The DAB syst em allows great flexib

The DAB syst em allows great flexibilit y in the choice of the proper error pro tection
for different applica tions and for diff erent phy sical trans mission chann els. Usin g rate
compat ible punc tured c onvolutio nal (RCP C) co des intr oduc ed by [Hag enauer,
1988], it is possible to use codes of diffe rent redunda ncy wi thout the ne cessity for
different decod ers. One has a fami ly of RC PC codes origin ated by a conv olutional
code of low rate that is called the mother code. The daughter co des will be generat ed
by omitti ng specific redu ndancy bits . This procedure is call ed punc turi ng. The
recei ver must know whi ch bits have been punc tured. Only one Viterbi deco der for
the mother code is necessa ry. The mother cod e used in the DAB system is define d by
the generat ors (133, 171,14 5,133) in octal notati on. The enco der is shown as a shif tregister diagram in Figure 2.6.
The mother code has the code rate R
c ¼ 1/4, that is for each data bit ai the encoder
produces four coded bits x0,i, x1,i, x2,i and x3,i. As an example, the encoder output
corresponding to the first eight data bits may be given by four parallel bit streams
written in the following matrix (first bit on the left hand side):
1 0 1 1 0 1 1 0
1 1 1 1 0 0 1 0
1 1 0 0 1 0 1 0
1 0 1 1 0 1 1 0
A code of rate 1/3 or 1/2, respectively, can be obtained by omitting the last one or two
rows of the matrix. A code of rate 2/3 ( ¼ 8/16) can be obtained by omitting the last
two columns and every second bit in the second column. If we shade every omitted
(punctured) bit, we get the matrix:
1 0 1 1 0 1 1 0
1 1 1 1 0 0 1 0
1 1 0 0 1 0 1 0
1 0 1 1 0 1 1 0
X
0,i
X
1,i
X
2,i
a
i
X
3,i
Figure 2.6 Encoder for the DAB mother code
34 Digital Audio Broadcasting: Principles and Applications of Digital Radio
For 8 data bits now only 12 encoded bits will be transmit ted: the code has rate 8 /12.
Using this method, one can generat e code rates 8/9, 8/10, 8/11, . . . . 8/31, 8/32. The
puncturing pattern can even be changed during the data stream, if the condition of
rate compatibility is taken into account [Hagenauer, 1988].
RCPC codes offer the possibility of Unequal Error Protection (UEP) of a data
stream: some bits in the data stream may require a very low bit error rate (BER),
others may be less sensitive against errors. Using RCPC codes, it is possible to save
capacity and add just as much redundancy as necessary.
UEP is especially useful for audio data. They are organised in frames of 24 ms. The
first bits are the header, the bit allocation (BAL) table, and the scale factor select
information (SCFSI). An error in this group would make the whole frame useless.
Thus it is necessary to use a strong (low-rate) code here. The next group consists
(mainly) of scale factors. Errors will cause annoying sounds (‘‘birdies’’), but these can
be concealed up to a certain amount on the audio level. The third group is the least
sensitive one. It consists of sub-band samples. A last group consists of Programmeassociated Data (PAD) and the Cyclic Redundancy Check (CRC) for error detection
in the scale factor (of the following frame). This group requires approximately the
same protection as the second one. The distribution of the redundancy over the audio
frame defines a protection profile. An example is shown in Figure 2.7.
The PAD may be extended to the so-called X-PAD. In this case, the PAD group
size increases and the audio sub-band sample group decreases. It is important to note
that the error protection does not take this into account. The X-PAD is thus worse
protected (see section 2.3.3.2).
For audio data with a sampling frequency of 48 kHz, the DAB system allows 14
different data rates between 32 and 384 kbit/s. The protection profiles for all these
date rates are grouped into five Protection Levels PL1 to PL5. Inside each protection
level different data rates are possible, but the robustness against errors is the same.
This means, for example, that if a broadcaster switches between 192 and 256 kbit/s,
the audio quality will change, but not the coverage area. PL1 is the most robust
protection level, PL5 the least robust one. All protection levels except PL5 are
24 ms
R
c = 8/24
R
c = 8/18
R
c = 8/14
R
c = 8/19
Group 1: Header, BAL, SCFSI
Group 2: Scale factors
Group 4: PAD,
Scale factor CRC
Group 3: Sub-band samples
Redundancy
Figure 2.7 Example of an audio UEP profile
System Concept 35
designe d for mo bile recep tion; 14 data rates and five protect ion level s lead to 70
possibl e co mbinations. For 64 of them, a protection profi le is define d. Tabl e 2.4
shows the possibl e co mbinations and the requir ed number of cap acity units.
The DAB system allows eight protect ion levels for Equal Error Protect ion (EEP) .
