Gscat w
i
R is selected within 0.9 and 0.95.
The channel estimator with lowpass filtering can achieve better performance in the penalty of higher computational complexity. This is because additional operations are necessary for the implementation of a lowpass filter, DFT/FFT, and IDFT/IFFT.
3.5 Computer Simulations
Computer simulations are conducted to evaluate the performance of synchronization and channel estimation in a DVB-T system. The channel model employed is given by the relative delay associated with ith path, respectively. Note that the first term in the numerator in (3.26) represents the line of sight ray. The parameter setting of channel response is summarized in Table 3.5. In simulation, the relationship between SNR and
0
When the system transmit power is normalized to one, then the noise power is given by σ corresponding to a specific 2 Eb N can be generated by 0
and M is the modulation order.
In the first simulation, the performance of synchronization is investigated in a DVB-T system. The simulation parameter setting is listed in Table 3.6. The result of cyclic-prefix-based coarse timing synchronization shown in Fig. 3.16 indicates that coarse timing synchronization may cause an error of one or two samples. Hence, a fine timing synchronization is required to enhance the performance. The result obtained by the pilot-based fine timing synchronization as shown in Fig. 3.17, demonstrates that a precise symbol timing position can be obtained.
In the second simulation, the performance of channel estimation is examined in a DVB-T system. The BER versus Eb N plots obtained by the LS channel estimation 0 with linear interpolation and lowpass filtering in the transform domain are shown in Fig.
3.18 (a)-(b) for the mobile speed of 0 and 20 m/s, respectively. Note that the Jake’s model is used to model the fading channel. From Fig. 3.18 (a), we observe that lowpass filtering can successfully suppress the noise and provides about 1dB improvement as compared with the LS channel estimation for the BER of
v
6 10× −4. This is because that the effect of ICI is negligible for low velocity. On the contrary, in the case of high velocity, the performance of both methods is degraded due to ICI, but the lowpass filtering slightly outperforms than the LS method. These simulation results confirm that the lowpass filtering can obtain better performance than the LS channel estimation.
3.6 Summary
A standard DVB-T system is introduced. The cyclic-prefix based and pilot-based schemes are then constructed to enhance the accuracy of timing and frequency offset estimation over a fading channel. In addition, LS channel estimation with linear interpolation and lowpass filtering in the transform domain are also included for correct estimation of the channel response. A DVB-T system incorporating the previously mentioned technique proves to be robust in a typical urban environment.
Serial
Figure 3.1: Digital implementation of appending cyclic prefix into OFDM signal in transmitter.
Figure 3.2: Transceiver architecture for DVB-T system.
Figure 3.3: Scrambler/descrambler schematic diagram in DVB-T system.
Figure 3.4: Conceptual diagram of outer interleaver and deinterleaver DVB-T system.
(a)
(b)
(c)
(d) (a)
(b)
(c)
(d)
Figure 3.5: Data format of DVB-T system. (a) MPEG-2 transport MUX packet. (b) Randomized transport packets: Sync bytes and randomized data bytes. (c) RS(204, 188, 8) error protected packets. (d) Data structure after outer interleaving.
Figure 3.6: Mother convolutional code of rate 1/2 in DVB-T system.
Figure 3.7: Mapping of input bits onto output modulation symbols, for non-hierarchical transmission modes in DVB-T system.
Figure 3.8: Block diagram of symbol interleaver address generation scheme for 2k mode in DVB-T system.
Figure 3.9: Frame structure in DVB-T system.
Guard
Figure 3.10: Block diagram of synchronization and channel estimation in DVB-T system.
Symbol 1 Symbol 2
Symbol 1
Symbol 1 Symbol 2
Symbol 1
Figure 3.11: Illustration of OFDM signal with cyclic prefix for coarse timing and frequency offset synchronization.
271 OFDM symbols
271 OFDM symbols
Corr.
271 OFDM symbols
271 OFDM symbols 271 OFDM symbols
271 OFDM symbols
Corr.
Figure 3.12: Illustration of pilot-based correlation for fine frequency synchronization in synchronization and channel estimation in DVB-T system.
Channel response at the pilot tone
f
Channel response at the pilot tone Channel response at the pilot tone
ff
Figure 3.13: Illustration of pilot-based channel estimation with linear interpolation in DVB-T system.
