Performance Evaluation of Layer Division Multiplexing (LDM) combined with Time Frequency Slicing (TFS) Eduardo Garro, Jordi Joan Gim´enez, David G´omez-Barquero
Sung-Ik Park
iTEAM Research Institute, Universitat Polit`ecnica de Val`encia, Val`encia, Spain Email: {edgarcre,jorgigan,dagobar}@iteam.upv.es
Electronics and Telecommunications Research Institute, Daejeon, Republic of Korea Email:
[email protected]
Abstract The Advanced Television System Committee (ATSC) is currently developing the next-generation U.S. Digital Terrestrial Television (DTT) standard, known as ATSC 3.0. Two disruptive technologies for the physical layer are being evaluated, Layer Division Multiplexing (LDM) and Time Frequency Slicing (TFS). LDM consist in the transmission of a signal composed of two independent signals (layers) which are superimposed together at different power levels. These two layer can be configured with the desired robustness and capacity. LDM enables the efficient provision of services addressed to mobile and fixed reception in a more efficient way than the classical Frequency Division Multiplexing (FDM) and Time Division Multiplexing (TDM) since full bandwidth and transmission time are used in both layers. However, practical operation is restricted due to several implementation constraints such as the use of common parameters and transmitter blocks for both layers (e.g. a common time interleaver). The use of TFS allows for an improved frequency diversity by the distribution of the service data across multiple Radio Frequency (RF) channels instead of using a single RF channel. The paper investigates the potential gains provided by the increased frequency diversity with TFS in conjunction with LDM. Index Terms Layer Division Multiplexing (LDM), Time Frequency Slicing (TFS), ATSC 3.0, mobile broadcasting.
I. I NTRODUCTION The next-generation Digital Terrestrial Television (DTT) looks forward increasing spectral efficiency to cope with the scarcity of spectrum, derived from the digital dividend. The Advanced Television System Committee (ATSC) is currently developing the next-generation U.S. DTT standard, known as ATSC 3.0 [1]. Layer Division Multiplexing (LDM) [2] is being considered for adoption in addition to Time Division Multiplexing (TDM). In LDM, the transmitted signal consists of two independent signals superimposed together at different power levels, with an injection level (∆) relation between them, forming a multilayer signal. The independence of the two layers allows each one to be configured with its own modulation and coding rate (MODCOD) parameters. The most robust layer, intended for mobile reception, is referred to as Core Layer (CL). The Enhanced Layer (EL) can be used to provide high data rate services to fixed roof-top receivers. The injection level is a parameter which enables the distribution of the transmission power between the layers and defines the ratio between the power assigned to the CL and EL. The selection of a particular ∆ involves a penalty on the Signal-to-Noise Ratio (SNR) of both layers, due to their mutual interference. When ∆ is high, the CL power increases and the interference from the EL is reduced, thus, increasing the SNR of the CL. Conversely, when ∆ decreases high the CL becomes less robust (higher required SNR threshold). The higher SNR threshold of the CL allows direct demodulation. The EL is demodulated once the CL signal has been demodulated and cancelled from the input signal. Each LDM layer uses the full Radio Frequency (RF) bandwidth and transmission time what leads to a higher spectral efficiency in comparison with TDM and FDM, for one or both services (e.g. LDM can achieve the same bitrate as TDM/FDM with increased robustness, or the same robustness as TDM/FDM with increased bitrate) [3]. Several restrictions have been imposed for LDM implementation in ATSC 3.0 in order to limit the receiver complexity. The number of carriers (FFT size), Guard Interval (GI), Pilot Pattern (PP) and TI configuration are common for both layers. Thus, flexibility in configuration of the transmission is reduced in comparison with TDM where these parameters can be selected for the reception conditions of each service independently. For example the use of M-PLPs (multiple physical layer pipes) in DVB-T2 (Digital Video Broadcasting - Terrestrial 2nd generation) allow configuring a different MODCOD and TI per PLP. With T2-Lite, FFT size, GI and PP can also be configured independently. As a result, there is a trade-off between the optimum Core Layer (CL) and Enhanced Layer (EL) configuration for delivering mobile and fixed services, respectively: • High data rate service delivery to fixed reception (EL) should be done with a low carrier spacing (i.e. high FFT size) and a low-dense PP, to reduce GI and PP overheads. A low TI depth is recommended in order to reduce the demodulation latency.
