Paper Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

Ryan Paderna †, Yafei Hou ††, Duong Quang Thang †, † † Takeshi Higashino ,MinoruOkada(member)

Abstract In order to realize next generation digital terrestrial television broadcasting (DTTB) systems, the currently deployed Integrated Services Digital Broadcasting Terrestrial (ISDB-T) is still essential due to the spectrum-scarcity. This paper proposes a layered division multiplexing (LDM) scheme to simultaneously transmit ISDB-T stream and a secondary data stream over a common frequency band. The next generation DTTB system can be obtained by combining the ISDB-T and the secondary data streams. In the proposed scheme, we deploy the primary layer for transmission of ISDB-T signal and the secondary layer for transmission of secondary data stream both employed space time block coding. In addition, we only adopt partial LDM to reduce the interference between two systems, since the transmission capacity of the secondary layer exceeded the minimum requirement due to reuse of primary layer.

Key words: ISDB-T, DTTB, partial layered division multiplexing, single frequency network, STBC

In this paper, we accomplish this task by exploiting 1. Introduction the non-orthogonal multiplexing capability of Layered The current Integrated Services Digital Broadcast- Division Multiplexing (LDM) technique3).LDMbe- ing - Terrestrial (ISDB-T) transmits high definition TV came popular for next generation DTTB because of its (HDTV) broadcasting in Japan1). Recently, next gener- efficient frequency utilization and robust performance. ation digital terrestrial television broadcasting (DTTB) In fact, the LDM was employed in Advanced Television systems for higher resolution TV, e.g., 4K-resolution Systems Committee (ATSC) 3.0 as a new technology and 8K-resolution, have been investigated. The field ex- for next generation DTTB system4).LDMiscomposed periment operated by Nippon Hoso Kyokai (NHK) suc- of two layers namely primary layer signal for high power cessfully transmitted a 8K signal using space time block signal transmission, and the secondary layer signal decoding (STBC) over a single frequency network (SFN) signed for high capacity transmission. The signal power system2). The experiment of the 8K DTTB system was assigned at secondary layer is sufficiently low compared conducted in a rural area so that TV subscribers will to the primary layer. not be affected by co-channel interference. However, an Although LDM has a similarity to hierarchical mod- issue needs to be considered is how to implement the ulation, LDM is more flexible compared to hierarchi- next generation DTTB system without requiring new cal modulation. LDM has more freedom on how much frequency resources. Currently, the UHF band from 470 transmission rate will be implemented for primary and MHz to 710 MHz are assigned for the ISDB-T1).One secondary layer. It is because LDM uses two indepen- of the promising way is to transmit the next genera- dent constellations with different transmission power tion DTTB signal with the current ISDB-T signal over overlapping with each other. While hierarchical mod- common frequency band simultaneously. The ISDB-T ulation uses only single constellation for two different receiver only requires simple STBC decoder in order to information (e.g. combination of 4-PSK for primary demodulate properly the ISDB-T stream. layer and 16QAM in the secondary layer results to a 64QAM)5). For 4-PSK in 64QAM hierarchical modu- ReceivedReceived June 13, 2017; Revised ; Revised September 29, 2017; Accepted ; Accepted November 9, 2017 lation, the number of bits per symbol in each subcar- † Graduate School of Information Science, Nara Institute of Science rier is 6, because 2 bits are modulated by using QPSK and Technology (8916-5Takayama, Ikoma-Shi, Nara, 630-0192 Japan) for primary layer, and additional of 4 bits are mod- †† Graduate School of Natural Science and Technology, Okayama ulated for secondary layer. Another example is the University (1-1-1, Tsushima-Naka, Kita-ku, Okayama-shi, 700-8530 Japan) QPSK/16QAM hierarchical modulation, where 4 bits

98 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

per symbol are used6). In QPSK/16QAM hierarchical stream does not lay down over the whole symbols, but modulation, 2 bits are used for primary layer, while partial secondary layer symbols are employed to trans- additional of 2 bits for the secondary data stream can mit additional data. At the receiver, the estimated be implemented. In summary, the conceptual differ- ISDB-T symbol is not only used for cancellation itself, ence between LDM and hierarchical modulation are: 1.) but also the demodulated ISDB-T data can be treated LDM is configured by use of spectrum overlap technol- as supplemental data of next generation DTTB system. ogy while the hierarchical modulation mainly focuses on The contribution of this research is threefold: backward compatibility and it does not based on spec- ( 1 ) The proposed system will reuse the current trum overlapping. 2.) LDM receiver basically equips ISDB-T frequency spectrum resulting in a very signal cancellation to retrieve the secondary layer sig- efficient utilization of the spectrum resources. nal, while the hierarchical receivers does not have such ( 2 ) The primary layer of LDM is both used as an signal cancellation. interference cancellation of secondary layer signal, and part of the transmitted data of the next generation DTTB system. In this way, the primary layer signal will not be wasted since it will be combine with the secondary layer data. ( 3 ) The interference between two layers of LDM will be reduced since not all but part of the transmitted signal are LDM implemented. In this way, the overall error rate of the received data will be reduced. The rest of the paper is organized as follows. Section Fig. 1 LDM for ISDB-T and secondary data stream 2 describes the system configuration of the proposal. Section 3 explains about LDM for ISDB-T receiver and This paper proposes a dual transmission scheme for the proposed next generation DTTB receiver. Section ISDB-T and secondary data stream using LDM as il- 4 will explain the channel estimation. Section 5 shows lustrated in fig. 1. In fig. 1, the primary layer is as- the channel capacity of LDM. Performance evaluation signed for ISDB-T, and the secondary layer is assigned is given in Section 6 and conclusions are drawn in the for secondary data stream. However, LDM is prone last section. to interference due to superposition between the pri- 2. System configuration mary and secondary layer. And also, the primary layer symbol is wasted because it is only use for cancella- Figure 2 shows the system block diagram of the two tion to obtain the secondary layer. In order to solve transmitters and the next generation DTTB receiver in these issues, the proposed receiver reuses the primary SFN. The next generation DTTB system adopts multi- layer to reconstruct higher data rate streams in com- ple transmitters that operates in an identical frequency bination with the secondary data stream. A related band8). Each of the transmitters composes of two work is the augmented data transmission (ADT)-based antennas with different polarities making the system ultrahigh definition (UHD) television transmission sys- equivalent to a multiple input multiple output (MIMO) tem using an existing ATSC terrestrial DTV system7). system. The dual polarized antennas transmit simul- The ADT combines the Moving Picture Experts Group- taneously using horizontal and vertical polarized waves 2 (MPEG-2) based DTV signal, High Efficiency Video for different data streams resulting to an increase of Coding (HEVC)-based DTV signal, and HEVC-based transmission capacity2). Meanwhile, MIMO with STBC ADT signal to form a 4K UHD video. for single stream enhances reliability because of diver- With the reuse of the ISDB-T data for next gener- sity gain. ation DTTB system, we can reduce the transmission The SFN requires the transmitters to be synchronized capacity requirement for secondary data stream result- with each other to avoid intersymbol interference. One ing to less implementation of LDM. With partial LDM of the techniques for synchronization of the transmit- implementation, we can reduce the interference between ters is longer duration cyclic prefix (CP) of OFDM sys- upper and lower layer. Hence, the secondary layer data tems. Longer duration CP makes the transmit delay

