe, Hansen has developed a straightforward set of guidelines to ensure that the loudness measurements his company takes are be both accurate and compliant.
This is important due to the fact that different equipment, nominally sold for the same purpose and developed to meet the same standards of loudness measurement, has been shown to produce significant variance in the results obtained. Like all technical facets of broadcasting, it is critical to know and implement best practices in loudness measurement and control.
Hansen’s first recommendation is to use loudness measurement equipment from a trusted manufacturer. This trust can be best established through some straightforward confirmation testing. It is a fact that there are significant variations among manufacturers, and so broadcasters are encouraged to rely on benchmark testing to confirm the accuracy of any loudness measurement device before purchasing.
The first step involves the digital interface: It is critical to prohibit sample rate converters within the system. Hansen showed this conclusively by measuring a known pair of test signals: one a 400Hz sine wave saturated by a brick wall, cutting off the top 3dB, and the other signal was a square wave with the same peak level. The 3dB difference between those tones was confirmed by taking the digital output back to an input without a sample rate converter.
However, when running the same test with a sample rate converter inserted in the signal path, results were markedly different. On the saturated sine wave, the result was virtually identical. But for the square wave, the measured result was significantly louder, by roughly 3.5 dB in this example. The only variable between tests was the insertion of a sample rate converter.
This phenomenon is caused by frequency folding, which occurs when frequencies above half the sampling rate are introduced. Thus, if the sampling rate is at 48kHz, it is mandatory then that there is no frequency above 24kHz in the audio stream. While that sounds simple enough, today’s digital systems make it easy to unintentionally violate this critical guideline.
Within a typical system, such as a mixing console, a host of changes are typically employed in the digital domain, such as compression, filtering and other modulation components. When added together, these digital manipulations can add high-frequency components to an otherwise legal audio signal, introducing artifacts that cause the signal to fold back on itself within the sampling window.
Thus, when a sample rate converter is introduced to the system, it introduces the possibility of “elastic” measurements and creates variable results, which is the antithesis of what any measurement device should do. Frequency folding can only be prevented by using very high sample rates, which is not a viable option in the 48kHz world of broadcasting. So, it is important to perform the two-tone test to ensure consistent results across all signal types.
The next test is for transient response. True-peak measurements must comply with benchmark testing using one-sample and two-sample -6dBFS tests. Hansen chose this test signal level as one that is typical for broadcast use. The test’s requirement is that the system measure true peak “on the mark” when reading one audio sample. When adding a second adjacent sample, the reading should be approximately 2dB more. This simple test will tell you whether the digital true peak performance of the measurement device is accurate.
Hansen notes that the reading with two samples is only approximate.
“Mathematically, it’s actually a bit more,” he said. “But for a broadcaster, it’s easy to remember: two samples, 2db more. That is the key.”
For static and dynamic performance, use a tone generator to perform channel sum and tone burst tests to confirm readings in sliding and integrating modes of the loudness meter.
To test the static performance of the system, use a steady tone from a reliable source to establish a reference-level loudness reading of -3dB for a single channel, and then check the summing values as more channels are added at the same level. For two channels at reference level, the reading should be 0dB. For 5+1 channels, the increase should measure approximately 5dB (technically, 4.7). Thus, a 5.1 surround system is 8dB louder than a mono system in static performance. If the system complies with these measurements, its static performance is correct.
To test dynamic performance, Hansen uses a tone-burst signal using a 300ms tone and a 3000ms pause. This 1:10 ratio produces a 10dB reduction in the loudness reading. The goal of this test is to analyze the overload capability, or crest factor, of the measurement equipment, so the generator level is raised 10dB. Readings are then taken in two modes: sliding and integrating. The integrating value is used for short duration measurements, while the sliding value calculates loudness with a sample time of up to 20s. The goal of the tone-burst measurement is to confirm equal values for both the sliding and integrating measurements. This confirms accuracy of the loudness measurement.
By performing these tests, broadcasters can assess the accuracy of a loudness measurement device under real-world conditions. With the increasing emphasis on loudness measurement and control, it is critical that broadcasters know conclusively that the measurements they are taking are both accurate and repeatable.
“I developed these tests strictly from an engineering perspective,” Karsten Hansen said. “The goal was to learn why broadcasters would get different loudness measurements. If your setup is incorrect, you can get all sorts of readings, like measuring length with a rubber band. But by using these few simple tests, it is a simple matter to select the right equipment and have confidence that your loudness measurements will be accurate and repeatable.”
See DK-Technologies at the 2011 NAB Show in Booth C7840.
| Want to use this article? Click here for options! |  |