CL MOD&COD CL Input BCH (IP, TS…)
BIL
LDPC
Mapper
LDM+TFS TRANSMITTER
FI
RF1
PP
IFFT
GI
Output RF1 D/A
Output RF2 EL Input (IP, TS…)
+
BCH
BIL
LDPC
Mapper
EL MOD&COD
Injection Level
Channel Estimation
Input RF1
CL+EL TI Framer/TFS Normalization RF2
LDM+TFS RECEIVER
Tuner Input RF2
CL DEMOD&DECOD H’(f)
A/D
GI
Time FFT & Freq. Synch.
Mapper-1 BIL-1 LDPC-1 BCH-1 CL Data
LDM buffer
FI-1 Framer-1 TI-1
CL REMOD Mapper BIL LDPC Mapper-1 BIL-1 LDPC-1 BCH-1 EL Data
H’(f) EL DEMOD&DECOD
Fig. 1. Transmitter and receiver implementation schemes for LDM + TFS operation. The block chains for a TFS-Mux of 2 RF channels are shown.
Mobile reception (CL) should involve a robust mode. A high carrier spacing (i.e. low FFT size), high-dense PP and a larger TI depth are recommended to cope with fast fading effects and to avoid Inter-Carrier Interference (ICI) caused by Doppler shift. Considering the common parameters with LDM, when the waveform is configured to favour fixed reception, the CL would have a penalty in mobility performance. However, the lack of an optimum transmission configuration for the CL can be partly compensated by a robust MODCOD, even with negative SNR threshold1 [5], [6]. Other techniques such as Time Frequency Slicing (TFS) [7] could be used together with LDM to enhance mobility performance. TFS has been analysed in the ATSC 3.0 standardization process although it was already introduced as an informative annex in the DVB-T2 standard, and was fully adopted in DVB-NGH (Next Generation Handheld). TFS distributes the data of each TV service across two or more RF channels instead of delivering each service over a single RF channel. Data recovery is performed at the receiver by means of frequency hopping. Though more than one tuner can be used for this purpose, single-tuner TFS may allow correct reception at the expense of several time constraints due to the necessity of guard times, where data cannot be transmitted,to allow frequency hopping operations. This approach enables to combine multiple RF channels into a high-capacity TFS-Multiplex (TFS-Mux), providing enhanced RF performance thanks to a frequency interleaving over multiple RF channels. The increased diversity is achieved by the extension of the time interleaving to all RF channels in the TFS-Mux. A time slot of each service is transmitted in each RF channel. The increased frequency diversity can be translated into a coverage gain against signal imbalances between RF channels and an increased robustness against interferences. With the proper error correction code and interleaving, it is even possible to recover from a lost RF signal provided that the other channels are good enough. Furthermore, the extra frequency diversity is particularly advantageous for pedestrian reception (slow time varying channels), where it may reduce the TI depth requirements. This paper investigated the potential advantage of TFS in connection with LDM and the limited TI configuration (common TI). Evaluation is performed by means of physical layer simulations considering the ATSC 3.0 standard. The paper also analyses the main implementation aspects of the joint configuration, assessing potential transmitter and receiver changes that might be taken into account. The paper is organized as follows: Section II evaluates the main implementation aspects of the joint configuration of both technologies; Section III studies the LDM CL and EL performance when TFS is added to the transmission. Finally, conclusions are summarized in Section V. •
II. LDM + TFS I MPLEMENTATION A SPECTS Different options for the implementation of TFS can be envisaged (i.e. implementation of TFS in both layers or only in one layer). However, according to the transmitter implementation scheme for LDM, the implementation of TFS for both layers is 1 In
ATSC 3.0, code rates 2/15, 3/15 and 4/15 code rates provide a negative SNR threshold for AWGN channel using QPSK modulation [4].
