Loudness – Part 1

//Loudness – Part 1

Loudness – Part 1

What is all this fuss about loudness?

In broadcast and audio for video environments, we have traditionally metered audio based on the peak (or quasi-peak

[1]) signal amplitude (or maximum sample values in the case of digital). This was great for making sure transmitters or analogue to digital converters were not over-modulated, but has a gaping flaw; it doesn’t actually tell us much about the audio content beyond that. If we use these peak level readings to normalize (or “match the volume” – which is inevitable if one sets a peak level) of different pieces of audio content, we come across some significant challenges.

Real–world audio has a varying “crest factor”; that is the measure of the largest signal amplitude against the average signal amplitude depends on the nature of the sound itself. Obviously if different program material has different crest factors but is normalized such that the peaks levels are the same, then the average level of the signal will change depending on the program.

It turns out that, as human beings, our perception of loudness has more to do with the average amplitude of an audio signal than the peak amplitude. The effect of standardizing our program output around peak audio levels means we see continual variance in loudness, which also happens to be a significant cause in complaints to broadcasters. Furthermore, if our perception of loudness is frequency dependent this needs to be addressed if we are to adequately measure it.

In order to try to control these factors, the European Broadcasting Union (EBU) has issued a recommended practice document (R-128) for metering and normalizing audio for broadcast based on an International Telecommunication Union (ITU) measurement standard. ITU BS 1770 specifies a mechanism for measuring the “perceptual loudness” of audio signals, which has been independently verified to correlate well with subjective observation. It achieves this by averaging the signal over time and also provides specifications for filters (k-weighting) which take account of the frequency selectivity of human hearing. It will also average over multiple channels ( i.e. you just get one set of readings whether you send it stereo or 5.1 audio) and accounts for the disturbing effect of loud sound coming from the rear with surround sources.

All major broadcasters in the UK and most in the EU are moving towards requiring compliance to R-128 and appear to be targeting a switch over in October, coincident with the move towards file based delivery via the Digital Production Partnership (DPP) AS-11 Delivery Standard. The Italian government has already legislated around this issue, as has the US (although to a slightly different standard, ATSC A-85).

[1] For IEC 60268-10 type 1 meters 5ms integration time quasi peak, for type 2 meters (traditional UK broadcast PPMs) 10 ms integration time quasi peak.

By |2016-10-29T08:55:11+00:00April 15th, 2014|Categories: Broadcast Engineering|Tags: , , , , |1 Comment

One Comment

  1. root6 blog | April 15, 2014 at 3:50 pm - Reply

    […] ← Loudness – Part 1 […]

Leave A Comment