Channel Estimation and Symbol Boundary Detection in DVB-T2 System
Ching-Sheng Ni, Pei-Jun Shih, Shin-Shiuan Cheng
Himax Media Solutions Inc., Taipei, Taiwan
Email: jennyshihpj@gmail.com, sean.cheng@stanfordalumni.org
ABSTRACT
A novel channel estimation and symbol boundary detection
algorithm is proposed herein for the DVB-T2 system. The proposed
channel estimation uses devised correlation patterns utilizing parts of
P1 symbol and is followed by proposed successive cancellation to
eliminate the artifacts generated during the correlation. With the
channel information, the inter-symbol interference is estimated and
symbol boundary position can be optimized according to the least ISI.
I. INTRODUCTION
A DVB-T2 signal is constructed by super frames, which consist
of several T2-frames, and each frame is composed of OFDM
symbols, including one P1 OFDM symbol and several P2 and data
OFDM symbols, as shown in Fig. 1. P1 symbol carries information
to indicate the transmission type and key transmission parameters. P2
symbols carry remaining parameters such as the guard interval type,
code rate, etc., and they also carry some data payload [1]. To receive
DVB-T2 signals, P1 symbol should first be detected and decoded for
key parameters. Once P1 symbol has been identified, the symbol
boundary detection could follow. Channel impulse response (CIR)
information can facilitate better symbol boundary detection, and
better CIR estimation can be achieved by taking advantage of P1
symbol. After the CIR is estimated, the symbol boundary can be
decided. The symbol boundary should be positioned so that the least-
possible inter-symbol interference (ISI) is incurred. The incurred ISI
can be estimated by the estimated CIR information. Finally, to avoid
the aliasing caused by frequency interpolation, a phase rotator is
employed and the rotation angle is computed after the proper symbol
boundary is decided.
II. CHANNEL ESTIMATION - CORRELATION
After the received P1 symbol is detected and decoded, P1 symbol
can be regenerated according to the decoded parameters. Parts of P1
symbol is utilized as the correlation pattern. Based on the structure of
P1 symbol, three different tailor-made patterns are devised to take
advantage of the special formatting of P1 symbol.
542
542 482 482
542
482
pattern 1
542
542 482
482
482
542
0
542 1024
0 482
542
0 60 542
P1 symbol
pattern 2
pattern 3
part 1 part 2 part 3 part 4
correlation result
correlation result
correlation result
c(n)
c(n)
c(n)
0 0
0
0 0
0
Figure 2. P1 symbol and correlation results of the devised patterns
As shown in Fig. 2, pattern 1 is constructed by the part2 and the
part3 of P1 symbol. Under perfect channel, the cross-correlation of
pattern 1 with the received signal yields the result shown next to
pattern 1 in Fig. 2. The cross-correlation result has three major pulses.
However, under perfect channel, there should only be one
propagation path, hence one pulse. Therefore, two extra undesired
pulses, due to the artifact of the cross-correlation, should be
eliminated for correct CIR estimation.
III. CHANNEL ESTIMATION – SUCCECIVE
CANCELLATION
Successive cancellation scheme can be employed here to help
eliminate the two undesired paths. In order to facilitate better
working of successive cancelation, two additional patterns - pattern 2
and 3 - for cross-correlation are proposed, as shown in Fig. 2. The
cross-correlation results of these two proposed patterns with the
received signal under perfect channel are also shown in Fig. 2, next
to the pattern respectively. When pattern 2 is employed for the cross-
correlation, the undesired two paths to eliminate are those to the right
of the maximum path, i.e. paths at 482 and 542 as shown in Fig. 2.
On the contrary, if pattern 3 is employed, the undesired two paths to
eliminate are those to the left of the maximum path, i.e. paths at 0
and 60.
Figure 1. DVB-T2 signal structure
The successive cancellation is implemented as follows. If pattern
2 is employed, the cross-correlation result is shown again in Fig. 3.
In this case, the main path at 0 and the Left-to-Right (LtoR)
cancellation are employed to eliminate the two undesired artifact
paths at 482 and 542, hence achieving correct CIR estimation. The
implementation is mathematically described in (1).
2
1
)()542()542(
)()482()482(
~0
D
D
u
u
tststs
tststs
Nt
(1)
,
978-1-4244-8499-7/11/$26.00 ©2011 IEEE