the solution with the lowest hardware complexity. Fig. 1 presents the ATSC 3.0 transmitter and receiver physical layer block diagrams when LDM and TFS are implemented. The transmitter scheme shows a different BICM chain per layer, but common TI, PP, FFT and GI. A TFS-Mux with 2 RF channels is considered. Regarding LDM implementation, the required cancellation process for the demodulation of the EL is the main aspect to bear in mind. This process should be faster than the time of arrival of new cells but it is only performed at those receivers oriented to demodulate the fixed service. Nonetheless, the number of iterations for the correct demodulation of the CL in such scenarios is quite small because the SNR may be higher than the required, so the cancellation process might consume less time than the arrival of new cells. Moreover, to obtain the cells of the EL, those on which the cancellation process is performed must be stored in the LDM buffer at the receiver while the cells of the CL are reconstructed. The implementation of TFS involves the necessity of extending time interleaving across the RF channels in the TFS-Mux. The limited amount of memory at the receiver for time de-interleaving, limits the peak bit rate per service. Furthermore, single-tuner TFS requires the reservation of time gaps where no data can be received. This overhead also limits the peak data rate per service. Single-tuner TFS requires a proper scheduling at transmitter (to be accomplished by the Framer) in order to guarantee the sequential reception of the service data with frequency hopping. The scheduler is in charge of ensuring a proper time margin between data slots to enable tuning and channel estimation operations at the receiver. After an overall evaluation of the constraints related to both techniques separately, the conclusion is that the combined operation of LDM and TFS does not involve, in general, additional implementation requirements for joint operation to those intrinsic to TFS and LDM if TFS is applied to the two LDM layers equally. However, to obtain the desired frequency diversity with TFS, the cells of the FEC word should be evenly distributed across all the RF channels of the TFS-Mux. This process is accomplished by the TI which works on a cell level. The number of cells in each layer depends on the FEC word size and the modulation order. Thus, the number of cells per layer can be different. Since the cells of both layers are added together before the TI, it is not guaranteed that both cells are evenly distributed across the TFS-Mux. Thus, the achieved frequency diversity may be compromised for one of the two layers. For instance, for the same FEC word size, a configuration with QPSK at the CL and 256QAM at the EL makes that several blocks of cells of the EL fit into one block of the CL. With a common TI performing interleaving on a cell level, it is possible that whereas those cells of the CL are evenly distributed across all RF channels, those from the EL only lie in one of the RF channels, causing a performance loss for the EL if this RF channel is degraded. III. P ERFORMANCE E VALUATION M ETHODOLOGY The performance of LDM and TFS is studied by means of physical layer simulations for mobile and fixed reception. Regarding TFS, the performance of the transmission across the TFS-Mux is compared with respect to the transmission over the RF channel in the worst SNR condition, the one that would be limiting the reception of the complete set of services in a classical non-TFS transmission. For fixed-rooftop reception, the worst channel is, in general, the one with the highest frequency [8]. In addition, for mobile reception, the degradation due to Doppler effect has to be taken into account. Intra-frame TFS is evaluated. The FEC words of a service are time interleaved. Thus, cells belonging to the same service are spread in a frame which length is set by the TI duration. The effect of transmitting slots of data across different RF channels is emulated taking into account the effect of the SNR imbalances between RF channels. This effect is included in the simulation for mobile and fixed reception, according to the model presented in [8]. With this, the imbalance between each pair of frequencies is determined by a statistical model developed from field measurements. Note that in these simulations only the average values of the imbalances are taken into account. In addition, for mobile reception, the FEC words are filtered by independent TU6 channel realizations, thus considering that the channels experienced by each frequency are mutually independent. The TU6 taps are generated taking into account the Doppler shift, using Eq. (1). fd (Hz) = v (m/s) ·
fc (Hz) c (m/s)
(1)
where fd is the Doppler shift, v is the receiver speed, fc is the carrier frequency of the RF channel and c is speed of light. The transmission and channel parameters assumed for the simulations are following described: • Results are based on ideal channel estimation. • 6 MHz channel bandwidth is used. • 16k FFT, 1/16 GI fraction and a block type TI of 100 ms depth are the waveform parameters shared by both layers. • The channel models considered are TU6 channel for mobile reception and Rice (F1 DVB-T2) channel for fixed reception. • The transmission mode for the mobile service adopted is QPSK 4/15, (2,7 Mbps). • The transmission modes for the fixed service assumed are 64NUQAM 10/15, with a data rate of 20,5 Mbps, and 256NUQAM 11/15, with a data rate of 30 Mbps. • An injection level of 4 dB is considered, which distributes the total transmission power according to 70% for the CL, and 30% for the EL, approximately.