99 ITE Trans. on MTA Vol. 6, No. 1 (2018)

" $ " !"#

Fig. 2 system block diagram among different transmitters fall within the duration of Thewaythevalueofg was chosen is by simulating CP. Another promising technique is the STBC applica- different values of g as shown in fig. 3. The simu- tion10). The STBC uses Alamouti code to increase the lation in fig. 3 shows the bit error rate (BER) per- diversity effect of MIMO. formance for the ISDB-T and secondary data stream The proposed system employs LDM where the pri- given the different values of injection coefficient ratio in mary layer is ISDB-T system while the secondary layer the Eb/No=25dB region. The injection coefficient ratio is for transmitting a part of high data rate stream can be calculated by 10 · log(g/(1 − g)). Here we can that can be combined with ISDB-T. The ISDB-T data observe that the optimum value of injection coefficient and secondary data stream are combined in frequency ratio is around 8dB where the BER for the primary and domain before inverse fast Fourier transform (IFFT). secondary layer are almost the same. Since the secondary data stream signal is placed in the The system model can be described as a 4 × 2MIMO secondary layer, it is necessary that the signal power system with a channel defined as hi,j from j-th antenna will be adjusted using the injection level g before com- of the transmitter to i-th antenna of the receiver. The bination with ISDB-T signal. received signal with noise n can be modeled as

y = Hx + n, (1)

where y is a vector deﬁned as ⎡ ⎤ 100 (t) y0 10-1 ⎢ ⎥ ⎢ y(t) ⎥ -2 ⎢ 1 ⎥ 10 ⎢ ⎥ ⎢ . ⎥ 10-3 ⎢ . ⎥ ⎢ (t+3) ⎥ -4 ⎢ y ⎥ 10 ⎢ 0 ⎥ -5 ISDB-T, 16QAM ⎢ (t+3) ⎥ 10 y1 BER 2nd stream, 16QAM ⎢ ⎥ y = ∗ . (2) -6 ⎢ (t+4) ⎥ 10 ⎢ y0 ⎥ ⎢ ∗ ⎥ -7 ⎢ (t+4) ⎥ 10 ⎢ y1 ⎥ ⎢ ⎥ 10-8 ⎢ . ⎥ ⎢ . ⎥ 10-9 ⎢ ∗ ⎥ ⎣ y (t+7) ⎦ -10 0 10 ∗ 2 4 6 8 10 12 14 16 18 (t+7) y1 injection coefficient ratio [dB]

Fig. 3 BER performance for diﬀerent values of injection Each entries of y is a received signal of each antenna. coeﬃcient ratio (t) (t) The y0 and y1 are received signals at time t in hor-

100 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

⎡ ⎤ izontal and vertical polarization, respectively. The no- g · a0 +(1− g) · b0 ∗ ⎢ ⎥ tation (·) denotes the complex conjugate. ⎢ · − · ⎥ ⎢ g a1 +(1 g) b1 ⎥ The channel matrix H is the equivalent channel ma- x(75) = ⎢ ⎥ , (7) ⎣ g · a2 +(1− g) · b2 ⎦ trix for STBC process for primary and secondary layer a3 as shown in ﬁg. 2. The channel matrix H can be deﬁned is for the case of 75% partial LDM where 3 out of 4 as J transmitted signals are LDM implemented. H = , (3) J∗ To avoid pilot interference between ISDB-T system where J is, and secondary data stream system, the pilot for ISDB-T ⎡ ⎤ is shared with the secondary data stream system. The h h h h ⎢ 00 01 02 03 ⎥ scattered pilots (SP) are transmitted using ⎢ ⎥ ⎢ h10 h11 h12 h13 ⎥ ⎡ ⎤ ⎢ ⎥ SPk ⎢ h01 −h00 h03 −h02 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ − − ⎥ SPk ⎢ h11 h10 h13 h12 ⎥ ⎢ ⎥ , (8) J = ⎢ ⎥ . (4) ⎢ ⎥ ⎢ h −h −h h ⎥ ⎣ 0 ⎦ ⎢ 02 03 00 01 ⎥ ⎢ − − ⎥ 0 ⎢ h12 h13 h10 h11 ⎥ ⎢ ⎥ ⎣ h03 h02 −h01 −h00 ⎦ in the k-th subcarrier of transmitted signal x. The sub-

