How DAB works
| DAB digital radio uses a transmission system known as
coded orthogonal frequency division multiplex (COFDM). This
divides the signals amongst several hundred low bit rate
carriers. |
|
Because the bit rates are low, the signal
can be subject to large multipath dispersion without any
adverse effect on reception. In fact, multipath reinforces
and improves reception.
For DAB, the bit rates on the individual carriers are
low enough for different transmitters carrying the same
programmes to broadcast on the same frequency, provided
they are correctly synchronised. This enables much more
efficient use of the radio spectrum. |
Each DAB transmission, known as a multiplex, is spread
over about 1.5 MHz, with 4 phase (2 bit) symbols. This gives
a data rate of about 2.4 Mbit/s, about half of which is used for error
correction (though this may be varied). Each multiplex carries between
6 and 11 radio stations, depending on the bit rate (and hence sound quality)
allocated to each station. The data for all the stations is
distributed amongst all of the carriers. This is to ensure
that interference to a few carriers does not cause more
disruption to a particular station's bit stream than the error
correction software can handle. Time interleaving is also used
to protect against short duration interference spikes.
A more detailed description of COFDM transmission can be
found in a BBC conference paper (hosted by the BBC).
A CD codes audio at a rate of about 1.2 Mbit/s. Using
the same coding standard on DAB would only allow one
station per multiplex - not very practical. Therefore
audio compression is used to reduce the data rate of each station. Audio
compression technqiues all make use of psycho-acoustic coding. This takes
advantage of the fact that when a sound on one frequency is heard,
the ear is rendered much less sensitive to quieter sounds on nearby
frequencies. A psycho-acoustic coder performs a Fourier analysis
of the sound to be coded and calculates a noise floor. Sound below
the noise floor can not be perceived by the listener, so can be
discarded. Only the sound above the noise floor need be calculated
and this requires much fewer bits than coding the whole sound,
as is done on CDs.
The audio coding standard used for DAB is MPEG (Motion Picture Expert
Group) 2 Layer 2, abbreviated to MP2. This samples at 48k samples per
second (twice the maximum audio frequency). Each block of 384 samples
is broken into 32 equal frequency bands of 12 samples each and a
separate noise floor and scale factor is set for each band. Noise
floor and scale factor information is shared between sets of three
blocks to save bits. Stereo stations mostly use a technique known
as joint stereo whereby lower bit rate stereo separation data is
added to a mono signal rather than coding the left and right
channels in full.
A more detailed description of audio coding by David Robinson
can be found here (569 kB, hosted on Digital Radio Tech).
|