10 −1
TFS
10 −2 BER
non−TFS
10 −3
10 −4
0
5
10
15
SNR (dB)
Fig. 2. Performance of a LDM core layer QPSK 4/15 with and without TFS for the four RF channels. Injection Level = 4 dB. TU6 channel 3 km/h. TABLE I D OPPLER SHIFT fd (H Z ) PER RF CHANNEL
Speed
•
RF1 503 MHz
RF2 533 MHz
RF3 563 MHz
RF 4 593 MHz
3 km/h
1,4
1,48
1,56
1,65
135 km/h
62,88
66.63
70,38
74,13
A TFS-Mux of 4 RF channels with a total frequency separation of 90 MHz among them is assigned (503, 533, 563 and 593 MHz). IV. R ESULTS AND D ISCUSSIONS
The results show the performance of LDM with and without TFS. A first study compares the performance of LDM mobile service due to the Doppler effect. Secondly, the mobile and fixed services performance with the inclusion of RF channel imbalances are also evaluated. A. LDM CL + TFS with Doppler Table I presents the Doppler shifts of the RF channels considered for the TFS-Mux. 3 km/h (pedestrian reception) and 135 km/h (vehicular reception) are set. 1) Pedestrian reception (3 km/h): Fig. 2 depicts the performance of 4 different signals which are independently transmitted in each RF channel of the TFS-Mux. A slow time-varying channel is considered (low Doppler shift). Differences in performance due to different Doppler shifts are minimum. In addition, the configured TI duration (100 ms) is not enough to provide low SNR values. The use of TFS offers a diversity increase which can cope with the lack of insufficient TI duration [9], obtaining a gain of about 7 dB. 2) Vehicular reception (135 km/h): Fig. 3 depicts the performance for the same configuration but a high-speed vehicular reception condition is considered. The performance gain achievable with TFS is almost non-existent since the TI duration already achieves enough time diversity. The low coherence time of the high-speed channel does not requires additional diversity and TFS does not offer an important improvement when only considering Doppler differences. In summary, two main important conclusions can be extracted from both scenarios: • The performance gain obtained with TFS in high speed reception is lower than for low speed reception (0,1 dB gain in vehicular conditions, whereas in low speed it was 7 dB). • When TFS is used, a higher TI duration is not essential for a better performance on low speed reception, since this case already achieves a high degree of diversity.
10 −1
TFS fd = 74,13 fd = 70,38 fd = 66,63 fd = 62,88
BER
10 −2
Hz Hz Hz Hz
10 −3
10 −4
1.5
2
2.5
3
SNR (dB)
Fig. 3. Performance of a LDM core layer QPSK 4/15 with and without TFS for the four RF channels. Injection Level = 4 dB. TU6 channel 135 km/h.
10 −1
RF1 RF2 TFS RF3 RF4 10 −2 BER
EL 64QAM 10/15
CL QPSK 4/15 10 −3
EL 256QAM 11/15 10 −4
0
5
10
15
20
25
SNR (dB)
Fig. 4. Performance for CL and EL with and without TFS for the four RF channels considering the RF imbalances. CL TU6 Channel (135 km/h), EL Rice Channel. Dotted lines: EL with same cells in the FEC words. Dashed lines: EL with different cells in the FEC words
B. LDM + TFS with RF channel imbalances The TFS-Mux constituted by the same 4 RF channels produces the RF imbalances exposed in Table II according to the aforementioned model [8]. Fig. 4 illustrates the mobile and fixed reception performance when TFS is used. For mobile reception, the TFS performance gain with respect to the worst RF channel is 1,5 dB. For fixed reception, the performance varies depending on the CL-EL FEC word size ratio. Table III shows the FEC word cells ratio (EL cell blocks that fit in one CL cell block) between layers for the LDM configurations assumed. As it was explained in Section II, when the number of cells of the FEC words is the same (QPSK-256NUQAM), the EL will benefit from the same diversity as the CL. With TFS this is translated to an equal distribution of the cells among the RF channels in the TFS-Mux, so the performance gain is around 1,5 dB again. On the other hand, if they were not the same (QPSK-64NUQAM), the expected TFS gain for the EL is not reached. Remind that these simulations are based on a conventional block type TI. V. C ONCLUSIONS AND F UTURE W ORK In this paper, the potential advantages of the introduction of TFS in conjunction with LDM are presented.