h13 h12 −h11 −h10 carrier position of the SP is the same all throughout This type of channel matrix for STBC will have an the transmitted signals. The section 4 about channel equivalent STBC rate of 1/2 which is 4 transmitted estimation will explain the details on how the received symbols per 8 time instances. The transmitted signal x pilots where combined to avoid interference with each is a composite symbol of primary and secondary layer. other. It can be represented by 3. Proposed LDM decoder ⎡ ⎤ g · a +(1− g) · b ⎢ 0 0 ⎥ 3. 1 Primary layer LDM decoder ⎢ · − · ⎥ ⎢ g a1 +(1 g) b1 ⎥ x = ⎢ ⎥ , (5) ⎣ g · a2 +(1− g) · b2 ⎦

g · a3 +(1− g) · b3 where ai and bi corresponds to transmitting symbols for primary and secondary layer in the i-th symbol, respectively. Here we can observed that both ISDB-T signal and secondary data stream signal are both multiplied to the channel matrix H for STBC. This implies that the ISDB-T signal also undergoes STBC process as shown in fig. 2. In this paper, we want to propose a partial LDM assuming that the achievable capacity of the secondary layer exceeds the minimum requirements due to the reuse of the ISDB-T data. The proof is provided in section 5. In that way, we need not to implement LDM for all the transmitted signal from ISDB-T transmission and next generation DTTB transmission. With partial Fig. 4 ISDB-T receiver with STBC decoder LDM, the Eq.(5) can be simplified into ⎡ ⎤ g · a +(1− g) · b ⎢ 0 0 ⎥ Figure 4 shows the block diagram of the ISDB-T re- ⎢ · − · ⎥ ⎢ g a1 +(1 g) b1 ⎥ x(50) = ⎢ ⎥ , (6) ceiver. Since the ISDB-T receiver has only single an- a ⎣ 2 ⎦ tenna, the received signal in Eq.(2) and the channel a 3 matrix in Eq.(3) are different for the case of ISDB-T. for 50% partial LDM where 2 out of 4 transmitted sig- The received signal for ISDB-T y(up) in the primary nals only use LDM, while layer LDM is defined to be

101 ITE Trans. on MTA Vol. 6, No. 1 (2018)

y(up) =gH(up)a +(1− g)H(up)b +n(up), (9) Hermitian transpose, respectively. The Eq.(12) also de-

T T scribes the STBC block in fig. 4 where the y(up) is the T T where a = a0 a1 a2 a3 and b = b0 b1 b2 b3 received signal from the buffer. are the transmitted ISDB-T data symbols and sec- 3. 2 Detection for secondary layer data ondary data stream symbols, respectively. The entries stream of y(up) are defined as, ⎡ ⎤ y(t) ⎢ 1 ⎥ ⎢ . ⎥ ⎢ . ⎥ ⎢ ⎥ ⎢ (t+3) ⎥ ⎢ y1 ⎥ y(up) = ⎢ ∗ ⎥ , (10) ⎢ y (t+4) ⎥ ⎢ 1 ⎥ ⎢ . ⎥ ⎣ . ⎦ ∗(t+7) y1 which are composed of symbols from received single antenna. And also, the channel matrix for ISDB-T H(up) can be defined as ⎡ ⎤ h h h h ⎢ 10 11 12 13 ⎥ ⎢ − − ⎥ ⎢ h11 h10 h13 h12 ⎥ ⎢ ⎥ ⎢ h −h −h h ⎥ ⎢ 12 13 10 11 ⎥ !" ⎢ − − ⎥ ⎢ h13 h12 h11 h10 ⎥ H(up) = ⎢ ⎥ . (11) ⎢ h∗ h∗ h∗ h∗ ⎥ ⎢ 10 11 12 13 ⎥ ⎢ ∗ − ∗ ∗ − ∗ ⎥ ⎢ h11 h10 h13 h12 ⎥ ⎢ ∗ − ∗ − ∗ ∗ ⎥ ⎣ h12 h13 h10 h11 ⎦ Fig. 5 receiver configuration ∗ ∗ − ∗ − ∗ h13 h12 h11 h10

The term (1 − g)H(up)b in Eq.(9) will add an inter- For initialization, the proposed receiver will estimate ference to the ISDB-T received signal. To lessen the the ISDB-T signal in the primary layer LDM which is interference to the ISDB-T signal, the proposed sys- basically the same procedure in the Section 3. 1. The tem uses only partial LDM where only part of the total channel estimation is also using compressed sensing transmitted signal uses LDM11).Inthisway,wecan based impulse response estimator as of section 3. 1. The minimize the value of b which results to a less inter- details about channel estimation is describe in section 4. ference to the received ISDB-T signal. In addition, the And then, the secondary data stream in the secondary injection coefficient g is properly controlled to lessen the layer will be estimated before combining the two data signal power of the secondary data stream for further streams to form a data for next generation DTTB sys- reduction of interference. tem. Using the received pilot signals, the estimated chan- Figure 5 shows the block diagram of the proposed ˆ nel for ISDB-T H(up) can be obtain using compressed next generation DTTB receiver. Here we can see that sensing in time domain12). The channel estimation is the receiver is composed of two antennas with different ˆ explain in section 4 where each entries of H(up) was ob- polarities. The equivalent low pass received signal and tained using compressed sensing based impulse response channel response are denoted by y and H, of equations estimator. And then, the estimated channel for ISDB-T (2) and (3), respectively. ˆ H(up) will be use to decode the ISDB-T symbols a1, a2, At first, using the received pilots, the channel can be 10) 12) a3,anda4 using STBC with the following equation estimated using compressed sensing .Andthen,diversity combined signal using STBC decoding can be Hˆ H aˆ = (up) y , (12) represented by, ˆ (up) H(up) Hˆ H xˆ = y, (13) T Hˆ T where aˆ = aˆ0 aˆ1 aˆ2 aˆ3 is the estimated signals T T for ISDB-T. The notation · and (·)H denote norm and where xˆ = xˆ0 xˆ1 xˆ2 xˆ3 . Estimated symbol of