Decimation &
Frequency domain Transform domain
Decimation &
Frequency domain Transform domain
Figure 3.14: Block diagram of channel estimation with lowpass filtering in transform domain in OFDM system.
High “virtual frequency” component High “virtual frequency” component
Figure 3.15: Estimated channel amplitude responses in transform domain of OFDM system.
Figure 3.16: Decision variable versus sample index with coarse timing synchronization in DVB-T system.
Figure 3.17: Decision variable versus sample index with fine timing synchronization in DVB-T system.
(a)
(b)
Figure 3.18: BER performances versus Eb N of channel estimation methods in 0 DVB-T system with mobile speed v of (a) 0 m/s. (b) 20 m/s.
Table 3.1: Bit permutations for (a) 2k mode (b) 8k mode in OFDM-based DVB system.
(a) 'i
R bit positions 9 8 7 6 5 4 3 2 1 0
R bit positions i 0 7 5 1 8 2 6 9 3 4
(b) 'i
R bit positions 11 10 9 8 7 6 5 4 3 2 1 0
R bit positions i 5 11 3 0 10 8 6 9 2 4 1 7
Table 3.2: Signal constellation and mapping in DVB-T system. (a) Constellation points are with values of (b) Normalization factors for data symbols.
{
z∈ +n jm} n m, . (a)
Modulation scheme n m
QPSK -1, 1 -1, 1
16QAM with α= 1 -3, -1, 1, 3 -3, -1, 1, 3 16QAM with α= 2 -4, -2, 2, 4 -4, -2, 2, 4 16QAM with α= 4 -6, -4, 4, 6 -6, -4, 4, 6 64QAM with α= 1 -7, -5, -3, -1, 1 , 3, 5, 7 -7, -5, -3, -1, 1 , 3, 5, 7 64QAM with α= 2 -8, -6, -4, -2, 2, 4, 6 ,8 -8, -6, -4, -2, 2, 4, 6 ,8 64QAM with α= 4 -10, -8, -6, -4, 4, 6 ,8, 10 -10, -8, -6, -4, 4, 6 ,8, 10
(b)
Modulation scheme Normalization factor
QPSK KMOD = z/ 2
16QAM with α= 1 KMOD = z/ 10 16QAM with α= 2 KMOD = z/ 20 16QAM with α= 4 KMOD = z/ 52 64QAM with α= 1 KMOD = z/ 42 64QAM with α= 2 KMOD = z/ 60 64QAM with α= 4 KMOD = z/ 108
Table 3.3: Numerical values for OFDM parameters in 8k and 2k modes of 8MHz channel in DVB-T system.
Parameter 8K mode 2K mode
Number of carrier K 6817 1705
Value of carrier number Kmin 0 0
Value of carrier number Kmax 6816 1704
Duration Tu 896 µ 224 s µ s
Carrier spacing 1 Tu 1116 Hz 4464 Hz Spacing between carriers Kmin and Kmax (K −1) Tu 7.61 MHz 7.61 MHz Note1: Values in italics are approximate values
Table 3.4: Required C/N for non-hierarchical transmission to achieve a after the Viterbi decoder for all combinations of coding rates and modulation types in DVB-T system.
BER= ×2 10−4
Table 3.5: Relative power, phase and delay values for standard channel model of DVB-T system.
i ρi [ s]τ µi θi [rad]
1 0.057662 1.003019 4.855121
2 0.176809 5.422091 3.419109
3 0.407163 0.518650 5.864470
4 0.303585 2.751772 2.215894
5 0.258782 0.602895 3.758058
6 0.061831 1.016585 5.430202
7 0.150340 0.143556 3.952093
8 0.051534 0.153832 1.093586
9 0.185074 3.324866 5.775198
10 0.400967 1.935570 0.154459
11 0.295723 0.429948 5.928383
12 0.350825 3.228872 3.053023
13 0.262909 0.848831 0.628578
14 0.225894 0.073883 2.128544
15 0.170996 0.203952 1.099463
16 0.149723 0.194207 3.462951
17 0.240140 0.924450 3.664773
18 0.116587 1.381320 2.833799
19 0.221155 0.640512 3.334290
20 0.259730 1.368671 0.393889
Table 3.6: Simulation parameters of DVB-T system.
