Audio Processing   

Audio processing has been used by broadcasters since the 1970s. It was originally used to improve the signal for AM signals but was carried over into the FM system. Research that was carried out when AM was the primary radio system for commercial radio showed that listeners preferred a louder signal than a quieter one. Ever since then, broadcasting companies have competed to be “the loudest on the dial”. One of the ways that this audio processing is implemented is that the signal is sent through a group of filters to separate the signal into a group of signals that are limited to different frequency bands. These signals then have their dynamic range compressed so that the overall sound is louder. One of the ways in which this dynamic range compression is carried out is that the tops of the waveforms are just chopped off and the resulting signal is amplified. The result is that the average value of the amplitude, and therefore the loudness, increases thus making an overall louder signal. An example of a station that uses high levels of dynamic range compression is Radio 1. If you listen to Radio 1 and then tune into a different station that doesn’t use dynamic range compression such as Radio 3 then although the signal should be about the same power (because the signals are from the same transmitter) the difference in loudness will be large. This dynamic range compression makes the station have a more ‘punchy’ sound. 

On FM stations, because the signal is analogue, this compression does not cause severe problems to the audio quality. All it does is make the sound more punchy and louder. Proponents of audio processing say that this sound is more entertaining than an uncompressed station while others intensely dislike this sound. On FM I can’t say I’m too bothered about it although I would prefer it if Radio 1 used lower levels of compression. 

On DAB, high levels of audio processing such as those used by Radio 1 cause problems with the encoding of the sound at low bit rates such as 128 kbps. The reason for this is that the aim of perceptual encoding is to use the psychoacoustic model’s masking curves to eliminate frequencies that cannot be perceived by the listener but because the aim of the dynamic range compression is to make the frequency components larger less of these frequency components will fall below the masking curves. Therefore, the result is that a far larger number of frequency components need to be encoded so more bits are required in order to do this. Different songs are easier for the signal to encode because the signal doesn’t contain many strong frequencies before compression so after amplification there still are few frequencies to be encoded. Other songs however contain a larger number of frequency components after dynamic range compression and at low bit rates such as 128 kbps, the encoder simply runs out of bits. The result for these songs is that the definition and clarity of the music is enormously reduced sometimes to the point where the song sounds worse than if it had been played back from an old cassette tape. 

Unfortunately, the majority of songs are affected significantly by this action and a few sound absolutely dreadful; very few songs actually sound good. Furthermore, the majority of the radio stations playing ‘pop’ music are carrying on with this practice of using high levels of audio processing and the audio quality of all of these stations that do so is poor.   

The problem is compounded when the music is played back from vinyl. The reason for there being a problem with vinyl is that on the vinyl, beneath the music there is a great deal of noise that normally cannot be perceived by the listener. This noise confuses the encoder and it goes ahead and codes this noise. Therefore, the effective bit rate is reduced. If this signal is then compressed using an audio processor the two problems add together to dramatically reduce the effective bit rate so for already low bit rate transmission the majority of tracks sound extremely poor to the point of being sub-cassette standard. Radio stations will usually use re-mastered copies off vinyl stored on CD. Dance music shows though are seriously affected because the DJ has to mix live off vinyl so the only thing that can be done is to increase the bit rate or lower the audio processing level. For example, on Radio 1 DAB, one of the worst sounding shows are Pete Tong's Essential Selection and Essential Mix. On these 2 shows there is a combination of high processing levels, low bit rate and vinyl.

The audio quality for the stations that use high levels of processing on their DAB signals is far below that of FM. Therefore, if you primarily listen to stations that are aimed at a younger audience then it is almost certain that that station will sound better on FM than it does on DAB so long as you have a good FM signal. So for people who this applies to, DAB for the home is not worth the investment. Thankfully, Freeview and digital satellite (DSat) carry higher bit rate versions of the stations on DAB so the audio quality is much better.

Dynamic Range Control (DRC)

Dyanamic range control was specifically designed for the DAB system by the BBC but unbelievably they are not using it on all of their stations. They use it on Radio 3, 4 and 5 and that is why the audio quality of these stations is very good. DRC is not used on Radios 1, 2 or 6 Music so these stations have audio processing applied to them. BBC 6 Music does not have as high levels of audio processing applied to it and therefore has a higher audio quality.

Fortunately there is always the option of not going for DAB at all when upgrading to a digital radio system. The alternative system is digital radio received via satellite which is described on the Digital Satellite Radio page.

For more information about audio processing, quality and bit rate see this article.