102 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

secondary layer bˆ can be obtained as, 1 bˆ = (xˆ − g · aˆ). (14) 1 −1 (t+2) (t+3) (t+6) (t+7) − δ = F E y +y +y +y , (22) 1 g 12 4 1 1 1 1 Finally, using the combiner, we can exploit the ISDB-T data aˆ and the secondary data stream bˆ to acquire the whole transmitting data. 1 −1 − (t+2) (t+3)− (t+6) (t+7) δ13= F E y1 +y1 y1 +y1 , (23) 4. Channel estimation using compressed 4 sensing where F and E are the N-size FFT matrix and pilot

This section will discuss the channel estimation for arrangement matrix, respectively. The vector yi is the both ISDB-T receiver and next generation DTTB re- OFDM received signal for all subcarriers of the i-th an- ceiver using compressed sensing (CS). Although other tenna in vector form with size N. The FFT matrix can channel estimation methods are also applicable in this be described as system, CS was chosen because the channel is consid- √1 − 2πkn ered as sparse. CS is a popular technique for estimating F = exp j − , (24) N N 0<=k<(N 1) 13) − a sparse information using only few measurements .A 0<=n<(N 1) sparse can be defined as a vector where most entries are while pilot arrangement matrix E is a diagonal zero and only few are non-zero. The sparse information matrix where the diagonal part contains e = that we are going to estimate is the channel impulse T 12) ··· response (CIR) . The CS equation can be defined as e0 e1 eN−1 . Each entries of e is defined as, ˆ ⎧ δ = Φδ + n, (15) ⎨ 1 (i =0) SP where δ, Φ,andδˆ is the observed CIR vector, measure- eCm+i = ⎩ m , (25) 0(0

1 − (t) (t+1) (t+4) (t+5) T δ = F 1E y − y + y − y , (16) ··· 00 4 0 0 0 0 where Q =diag(q), q = q0 q1 qN−1 that indicates the position of the pilots in the subcarriers. Each entries of q is deﬁned as, 1 δ = F−1E y(t) + y(t+1) + y(t+4) + y(t+5) , (17) 01 4 0 0 0 0 1(i =0) qCm+i = . (27) 0(0

1 The algorithm 1 shows the details about the OMP. δ = F−1E y(t) − y(t+1) + y(t+4) − y(t+5) , (20) 10 4 1 1 1 1 For the initialization of OMP, the algorithm needs only the observed CIR δ and the number of iteration η.After obtaining the estimated CIR δˆ, the estimated channel 1 −1 (t) (t+1) (t+4) (t+5) δ = F E y + y + y + y , (21) ˆ ˆ −1 ˆ 11 4 1 1 1 1 h can now be acquire by h = F δ.

103 ITE Trans. on MTA Vol. 6, No. 1 (2018)

Algorithm 1 Orthogonal Matching Pursuit (OMP) respectively. We assume that the SNRu = 20dB, and Require: δobserved vector SNRl = 25dB which is a reasonable value for ﬁxed re- ˆ Ensure: δestimated sparse vector ceiver with roof-top antennas3). ri ←δ Φ = φ0 φ1 ··· φN For the DTTB system using ADT, the capacity will ← Λi [] only require 12Mbps for MPEG-2 signal and 12.5Mbps Yi ← [] 7) i ← 1 for HEVC signal to form a 4K (30 frames) UHDTV . while i <= η do H Assuming that the ISDB-T only requires 3 bps/Hz for λi = arg maxk |ri−1φk| Λi = Λi−1 ∪{λi} a 6 MHz bandwidth, the achievable spectral eﬃciency

Yi =[Yi−1,φλi ] 2 for secondary data stream will be around 24Mbps based wi =argminw ||Yiw − δ||2 ri = δ − Yiwi on the ﬁg. 6. A 24Mbps channel capacity is assumed i = i +1 to have an ideal channel coding, and a perfect chan- end while return δˆ = wη nel estimation wherein pilots are unnecessary. We also assumed that the receiver side uses a maximum ratio combiner for the received signals. A 24Mbps for sec- 5. Achievable channel capacity ondary data stream is too high since the requirement is only around 12.5Mbps. With the reuse of the ISDB-T data for next generation DTTB system, we can reduce

7 the capacity load for the secondary data stream. In this way, we proposed a partial LDM to reduce the achiev- 6 able capacity of the secondary data stream while reduc- 5 ing the error rate due to interference between primary

4 and secondary layer of LDM.