Number of transmit/receive antenna 1/1
Data rate 4.98 Mbps
Modulation QPSK
Coding rate 1/2
Symbol time 280 sµ
Guard time 56 sµ
Subcarrier spacing 4.464 KHz
FFT length 2048
System clock 64/7 MHz
OFDM symbols 272 symbols (4 frames/per superframe)
Carrier frequency 474 MHz
Chapter 4
Diversity Reception and Phase Noise Compensation in DVB-T Systems
As mentioned in Section 2.1, diversity is effective to combat detrimental effects in wireless fading channels because the probability of two or more independent faded channels in a deep fade simultaneously is small. On the other hand, oscillator phase noise is a potentially serious problem because of the common necessity to employ relatively low cost tuners in the receivers. It affects communication systems in two ways: CPE and ICI (referred to Section 2.2). We propose some solutions to combat against these effects. They are MRC, low-pass filtering in the transfer domain, and decision feedback phase noise compensator with batch processing. The details will be discussed in sections to come.
4.1 Diversity Reception at Different Stages in DVB-T Systems
We will focus on space diversity reception in this chapter. Space diversity is also known as antenna diversity. In practice, we consider dual receive antennas employed in a DVB-T system. Dual antennas separated physically by a distance can guarantee that the transmitted signals from different antennas go through low correlated fading paths.
This kind of diversity scheme is widely used due to simplicity and low-cost.
Furthermore, it requires no extra frequency spectrum. Space diversity can be divided into two categories: selective and combining diversity schemes. A more robust reception is achieved when using more than one receive antenna by combining or
selecting the signals from different branches. The improved reception can be used to increase the transmission reliability for both portable and mobile television receivers, such as those found in cars. Selective diversity is to select the strongest signal among diversity branches, which is easy for implementation. On the other hand, combining diversity combines signals from different branches to get better performance. The trade-off between BER performance and computational complexity determines which diversity scheme is used. In Fig. 4.1, we use various diversity reception schemes at different stages in a DVB-T system. The discussed diversity reception schemes include MRC, symbol-based selective diversity (SBSD), cyclic delay diversity (CDD), MRC with CDD, in-/post-Viterbi selective diversity (IVSD/PVSD), and packet-based selective diversity (PBSD). For convenience, we assume perfect synchronization and perfect channel estimation. These schemes are described as follows.
4.1.1 Maximal Ratio Combining
In the MRC scheme, the outputs of the dual receive antennas are linearly combined so as to maximize the instantaneous SNR. The coefficients that yield the maximum SNR can be found from the optimization theory in Chapter 2. They are the complex conjugate of the channel responses. The algorithm in the frequency domain is given by
where the superscript denotes the antenna index. From (4.1), we can observe that “full diversity” gain is exploited. Therefore, MRC can get better performance. In an OFDM-based DVB system, each tone needs its own one-tap equalizer for channel compensation. Therefore, MRC needs twice the number of multipliers for channel compensation with single antenna. After introducing the principle and hardware cost of MRC, we compare other diversity schemes with MRC in the following sections.
4.1.2 Symbol-Based Selective Diversity
The average power pj of the OFDM symbol in the jth branch is estimated by the receiver [1]. This estimation is done for each branch, i.e. j = 1, 2. The branch with larger power will be selected
ˆj
ˆ arg max j
j= j p (4.2)
In SBSD, we assume that the signal power and noise power are the same in different branches. Only the channel responses are different in different branches. Moreover, channel responses don’t change significantly in the frequency domain. Therefore, the scheme can suppress deep fading phenomenon by selecting larger symbol power which doesn’t need channel information. As mentioned before, the channel information must be obtained in frequency domain for a DVB-T system. Therefore, SBSD can be proposed to operate before FFT. Fig. 4.1 shows that a receiver using SBSD obtains the same computational complexity as the single antenna case after FFT. Thus, SBSD would need lower computational complexity.
4.1.3 Cyclic delay Diversity
Before referring to CDD, we introduce delay diversity (DD) and phase diversity (PD) whose concepts are similar to that of CDD [2]-[5]. Usually, DD and PD are applied at multiple-antenna transmitter side.