TABLE II RF CHANNEL IMBALANCES (dB) RESPECT TO THE LOWEST FREQUENCY RF1 503 MHz
RF2 533 MHz
RF3 563 MHz
RF 4 593 MHz
0
-1,1
-2,15
-3,15
TABLE III FEC WORD SIZES LDM Layer
FEC (bits)
MOD (bpc)
FEC (cells)
CL EL
64800
QPSK (2)
32400
64800
64QAM (6)
10800
CL EL
16200
QPSK (2)
8100
64800
256QAM (8)
8100
CL:EL ratio 3:1 1:1
The main advantage, for the use cases studied, is obtained in pedestrian reception where TFS offers a diversity increase which can cope with the lack of enough time interleaving duration in slow time-varying channels. The improved performance by TFS increases the probability of receiving the complete set of services in the TFS-Mux. However, when using a common time interleaver approach working on a cell level and with a different number of cells per layer, it may occur that the desired performance with TFS is not achieved equally in each layer. The number of cells depends on the FEC word size and the modulation order, which may be different for each layer. The gain obtained thanks to TFS implementation requires an equal distribution of cells across the RF channels of the TFS-Mux, which mainly depends on the TI configuration used. When the LDM configuration assumes a common TI approach, and the FEC word cells layers ratio is not 1:1, the performance of the EL may be compromised. The simulations were made assuming a block type TI. Further investigations will be made with different TI configurations, such as convolutional type, twisted block type, and hybrid combinations, in order to find out the best LDM and TFS implementation. The potential TFS gain with LDM with other TI durations, Doppler shifts (medium and very high speeds), and different number of RF channels in the TFS-Mux will be investigated. Finally, real channel estimation is required for assesing the impact of different FFT sizes and PP. VI. ACKNOWLEDGMENT This work was partially supported by the ICT R&D program of MSIP/IITP. [R0101-15-294, Development of Service and Transmission Technology for Convergent Realistic Broadcast] R EFERENCES [1] D. G´omez-Barquero and W. Caldwell, “Broadcast Television Spectrum Incentive Auctions in the U.S.: Trends, Challenges and Opportunities,” Communications Magazine, IEEE, vol. 53, no. 7, July 2015. [2] Y. Wu, B. Rong, K. Salehian, and G. Gagnon, “Cloud Transmission: A New Spectrum-Reuse Friendly Digital Terrestrial Broadcasting Transmission System,” Broadcasting, IEEE Transactions on, vol. 58, no. 3, pp. 329–337, Sept 2012. [3] J. Montalb´an et al., “Cloud Transmission: System Performance and Application Scenarios,” Broadcasting, IEEE Transactions on, vol. 60, no. 2, pp. 170–184, June 2014. [4] L. Michael and D. G´omez-Barquero, “Modulation and Coding for ATSC 3.0,” in Broadband Multimedia Systems and Broadcasting (BMSB), 2015 IEEE International Symposium on, June 2015. [5] S. I. Park, H. M. Kim, Y. Wu, and J. Kim, “A Newly Designed Quarter-Rate QC-LDPC Code for the Cloud Transmission System,” Broadcasting, IEEE Transactions on, vol. 59, no. 1, pp. 155–159, March 2013. [6] S. I. Park, Y. Wu, H. M. Kim, N. Hur, and J. Kim, “Raptor-Like Rate Compatible LDPC Codes and their Puncturing Performance for the Cloud Transmission System,” Broadcasting, IEEE Transactions on, vol. 60, no. 2, pp. 239–245, June 2014. [7] J. J. Gim´enez, E. Stare, S. Bergsmark, and D. G´omez-Barquero, “Time Frequency Slicing for Future Digital Terrestrial Broadcasting Networks,” Broadcasting, IEEE Transactions on, vol. 60, no. 2, pp. 227–238, June 2014. [8] J. J. Gim´enez, D. Goz´alvez, D. G´omez-Barquero, and N. Cardona, “Statistical Model of Signal Strength Imbalance Between RF Channels in DTT Network,” Electronics Letters, vol. 48, no. 12, pp. 731–732, 2012. [9] D. Goz´alvez, D. G´omez-Barquero, I. Eizmendi, G. Berj´on-Eriz, and M. V´elez, “DVB-T2 for Mobile and Mobile DVB-T2 (T2-Lite),” in Next Generation Mobile Broadcasting, D. G´omez-Barquero, Ed. Boca Raton, FL, USA: CRC Press, 2013, pp. 151–183.