3 6. Numerical analysis ISDB-T (bps/Hz) 2

1 Table 1 simulation parameters LDM layer primary secondary 0 0 1 2 3 4 5 6 7 data ISDB-T secondary data stream secondary data stream (bps/Hz) modulation QPSK, QPSK, 16QAM, 64QAM, 16QAM 256QAM Fig. 6 ISDB-T and secondary data stream capacity channel estimation OMP OMP with perfect trade-off for LDM (CE) ISDB-T detector OMP iteration (η) 6 6 injection coefficient ratio QPSK in fig.7→ 12dB 10 · log(g/(1 − g)) QPSK in fig.8,10 → 6dB Figure 6 shows the achievable channel capacity trade- 16QAM in fig.9,11,12 → 8dB off between ISDB-T and secondary data stream using 64QAM in fig.12 → 7.5dB 256QAM in fig.12 → 6.9dB LDM. The channel capacity for the primary layer can FFT size (N) 8192 be obtained using number of pilot (M) 1024 pilot type scattered R =R · E log 1+ guard interval N/16 u OFDM frame 10,000 channel model TU6 7 2 SNRu || ||2 (28) g [H(up)]nm noise type AWGN Tx , 1+(1− g)2 SNRu ||[H ] ||2 nm=0 Tx (up) nm while the channel capacity for the secondary layer is 2 SNRl 2 Rl = R · E log 1+(1− g) ||[H]|| . (29) Tx Table 2 TU6 delay profile A derivation of the eq.(28) is provided in the appendix delay (μs) power (dB) 0 -3 section A for further explanation. The R and Tx are the 0.2 0 STBC maximal rate and number of transmitter anten- 0.5 -2 1.6 -6 9) nas, respectively . While the SNRu and SNRl refers 2.3 -8 to the SNR threshold for primary and secondary layer, 5.0 -10

104 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

Table 3 bit rate order modulation such as QPSK. In the next simula- modulation type full LDM 75% LDM 50% LDM tions, a STBC decoder will be added to the ISDB-T QPSK 7.11Mbps 5.33Mbps 3.56Mbps 16QAM 14.22Mbps 10.67Mbps 7.11Mbps receiver to lessen the error rate, and to allow higher 64QAM 21.33Mbps 16.00Mbps 10.67Mbps order modulation for secondary data stream. 256QAM 28.44Mbps 21.33Mbps 14.22Mbps 1024QAM 35.56Mbps 26.67Mbps 17.78Mbps For the ISDB-T with STBC decoder, the simulation compares the BER in cases of two transmitters and single transmitter. And then, we will evaluate the perfor- The computer simulation was performed using C++ mance of the proposed partial LDM in comparison with library for signal processing and communication16).Ta- full LDM. The performance for higher modulation will ble 1 shows the simulation parameters of the experi- be discuss in the last part of the simulation. The pairs ment for both primary and secondary layers of LDM. in the simulations (e.g. pair A, pair B, and pair C) indi- For simplicity purposes, a forward error correction for cate the pair of primary and secondary layers of LDM. primary layer is assumed to be ideal e.g. perfect ISDB- For example in fig. 8, pair A describes the ISDB-T as T detector. Table 2 shows the delay profile of the TU6 a primary layer, and 2nd stream as a secondary layer channel model used in the simulation17). In this simula- wherein the system has 2 transmitters. On the other tion, the performance of the ISDB-T system for primary hand, the pair B indicates that the primary layer is layer, and the secondary layer are investigated sepa- theISDB-T,andthesecondarylayeristhe2ndstream rately. The Eb/No that was used in this experiment is where the system only uses 1 transmitter. the Eb/No of the secondary layer. Table 3 shows the From all the results of the shown figures, we always bit rate for different modulation type vs. LDM type observed that the ISDB-T receiver has low error rate ifthechannelcharacteristicisignore.Thebitrateis compared to that of secondary data stream because pri- calculated using mary layer has higher power compared to the secondary · − · (STBC rate) (N M) log2(modulation type) layer. duration of effective OFDM symbol Figure8showstheBERperformanceoftheISDB-T where the STBC rate are 1/2, 3/8, and 1/4 for full receiver and secondary data stream using QPSK. In fig. LDM, 75% LDM, and 50% LDM, respectively. The 8, we compare the performance with two transmitters OFDM effective symbol period is 1008μs1). The bit rate versus single transmitter using full LDM. In order to re- shown in the table is both applicable for primary and alizesingletransmitter,thevaluesofh02,h12,h03,h13 of secondary layer since both layer exhibit STBC trans- J and H(up) will be set to zero. The h02,h12,h03,h13 are mission. the channels for transmitter B in fig. 2. While for the

At first, we will evaluate the performance of the sys- case of two transmitter, the values of h02,h12,h03,h13 tem using the conventional ISDB-T receiver without of J and H(up) will be nonzero. In fig. 8, the error rate STBC decoder.The conventional ISDB-T receiver can for 2 transmitters is less compared to single transmit- demodulate without any problem using ter. It is found that the diversity improves the perfor- ⎡ ⎤ mance when the number of transmitter increases. In g · a +(1− g) · b0 ⎢ ⎥ the Eb/No region of 25dB, we can improve up to 10dB ⎢ · − · ⎥ ⎢ g a +(1 g) b1 ⎥ ⎢ ⎥ , (30) for ISDB-T, while 5dB for secondary data stream. ⎣ (1 − g) · b2 ⎦ Similarly, the BER performance of 16QAM is shown (1 − g) · b3 in fig. 9. In this figure, the performance for 2 transmit- as a transmitted signal for Eq.(5). For simplicity, we use ters has less error rate compared with single transmit- thesameISDB-Tsymbola for all time instances t to ter. In Eb/No region of 25dB, the diversity improves t + 7. The conventional ISDB-T detection is described the performance of both layers by around 5dB. in appendix section B. Figure 7 shows the bit error rate Figure 10 shows the BER performance of partial (BER) performance for the conventional ISDB-T, and LDM. In this figure, we compare partial LDM of 50% the secondary data stream for next generation DTTB and 75%, with the full LDM. The transmitted signal receiver. The BER performance for ISDB-T is within for partial LDM of 50% and 75% can be described in acceptable error rate in the Eb/No=25dB region using Eq.(6) and Eq.(7), respectively. Here we observed that QPSK for both layers. However, the performance for the performance of the secondary data stream is the the secondary data stream is only limited to the low same in spite of partial use in LDM. The partial LDM