Delay Diversity
In DD, the transmitted multi-carrier modulated signal differs only with the specific delay δm at the mth antenna, where m = 0,…, M-1. The block diagram of an OFDM system with the spatial transmit diversity applying DD is shown in Fig. 4.2. In order to achieve constructive and destructive superposition of the signals, the delay δm has to fulfill the condition
1,
m m
δ ≥ B ∀ (4.3)
where B is the channel coherence bandwidth. Therefore, frequency selectivity is
generated, and long burst errors induced from deep flat fading are avoidable. Fig. 4.3 shows the effect of DD over a typical indoor channel. But the disadvantage of DD is that the additional delays increase the total delay spread at the receiver antenna, i.e. the effective cyclic prefix length is reduced relatively. To overcome this issue, phase diversity scheme is presented.
Phase Diversity
In PD, the transmitted multi-carrier modulated signal differs only with the specific phase offset. The block diagram of an OFDM system with the spatial transmit diversity applying PD is shown in Fig. 4.4. In order to achieve constructive and destructive superposition of the signals, the phase offset Φm k, has to fulfill the condition
, the cyclic prefix. Therefore, frequency selectivity is generated. But the disadvantage of PD is that the scheme operates in the frequency domain, which needs the M OFDM modulation blocks. To prevent additional hardware induced from PD and the effective cyclic prefix length from being shortened by DD, the CDD is exploited in place of PD and DD.
Cyclic Delay Diversity
In CDD, the signal is not truly delayed but cyclically shifted. The basic idea behind the new scheme is to convert time-selectivity of the channel into frequency-selectivity so that channel coding which is applied mainly in frequency domain in the DVB-T system is more effective. The equivalence between PD and CDD is a property of DFT and can directly be seen from the length Nc IDFT definition
1 2
complex-valued signals in time and frequency domain respectively with l, k. δ cy stands for cyclic time shift. From (4.5), the operation for PD has to be done before the OFDM modulation. So for an M-antennas system with PD, the M OFDM modulation blocks have to be constructed. On the other hand, only one OFDM block is required in CDD. Thus the implementation of CDD is more efficient, and has the same functionality as those of DD/PD. However, all schemes of DD, PD and CDD are often applied in the transmitter. To implement them in the receiver side, we derive the equivalence between the transmitter and receiver with CDD. The received signals with CDD applied at the transmitter side can be written as
(1) (1) (1)
And the received signals with CDD applied at the receiver side can be written as
(1) (1) (1)
where Rx denotes the receive antenna. From (4.6a) and (4.6b), the equivalence is mainly based on the linear characteristics of convolution and delay. Fig. 4.5 shows the equivalence model of the transmitter and receiver using CDD. Thus, CDD can be applied in the receiver. From Fig. 4.5, we know that the implementation of CDD only needs an adder and a man-made cyclic delay. The receiver structure remains unchanged compared with the single antenna case after FFT. In addition to lower computational complexity, CDD can also incorporate other diversity schemes if hardware loading can be increased. Finally, we summarize CDD as follows:
(1) Standard compatibility, i.e. CDD can be applied without changing the transmitter.
(2) No necessity of the multiple channel estimators; therefore lower the receiver complexity.
(3) The number of receive antennas is arbitrary.
(4) Low implementation complexity, due to the simple cyclic shifts in the time domain.
(5) CDD can incorporate other diversity schemes to enhance the system performance.
Maximal Ratio Combining with Cyclic Delay Diversity
As discussed above, it is easy to combine CDD with other diversity reception schemes [5], e.g. MRC with CDD, as shown in Fig 4.6. Fig. 4.7 shows the BER performances of MRC, CDD, and MRC with CDD in a DVB-T system with cyclic delay varied as a parameter. We observe that the value of cyclic delay is important, which should be larger than RMS delay spread of a channel model. The RMS delay spread of Rayleigh fading channel model in a DVB-T system is about 1.2689μs (see Section 4.3). Therefore, the system with the cyclic delay δcy =1.8μs gets better performance. On the other hand, MRC with CDD outperforms conventional MRC. This is because MRC with CDD not only gets full diversity gain, but also rejects long burst errors.
