features of scalable video coding (SVC) into current OFDM based ... broadcast transmission schemes in cellular MIMO networks are presented. Important ...
Scalable Video Transmission in Multiantenna Broadcast Systems Stephan Jaeckel and Volker Jungnickel Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institut Einsteinufer 37, 10587 Berlin, Germany {stephan.jaeckel ; volker.jungnickel}@hhi.fraunhofer.de
Abstract—In this paper, we make an attempt to integrate the features of scalable video coding (SVC) into current OFDM based mobile communications systems for improving the broadcast data rates and to provide a better coverage. Several physical- and MAC-layer concepts that qualify for the use in combination with SVC are evaluated. Space time coding in combination with single carrier frequency division multiplexing (SC-FDM) based on DFT precoding are promising.
I. I NTRODUCTION Typically, in mobile networks, transmission errors of a connection depend on the actual quality of the reception conditions. Current standards like UMTS, HSPA and LTE try to take maximum advantage of the available bandwidth by adjusting the transmission rate to the actual channel quality by means of user feedback. This feedback is, however, not available in broadcast scenarios. Techniques such as multipleinput multiple-output (MIMO), space-time codes (STCs) or spreading will help to increase the data rate, coverage and reliability. At the same time, innovations in video coding such as the advanced- and scalable video codes (AVC/SVC) can take maximum advantage of the provided datarate. Both parts, the mobile communication systems and the source coding schemes may be combined in a way that allows adaptive, flexible and reliable content distribution to many users at once. SVC is an extension of the H.264/AVC video compression standard [1]. It allows efficient scalability of temporal, spatial and quality resolution of a video signal. Scalability is achieved when parts of the data stream can be removed in a way that the resulting stream composes another valid video signal. This video stream represents the source content with a reduced reconstruction quality. Layer 1 - 160 kBit/s Layer 2 - 200 kBit/s Layer 3 - 200 kBit/s Increase size or quality Increase size or quality
Layer 1 - 160 kBit/s
Layer 1,2 - 360 kBit/s
Layer 1,2,3 - 560 kBit/s
Figure 1. Principle of SVC: A video stream is divided into several layers with different size. The receiver can combine these layers in order to increase the framerate, the size and/or the quality of the resulting video. Layer 1 is the most important substream - it is needed for each quality level.
Scalability might become an important feature for video transmission over mobile networks. The source content has to be encoded only once with the highest resolution and bitrate. Representations of lower quality can be obtained by partial decoding (see [2]). This is important especially for mobile clients with restricted resources (computing and battery power, display resolution, etc.). At the same time, content scalability together with an unequal error protection and power allocation allows rate and quality adaption not at the transmitter, but at the receiver. The paper is organized as follows: In section II several broadcast transmission schemes in cellular MIMO networks are presented. Important transmission techniques are assessed in section III followed by numerical results in Section IV. According to [3], a suitable 3-layer SVC may consist of a base layer with 160 kBit/s and the two enhancement layers having 200 KBit/s, each. This example is also used here to demonstrate the effects of MIMO-SVC. II. V IDEO B ROADCAST C ONCEPTS IN MIMO N ETWORKS Currently deployed terrestrial broadcasting concepts are mainly based on DVB-T [4] which uses orthogonal frequency division multiplexing (OFDM) in a cellular single frequency network (SFN). This allows the reuse of frequencies in each cell but lacks flexibility when it comes to localized content distribution (i.e. personalized advertisements). In recent years there have been several developments in the field of mobile communications which provide higher spectral efficiencies and better coverage. Important developments are MIMO [5], STCs [6] and base station (BS) cooperation [7]. Results in [8] show that additional capacity gains are possible even in rank-reduced line of sight (LOS) MIMO channels due to polarization multiplexing. On the other hand, video coding technologies also improve substantially. Both concepts use layered transmission schemes. MIMO systems provide multiple layers in the space-time-frequency domain whereas SVC splits the video stream in multiple layers as well. Bringing these two worlds together implies the need of new concepts for video transmission over wireless channels. High definition video needs transmission rates of up to several Mbit/s but high-rate wireless channels are unreliable especially when no channel feedback is available. The following techniques may be used to satisfy these demands.