105 ITE Trans. on MTA Vol. 6, No. 1 (2018)

slightly changes the BER performance of the ISDB-T. 0 10 In fig. 10, we observed a higher error rate in case of ISDB-T full LDM compared to ISDB-T partial LDM. The difference of the ISDB-T full LDM and ISDB-T 10-1 partial LDM is of 1dB because the modulation that was implemented is only QPSK. BER However, for the case of 16QAM as shown in fig. 11, -2 we observed that the error rate for the ISDB-T partial 10 LDM of 50% and 75% is around 2dB and 1dB lower compared to full LDM in the Eb/No region of 25dB. 2nd stream ISDB-T without STBC decoder Theoretically, partial LDM realizes lower interference 10-3 5 10 15 20 25 compared to full LDM because the value of the term Eb/No [dB] gH(up)b in Eq. (9) is lesser which represents the sec- Fig. 7 BER performance using conventional ISDB-T re- ondary layer signal. ceiver without STBC decoder, and the secondary data stream To fully qualify for 4K transmission, the primary layer should have at least 12Mbps, while the secondary layer should be at least 12.5Mbps7). Based on table 3, the primary layer and secondary layer can use 16QAM or higher order using full LDM. In fig. 11, the full LDM 100 for both layers using 16QAM qualifies to transmit a 4K 10-1 transmission. Other option is to use a 16QAM for pri- 10-2 mary layer, and 64QAM in the secondary layer. How- -3 10 ever, the transmission capacity using 64QAM full LDM 10-4

is around 21.33Mbps which is high enough in compar- BER 10-5 ison to a minimum requirement of 12.5Mbps. In that 10-6 way, a 75% partial LDM can be applied to reduce the 10-7 transmission capacity to 16Mbps and lessen the bit er- ISDB-T, 2 transmitters, pair A 10-8 2nd stream, 2 transmitters, pair A ror rate in the primary layer. Figure 12 shows the BER ISDB-T, 1 transmitter, pair B 2nd stream, 1 transmitter, pair B 10-9 performance of the primary layer set to 16QAM, to- 5 10 15 20 25 gether with various modulation type in the secondary Eb/No [dB] layer. All the parameters in ﬁg. 12 is capable of 4K Fig. 8 BER performance for 2 transmitters vs. 1 transmitter using QPSK full LDM transmission. In this ﬁgure, the Eb/No is set to 25dB. Here we can observe that we can gradually reduce the error for the primary layer using partial LDM. In summary, the BER performance for ISDB-T without STBC decoder is within acceptable level. However, 100 the modulation type for secondary layer is only limited to low order type, e.g. QPSK. With the ISDB- -1 T receiver with STBC decoder, it can lessen the error 10 rate which opens an opportunity for the secondary layer to implement higher order modulation. Increasing the 10-2 BER number of transmitters can also lessen the error rate for both primary and secondary layer due to STBC di- 10-3 versity. In addition, implementing a partial LDM can ISDB-T, 2 transmitters, pair A 2nd stream, 2 transmitters, pair A lessen further the error rate due to less interference be- ISDB-T, 1 transmitter, pair B 2nd stream, 1 transmitter, pair B tween primary and secondary layer. 10-4 5 10 15 20 25 Eb/No [dB]

Fig. 9 BER performance for 2 transmitter vs. 1 transmitter using 16QAM full LDM

106 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

100 7. Conclusions 10-1 This paper proposed an eﬃcient broadcast transmis- 10-2 sion system for the ISDB-T system and next genera- 10-3 tion terrestrial DTTB system. For eﬃcient frequency 10-4 spectrum utilization, we implemented LDM to transmit

BER -5 10 the ISDB-T in the primary layer and a secondary data -6 10 stream in the secondary layer of LDM. The proposed ISDB-T, full LDM, pair A 10-7 2nd stream, full LDM, pair A ISDB-T, partial 75%, pair B system is carried out using space time block coding in a 10-8 2nd stream, partial 75%, pair B ISDB-T, partial 50%, pair C single frequency network to increase the diversity gain. 2nd stream, partial 50%, pair C 10-9 In order not to waste the primary layer of LDM, the 5 10 15 20 25 Eb/No [dB] next generation DTTB system reuses the primary layer

Fig. 10 BER performance for full LDM vs. partial by combining the ISDB-T data and the secondary data LDM using QPSK stream. In this way, we can lessen the interference between two layers of LDM by implementing partial LDM where only few transmitted signal uses LDM. Partial LDM can be applied since the achievable capacity for the secondary layer exceeds the required capacity due 100 to reuse of primary layer. The simulation of the conventional ISDB-T without 10-1 STBC decoder shows an acceptable error rate of 10−3 to 10−2 in the Eb/No=25dB region. However, the mod-

10-2 ulation type for the secondary layer is only limited to BER QPSK, which suggest an implementation of STBC decoder for ISDB-T. The simulation for LDM-STBC with 10-3 ISDB-T, full LDM, pair A 2nd stream, full LDM, pair A multiple transmitters shows that error rate for both lay- ISDB-T, partial 75%, pair B 2nd stream, partial 75%, pair B ISDB-T, partial 50%, pair C ers can be lessen compared to the system with only 2nd stream, partial 50%, pair C 10-4 single transmitter. In addition, the BER performance 5 10 15 20 25 Eb/No [dB] of primary layer was further improved by introducing

Fig. 11 BER performance for full LDM vs. partial partial LDM. To fully qualify for 4K transmission, both LDM using 16QAM layers should be at least 16QAM, or a partial LDM can be applied in the secondary layer when using modulation type higher than 16QAM.