They are intende d for data transmis sion. For the so-called A-pr ofiles 1-A , 2-A, 3-A ,
4-A, all data rates are pos sible that are integ er multiple s of 8 kbit/ s. For the B-profi les
the data rate must be a multiple of 32 kbit/s. Table 2.6 sho ws the eight pro tection
levels and their co de rates . The third column sho ws the number of CUs required for a
64 kbit/ s da ta stre am. The fourth column shows the requ ired SNR to reach a BER of
2  10  4 for TM II in a Rayleigh fading channel with fDmax ¼ 40 Hz. The fifth column
shows the same for fDmax ¼ 125 Hz.
Table 2.4 Capacity needed for the possible combinations of audio
data rates and protection levels
Data Rate PL1 PL2 PL3 PL3 PL5
32 kbit/s 35 CUs 29 CUs 24 CUs 21 CUs 16 CUs
48 kbit/s 52 CUs 42 CUs 35 CUs 29 CUs 24 CUs
56 kbit/s X 52 CUs 42 CUs 35 CUs 29 CUs
64 kbit/s 70 CUs 58 CUs 48 CUs 42 CUs 32 CUs
80 kbit/s 84 CUs 70 CUs 58 CUs 52 CUs 40 CUs
96 kbit/s 104 CUs 84 CUs 70 CUs 58 CUs 48 CUs
112 kbit/s X 104 CUs 84 CUs 70 CUs 58 CUs
128 kbit/s 140 CUs 116 CUs 96 CUs 84 CUs 64 CUs
160 kbit/s 168 CUs 140 CUs 116 CUs 104 CUs 80 CUs
192 kbit/s 208 CUs 168 CUs 140 CUs 116 CUs 96 CUs
224 kbit/s 232 CUs 208 CUs 168 CUs 140 CUs 116 CUs
256 kbit/s 280 CUs 232 Cus 192 CUs 168 CUs 128 CUs
320 kbit/s X 280 CUs X 208 CUs 160 CUs
384 kbit/s 416 Cus X 280 CUs X 192 CUs
Note: It can be seen from the figures in Table 2.4 that the coding strategy supports
many possible changes of configuration. For example, if a 256 kbit/s audio channel is
split up into two 128 kbit/s channels at the same protection level, they will require the
same capacity. Furthermore, in most cases one can increase the protection to the next
better level and lower the audio data rate by one step without changing the required
capacity. Such a diagonal of constant capacity of 140 CUs has been marked by
shading in Table 2.4. It is possible to multiplex several audio channels of different
size together, as long as their total size does not exceed 864 CUs. Table 2.5 shows as
an example the number of 192 kbit/s audio programmes that can be transmitted for
the different protection levels and the signal-to-noise ratio (SNR) that is needed at
the receiver in a typical (not fast) fading channel [Schulze, 1995]. A small capacity
for data services is always left.
Table 2.5 Number of 192 kbit/s audio programmes and
required SNR
Protection Level Number of Programmes SNR
PL1 4 7.4 dB
PL2 5 9.0 dB
PL3 6 11.0 dB
PL4 7 12.7 dB
PL5 8 16.5 dB
36 Digital Audio Broadcasting: Principles and Applications of Digital Radio
The pro tection level s 4 -A and 4-B are very sensitiv e to fast fading . They sho uld not
be us ed for mobil e ap plications .
All the channe l co ding is ba sed on a frame struc ture of 24 ms. These fram es
are called logical fram es. They are synchroni sed with the transmis sion frame s, an d
(for audio) with the aud io fram es. At the beginni ng of one logical frame the co ding
starts with the shift regis ters in the all-zero state. At the en d, the shif t register will
be forced back to the all-zero state by appending 6 additiona l bits (so-call ed tail bits ) to
the useful data to help the Viterbi deco der. After en coding su ch a 24 ms logic al fram e
builds up a punctured co deword . It always contai ns an integ er multiple of 64 bits,
that is an integer num ber of CU s. W henever necessa ry, so me ad ditional punc turing is
done to achieve this . A data stre am of subsequent logic al frames that is cod ed independently of other data stre ams is call ed a sub -channel . For exampl e, an aud io
data stream of 192 kbit/s is such a possible sub-channel . A PAD data stre am is always
only a part of a sub-channel . After the ch annel en coder, each sub -channel wi ll be
time interleave d independ ently as descri bed in the next subsect ion. Aft er tim e
interleavin g, all sub-chan nels are multiple xed togeth er into the common inter leaved
frame (CIF).
Convolu tional codes and their de coding by the Vi terbi algorithm are treated in
textboo ks ab out cod ing, see for exampl e [C lark, 1988], [Bosser t, 1998], [Proaki s,
1995]. The paper of [Hoeher , 199 1b] gives some insi ght into how the chan nel c oding
for DAB audio has be en developed . It reflects the state of the researc h work on this
topic a few mo nths be fore the parame ters were fixe d. Bit error curves of the final
DAB coding scheme in a mobile fading channel and a discus sion of the limit s of the
system can be found in the paper of [Schulze , 1995