a) Single Frequency Networks: Coherent SFNs simultaneously send the same signal over the same frequency channel. This allows a higher number of radio and TV programs compared to traditional multi frequency networks (MFNs). At the receiver, SFN transmission can be seen as a special form of multipath propagation. As long as the artificial multipath components are covered by the OFDM guard interval, they increases the overall received power. Due to the independent paths, the diversity order is increased as well. Terrestrial digital TV broadcasting systems widely use OFDM-SFNs. They are able to take advantage of frequency diversity and are easy to implement since the underlying link concept is OFDM. Nevertheless, these systems do not include gains in the spatial domain which are provided by the uncorrelated paths from several BSs to the receiver. The DVB-T standard also does not consider multiple antennas or spatial multiplexing and diversity encoding schemes. Since SFNs can be seen as a special form of multipath reception, one can model them as Rayleigh fading channels.
BS1
BS3
BS2
Channel Estimator Combiner Detektor
Input Data
Figure 3. Alamouti Encoding: The two complex transmission signals x1 and x2 are encoded with the Alamouti scheme and simultaneously transmitted from the two BSs or from two antennas of the same BS. The known channel coefficients h1 and h2 can be used to combine the signals at the receiver.
the way to the terminal but when using channel estimation together with MIMO equalization, both streams can be separated again. Adjacent BSs can either transmit the same signals in SFN mode or one implements a 6x2 STC for joint transmission in one cell. Both schemes will be able to exploit frequency diversity and spatial diversity simultaneously. d) Macrodiversity: An even higher degree of diversity can be achieved by implementing a 3x1 or 6x1 STC which covers several adjacent BSs (see fig. 4). The mobile terminal can use the knowledge of the channel conditions to coherently combine the power of the independent propagation paths. This leads to a better signal to noise ratio (SNR) and could enable the network provider to use more advanced modulation schemes such as 64- or 256-QAM to increase the throughput. 1500 6 1000
b) Polarization Diversity: Another possibility to improve the diversity gain lies in the use of two or more antennas at the mobile terminal. Results from [8] show that there are at least two degrees of freedom even in a rank reduced LOS channel if cross-polarized antennas are used at both, the transmitter and the receiver. This comes from the fact that cross-polarized channels are uncorrelated to a certain degree. If the receiver is able to separate signals arriving from the horizontal and from the vertical plane, it can take advantage from maximum ratio combining (MRC) which sums up the two signals in a coherent way. Another option is to use a STC such as the Alamouti scheme [9] to encode the signals from each transmitter. This allows the coherent combination of two signal paths even if only one receive antenna is available. c) Polarization Multiplex: One has at least two (crosspolarized) antennas per BS. If the mobile terminal also has two cross-polarized antennas, multiplexing of two data streams in the spatial domain becomes possible. Each BS then transmits one stream on the H-plane and another one on the V-plane. Multipath propagation may change the polarization vectors on
4 5 21
y-position [m]
Figure 2. Concept of a SFN: All BSs transmit the same signal. The center region is covered by multiple signals that can be seen as additional multipath components.
9 19
500
3x1 Space Time Coding
7
20
8
Tx 1 Tx 2 Tx 3
3 1
0 2 18 -500
12 16
10
17
11 15 13
-1000 14 -1500 -1500
-1000
-500
0 500 x-position [m]
1000
1500
Figure 4. Macrodiversity network based on the SCME channel model: A 3x1 STC encodes the streams of three adjacent sectors.
III. T RANSMISSION T ECHNIQUES a) DFT Spreading: Spreading provides a promising method for enhancing the performance of different bit streams. The concept is known from code-division multiple access (CDMA) where a low-rate modulation signal is multiplied with a high-rate binary code sequence. The output waveform covers a larger bandwidth and is therefore more robust against transmission errors in the presence of multipath fading. When the codes can be considered as orthogonal, reconstructing the input symbol stream is a simple process of multiplying the received signal with the same sequence and summation over time. With the introduction of more powerful processors, the
Layer 1 Data
Demux
SC-FDM
Mod
...
IFFT IFFT
CP
Demux
STC Channel Estimation
MMSE Weight Calculation
LO
LO
STC
FDE / DFT-Despreading
SFN
CP
Mod Mod
Receiver
LO
... Layer S Data
STC / SFN Transmission
...
A/D
FFT
A/D
FFT
LO
Demod / Demux Layer 1 ... Layer S Data Data
Mod
Figure 5.