Appendix

10-1 A. Derivation of Eq.(28) Let us first define a ergodic spectral efficiency of STBC as9) -2 10 R · E log(1 + SNR) ,

BER where the STBC maximal rate R

10-3 T /2+1 R = x , Tx

256QAM 64QAM for Tx even, and 16QAM 10-4 50 60 70 80 90 100 (Tx +1)/2+1 partial LDM [%] R = , Tx +1 Fig. 12 BER performance for primary layer using various modulation type in secondary layer; for Tx odd. Eb/No=25dB, 16QAM for primary layer For simplicity, let us assumed that Tx =2.The

107 ITE Trans. on MTA Vol. 6, No. 1 (2018)

Alamouti scheme operates across two successive chan- Here we can measure the channel (h10 + h11) because a nels i.e. subcarriers j =1, 2. The received signal for contains the pilots based on Eq. 8. Dividing the whole the primary layer can be described as equation above with (h10+h11) will result to a detection ofa ˜ y [j]=g SNR H a[j]+ (up) u (up) − y(t) n (1 g) SNRuH(up)b[j]+n[j]. a˜ = 1 = g · a · + eff . (h10 + h11) (h10 + h11) The b[j] can be treated as Gaussian noise with zero mean, covariance matrix I/2where1/2 accounts for the The value of g should be high enough for decent detec- power allocation of 2 antennas. As a result, the effective tion of a. The process above is also the same in other √ (t+1) (t+7) receive signals y1 , ..., y1 . noise neff[j]=(1− g) SNRuH(up)b[j]+n[j]. For each received signal, the effective noise power can References 1) Association of Radio Industries and Businesses : “Transmis- − 2 || ||2 be approximated to be 1+(1 g) SNRu [H(up)]nm /2, sion system for digital terrestrial television broadcasting”, ARIB STD-B31 version 1.6-E2 (2005) where [H(up)]nm is the nm-th row of H(up). 2) S. Saito, T. Shitomi, S. Asakura, A. Satou, M. Okano, K. Mu- For example, in the case of two channels e.g. j =1, 2 rayama, and K. Tsuchida : “8K terrestrial transmission field using dual-polarized MIMO and higher-order modulation OFDM”, , the received signal can be written as IEEE Transactions on Broadcasting, vol. 62, no. 1, pp. 306-315 (March 2016) 3) L. Zhang, W. Li, Y. Wu, X. Wang, S. I. Park, H. M. Kim, y [1] H( ) H( ) (up) nm up nm,1 up nm,2 ∗ =g SNRu ∗ ∗ · J. Y. Lee, P. Angueira, and J. Montalban : “Layered-division- y [2] H − H multiplexing: theory and practice”, IEEE Transactions on (up) nm (up) nm,2 (up) nm,1 Broadcasting, vol. 62, no. 1, pp. 216-232 (March 2016) 4) S. I. Park, J. Y. Lee, S. Myoung, L. Zhang, Y. Wu, J. Montalban, u1 [neff[1]]nm · + S.Kwon,B.M.Lim,P.Angueira,H.M.Kim,N.Hur,andJ.Kim u2 [neff[2]]nm : “Low complexity layered division multiplexing for ATSC 3.0”, IEEE Transactions on Broadcasting, vol. 62, no. 1, pp. 233-243 T (March 2016) where u1 u2 is the transmitted information sym- 5) A. Schertz, and C. Weck : “Hierarchical modulation - The trans- bols with power 1/2. And then, projecting the transmission of two independent DVB-T multiplexes on a single frequency”, EBU Technical Review (April 2003) mitted vector into k-th column of the channel matrix 6) H. Jiang, and P. Wilford : “A Hierarchical Modulation for Up- grading Digital Broadcasting Systems”, IEEE Transactions on will result to a signal equal to Broadcasting, vol. 51, no. 2, pp.223-229 (June 2005) 7) S. I. Park, G. Lee, H. M. Kim, N. Hur, S. Kwon and J. Kim : || ||2 “ADT-based UHDTV transmission for the existing ATSC terres- g SNRu [H(up)]nm uk + wk trial DTV broadcasting”, IEEE Transactions on Broadcasting, − vol. 61, no. 1, pp. 105-110 (March 2015) where noise wk has a power equivalent to 1 + (1 8) A. Mattsson : “Single frequency networks in DTV”, IEEE Trans- 2 || ||2 actions on Broadcasting, vol. 51, no. 4, pp. 413-422, (Dec. 2005) g) SNRu [H(up)]nm /2. Finally, the SNR can be cal- 9) D. Gomez-Barquero, and O. Simeone : “LDM versus FDM/TDM culated as for unequal error protection in terrestrial broadcasting systems: an information-theoretic view”, IEEE Transactions on Broad- g2 · SNR ||[H ] ||2 SNR = u (up) nm casting, vol. 61, no. 4, pp. 571-579 (Dec. 2015) − 2 · || ||2 10) S. M. Alamouti, “A simple transmit diversity technique for wire- 2(1 + (1 g) SNRu [H(up)]nm /2) less communications”, IEEE Journal on Selected Areas in Com- B. Conventional ISDB-T detection munications, vol. 16, no. 8, pp. 1451-1458 (Oct. 1998) 11) R. Paderna, Y. Hou, and M. Okada : “Space-time block coding for Eq.(30) in a single frequency network with partial layered division mul- The way the detection of conventional ISDB-T re- tiplexing for DTV terrestrial systems”, ITE Tech. Rep., vol. 41, no. 11, BCT2017-48, pp. 37-40, (March 2017) ceiver is for every time instance. For example in time 12) R. Paderna, D. Q. Thang, Y. Hou, T. Higashino, and M. Okada (t) : “Low-complexity compressed sensing-based channel estimation instance t at the received signal y1 , the received signal with virtual oversampling for digital terrestrial television broad- will be casting”, IEEE Transactions on Broadcasting, vol. 63, no. 1, pp. 82-91 (March 2017) (t) 13) D. L. Donoho : “Compressed sensing”, IEEE Transactions on y =g · a · h10 +(1− g) · b0 · h10 + g · a · h11+ 1 Information Theory, vol. 52, no. 4, pp. 1289-1306 (April 2006) 14) J. A. Tropp, and A. C. Gilbert : “Signal recovery from random (1 − g) · b1 · h11 +(1− g) · b2 · h12+ measurements via orthogonal matching Ppursuit”, IEEE Trans- actions on Information Theory, vol. 53, no. 12, pp. 4655-4666 (1 − g) · b3 · h13 + n (Dec. 2007) Theeffectivenoisen will be 15) V. Tarokh, H. Jafarkhani, and A. R. Calderbank : “Space-time eff block coding for wireless communications: performance results”, IEEE Journal on Selected Areas in Communications, vol. 17, no. neff =(1 − g) · b0 · h10 +(1− g) · b1 · h11+ 3, pp. 451-460 (March 1999) 16) IT++ : http://itpp.sourceforge.net/4.3.1/index.html (1 − g) · b2 · h12 +(1− g) · b3 · h13 + n. 17) COST207 : “Digital land mobile communications”, Commission of the European Communities, Directorate General Communica- The received signal can now be simplified as tions, Information Industries and Innovation, pp. 135-147 (1989)