4.1.4 In-Viterbi Selective Diversity
In wireless communication, channel fading and random noise effects make points on the constellation shift and thus cause decision errors. In Fig.4.8, we can know that the probability of the effects making signals of different branches shift toward error areas simultaneously is small. Therefore, we exploit the diversity scheme to increase freedom of add-compare-select (ACS) in Viterbi to help suppress the deep fading and random noise effects as shown in Fig.4.9 [6]-[7]. By the Viterbi algorithm, the branch metrics associated with the state transitions are computed and then added to the previous path metrics. The contending path metrics are compared and the path with the largest metrics is as the survivor. With the aid of diversity scheme, the path metrics are computed for each diversity branch, and then compared to select the survivor and the criterion is given by
ˆ arg min j
j= j d (4.7)
where dj denotes the Euclidian distance at the jth diversity path. Moreover, IVSD incorporating interleaver can achieve better performance. This is because interleaver combating deep fading helps IVSD more concentrate on random noise. Therefore, IVSD can suppress deep fading and random noise effects effectively.
4.1.5 Post-Viterbi Selective Diversity
In PVSD, outputs from the two Viterbi decoders are compared and selected based on path metrics [6]-[7], which is similar as IVSD. The selection strategy is also given by the (4.7). There is a main difference between IVSD and PVSD. IVSD is more instantaneous than PVSD. In other words, IVSD selects the optimum survivor state by state and PVSD selects the optimum survivor section by section. In PVSD, each received branch sequence is individually compared and selected after passing through a long path as shown in Fig 4.10.
4.1.6 Packet-Based Selective Diversity
The selection strategy of PBSD is based on the number of errors from the two RS decoders [1]. Let us assume that the number of errors in one packet of each diversity branch is given by e1, e2. The packet selection criterion is given by
ˆ arg min j
j= j e (4.8)
Because the scheme will select the best syndrome, all effects are eliminated to be best of its capacity where RS decoder can correct 8 errors of one packet in a DVB-T system.
Furthermore, there are three possible cases. First, received packets in two branches are equal or less than 8 errors. Second, received packets in two branches are more than 8 errors. Third, received packet in one branch is equal or less than 8 errors and that in the other branch is more than 8 errors. In consequence, in the first case, since either packet is error-free, whether we use PBSD or not doesn’t make any difference. In the second case, since errors in the two packets exceed what RS decoder can tolerate, again whether we use PBSD or not doesn’t make any difference. In the third case, the error-free packet will be the desired one by RS decoder and PBSD will achieve its diversity efficacy. Moreover, when SNR is higher, the random noise effect becomes smaller and other effects, e.g. ICI can be eliminated effectively. Because RS decoder is located at the end of the receiver, a receiver using PBSD will actually consist of the multiple single antenna receivers as shown in Fig. 4.1. Therefore, PBSD needs higher computational complexity than other diversity reception schemes.
4.2 Comparison of Performance and Complexity for Diversity Reception
In principle, the diversity process at the front stage needs less computational complexity. Hardware loading at the posterior stage of SBSD and CDD is the same as that of the single receive antenna. In particular, CDD only requires extra one adder, one man-made cyclic delay, and two radio frequency (RF) components, compared to the original single antenna receiver.
On the other hand, performance comparisons between each diversity scheme are illustrated in Figs. 4.7, 4.11-4.12. There are some superior schemes worth noticing. For instance, MRC implemented after the channel estimator exploits the channel information and thus provides full diversity gain. Moreover, MRC with CDD results in better performance than MRC because the effect of CDD avoids long burst error induced from deep flat fading. Another superior scheme is IVSD incorporating the interleaver, it can combat against the effects of deep fading and random noise effectively.
In summary, we list the complexity and performance comparisons between each diversity scheme as follows:
Complexity: PVSD > IVSD > MRC with CDD > MRC > SBSD > CDD Performance: MRC with CDD > MRC > IVSD > PVSD > SBSD ≈ CDD
4.3 Phase Noise Compensation
The undesirable frequency drift introduced by the local oscillator at the receiver, which is usually called as carrier phase noise, will significantly affect an OFDM signal.
As a remedy, the phase noise compensation scheme is used for enhancement of robustness against phase noise and improvement of system performance. Phase noise basically leads to the detrimental effects CPE and ICI. The effect of CPE will make
As a remedy, the phase noise compensation scheme is used for enhancement of robustness against phase noise and improvement of system performance. Phase noise basically leads to the detrimental effects CPE and ICI. The effect of CPE will make