Signal processing chain
use of non-binary sequences in realtime becomes possible. Suitable candidates for such spreading codes can be derived from the discrete Fourier transform (DFT), hence the name: DFT spreading. It is done by an unitary DFT of the data symbol vector: Ndf t 1 X Wn,m · d~m ~xn = √ N m=1
(1)
where Wn,m are the scalarcoefficients of the DFT spreading sequences Wn,m = exp −2Πj (n−1)(m−1) , Ndf t is the N number of√the input symbols of the DFT and the power scaling factor 1/ N is for holding the transmit power constraint of the OFDM system. The lengths of the vectors d~m and ~xn equal the number of parallel transmission streams in the spatial domain (e.g. the number of transmit antennas in a MIMO multiplexing system). After transmission through the wireless MIMO-OFDM channel, each of the N transmitted symbols suffers from frequency-selective fading which can be expressed with the (narrowband) MIMO transmission equation [5] on each of the OFDM carriers: ~yn = Hn · ~xn + ~vn
(2)
~yn is the received symbol vector, Hn is the (nT x × nRx ) channel matrix and ~vn is the additive noise. Combined minimum mean square error (MMSE) equalization [10] and DFT despreading leads to an estimated data symbol at the receiver: N −1 1 X ∗ ntx ˆ d~m = √ · Wn,m · Intx + HH H · HH yn n n n ·~ SNR N n=1 (3) ∗ where Wn,m denotes the complex conjugate of Wn,m . It has been proven very recently that frequency domain equalization and subsequent IDFT despreading has the same performance as the RAKE multiuser detector based on DFT spreading sequences [11]. It realizes the maximum achievable diversity by coherently combining the power of several multipath components. This is especially helpful when SFN transmission is applied in a wireless broadcast network to increases the number of multipath components.
b) Single Carrier Multiplexing: A mobile communication system (i.e. 3G LTE) may feature single user and broadcast traffic simultaneously. Both traffic types can share
the same frequency range in a time-division multiple access (TDMA) manner. The duration of the timeslots can be adjusted depending on the amount of traffic available for broadcasting (i.e. number of programs) or point-to-point communication. During one timeslot, the broadcast system has full channel access and can therefore use all diversity-achieving techniques such as special pilot tones, space-time and channel codes, extended guard intervals, modulation schemes etc. SC-FDM could be used to multiplex multiple layers of a video stream into one single carrier waveform [12]. A generalized SC-FDM transmission chain is shown in figure 5. The layers of the scalable video stream are separated by assigning different modulation and coding schemes to each layer. Additional power scaling is applied in a way that higher layers have a higher QAM modulation order and get less power at the same time. In this way, we do not only enhance the dynamics range for the unequal error protection, we also reduce the overall peak to average power ratio (PAPR) in the compound single-carrier waveform which mainly comes from the higher order modulation schemes such as 16- or 64QAM. c) Unequal error protection: Since SVC is a hierarchical coding scheme, it is only possible to decode layer 2 when layer 1 is decoded correctly and layer 3 can only be decoded when layer 1 and 2 are correct. Although it is impossible in a broadcast system to adjust the transmission scheme to the channel, adaption to the source is still an option. Assigning more resources (i.e by using more bandwidth, QPSK instead of 16QAM modulation etc.) is crucial to increase the performance of layer 1. At the same time, the performance of the highest stream is decreased accordingly. For calculating the modulation and coding parameters, one has to consider the following constraints: •
Number of available resources: If N resources (i.e. inputs of the DFT coder) are available, one can create N complex modulation symbols. The number of available resources for each layer is limited by: S X
Ns ≤ N
(4)
s=1
Where S is the number of layers and Ns is the number of resources assigned to the sth layer.