(t) · · y1 = g a (h10 + h11)+neff .

108 Paper » Partial Layered Division Multiplexing with STBC for ISDB-T and Next Generation DTTB Systems

Ryan Paderna received the degree in Bach- elor of Science in Electronics and Communications Engineering from University of San Carlos, Cebu, Philippines, in 2009, and the M.E. degree from Nara Institute of Science and Technology (NAIST), Nara, Japan, in 2015. He is currently pursuing a Ph.D. degree in the Graduate School of Information Sci- ence at NAIST. His research interests are OFDM, MIMO systems, and Digital Terrestrial Television System. Yafei Hou received his B.S. degree in Elec- tronic Engineering from Anhui Polytechnic Univer- sity, China, in 1999. The M.S. degree in Computer Science was received in 2002 fromWuhan University, China. He received the Ph.D. degrees from Fudan University, China and KochiUniversity of Technol- ogy (KUT), Japan in 2007. He was a post-doctoral research fellow at Ryukoku University, Japan from August 2007 to September 2010. He was a research scientist at Wave Engineering Laboratories, ATR Institute International, Japan from October 2010 to March 2014. He became an Assistant Profes- sor at the Graduate School of Information Science, Nara Institute of Science and Technology, Japan from April 2014 to March 2017. Currently, he is an Assistant Professor at Graduate School of Nat- ural Science and Technology, Okayama University. He has obtained twice IEICE Communications So- ciety Excellent Paper Awards in 2016 and 2017. His research interest are communication systems, wireless networks, and signal processing. Dr. Hou is a senior member of IEEE and member of IEICE. Duong Quang Thang received his B.E., M.E. and Ph.D. degrees in Communications Engi- neering from Osaka University, Osaka, Japan, in 2009, 2011 and 2014, respectively. He is currently working as an assistant professor at Nara Institute of Science and Technology. His research interests include broadband wireless access techniques, channel estimation, information theory, correcting code, wireless power transfer and simultaneous wireless information and power transfer. He is a member of the IEICE and the IEEE. Takeshi Higashino received the B.E., M.E. and Ph.D. degrees in Communications Engi- neering from Osaka University, in 2001, 2002 and 2005 respectively. From 2005 to 2012, he was an As- sistant Professor in the Division of Electrical, Elec- tronic and Information Engineering at Osaka Uni- versity researching in microwave photonics, CDMA technique. He is currently an Associate Professor in the Information Science of the Nara Institute of Science and Technology (NAIST), with research interest on radio and optical communication and applications including MIMO-OFDM technologies, digital broadcasting, wireless position location. Dr. Higashino is a member of IEICE and IEEE. Minoru Okada received the B.E. degree from the University of Electro-Communications, Tokyo, Japan, in 1990. The M.E and Ph.D. degrees was received from Osaka University, Osaka, Japan, in 1992 and 1998, respectively, all in communications engineering. In 2000, he joined the Gradu- ate School of Information Science, Nara Institute of Science and Technology, Nara, Japan, as an As- sociate Professor and became a Professor. From 1993 to 2000, he was a Research Associate with Os- aka University. From 1999 to 2000, he was Visit- ing Research Fellow at University of Southampton, Southampton, U.K.. His research interest is wireless communications, including WLAN, WiMAX, CDMA, OFDM, and satellite communications. Dr. Okada is a member of the Institute of Television Engineers of Japan, the Institute of Electrical, In- formation, and Communication Engineers (IEICE) of Japan, and the Information Processing Society of Japan. He was the recipient of the Young Engineer Award from IEICE in 1999.

109