•
•
•
Desired bit error rate (BER) separation ∆BER between s streams: Two SVC layers have to achieve different performance results. The SNR offset between these streams is due to the modulation format (16QAM achieves less performance then QPSK) and the transmit power Ps assigned to each stream. OFDM sum power constraint: The transmission power of a mobile communication system is limited. When scaling the power of a stream to adjust the performance, one must not violate the sum power constraint. Therefore ! nT x Ns X S X 1 X ∗ xs,t,n · xs,t,n ≤ PT x (5) E Ns n=1 t=1 s=1 The PAPR: Another constraint arises from the PAPR of the time domain signal. The inverse fast Fourier transform (IFFT) generates a NT-point discrete signal whose PAPR [13] is defined as: PAPR =
Peak power maxk=1...N ·T |x(k)|2 = PN ·T 1 2 Average power k=1 |x(k)|
(6)
N ·T
where N is the number of carriers in the OFDM system and T is the oversampling factor used for the PAPR calculation. Two parameters have to be calculated for each of the S layers: The modulation format and the power scaling coefficient. Comparing all possible combinations of the S streams and M modulation formats results in an asymptotic complexity of O(M S ). The example stated in section I contains three layers of 160, 200 and 200 kBit/s, respectively. Lets assume for simplicity that for this example 660 bits (160 for Layer 1, 200 for layer 2 and 3) have to be transmitted in the next TTI of a 256 carrier system. A minimum layer separation of 8 dB should be guaranteed to capture most of the dynamics in the channel and the power scaling values should be as small as possible in order to keep the PAPR small. Four modulation formats are allowed: BPSK, QPSK, 16QAM and 64QAM. In an uncoded OFDM system, QPSK needs a 3 dB better SNR compared to BPSK to be decoded with the same BER. 16QAM needs additional 6 dB and 64 QAM another 5.2 dB. When comparing all 43 = 64 combinations one finds that BPSK for layer 1, 16QAM for layer 2 and 64QAM for layer 3 fits best. For the desired 8 dB separation, one has to scale all symbols of layer 2 by an additional factor of 1.25 and layer 3 by 0.66. Simulation results for this example are presented in fig. 6 where a Rayleigh fading channel with L = 3 independent
SC−FDM performance for 3 SVC layers : 1x1, N=256, L=3, DFT−MMSE
0
10
Layer 1 (BPSK) Layer 2 (16QAM * 1.25) Layer 3 (64QAM * 0.66)
−1
uncoded bit error rate
Incoming SVC data rate per stream: Each SVC stream has a certain amount of bits bs to be processed during the following transmission time interval (TTI). The modulation format (BPSK, QPSK, 16QAM etc.) and the number of assigned resources must be set accordingly. Additionally, the forward error correction (FEC) may also need a minimum number of bits per packet data unit (PDU) to work efficiently.
10
−2
10
−3
10
−4
10
0
5
10
15
20
25
30
35
40
45
50
SNR [dB]
Figure 6. Performance of a simulated 3-Layer SC-FDM transmission system: The figure shows the performance for each of the three layers. The overall transmission rate is 660 bit per channel use in a 256 carrier OFDM system. The underlying channel model is Rayleigh with L = 3 taps and a DFT-spread MMSE equalizer.
fading taps was used. At an uncoded BER of 2 · 10−4 , the performance curves of the three data streams are separated by approx. 12.2 dB (Layer 1 and 2) and 11.5 dB (Layer 2 and 3). That means that the coding covers almost 25 dB of SNR. A sudden decrease of signal strength would not interrupt the video stream but merely lower its quality. The additional separation comes from the DFT precoding which also captures the multipath diversity due to its relationship with the RAKE receiver. In addition, the PAPR is reduced by the single carrier transmission. Fig. 7 shows the complementary cumulative distribution function (CCDF) of the PAPR for the proposed 3-layer example (black line) and for an uncoded 256 carrier OFDM system with QPSK modulation (dashed blue line). As suggested in [13], an oversampling factor of T = 4 was applied to calculate these values. The 0.1 percent PAPR of the SVC example was 9.5 dB while the OFDM-QPSK reference case needs as much as 11.3 dB. The proposed concept does not only take advantage of the multipath diversity, it also shows better PAPR performance due the underlying SC-FDM concepts. IV. S YSTEM L EVEL S IMULATION R ESULTS In order to get realistic user channels in a multicell multiple access scenario, we performed system level simulations based on the 3GPP Spatial Channel Model (SCM, see [14]). In SC−FDM PAPR for 3 SVC layers : 1x1, N=256, L=3, DFT−MMSE
0
10
Prob. (PAPR > PAPR0)
•
−1
10
−2
10
−3
10
−4
10
4
3−layer Example OFDM QPSK 5
6
7
8
9
10
11
12
PAPR [dB]
Figure 7.
PAPR of the 3-Layer SC-FDM transmission system
13
addition to the original version, several extensions such as support for polarized antennas [15], tilt and scenario mixing [16] have been implemented to get realistic statistics. Table I gives an overview of the simulations settings: Table I S YSTEM L EVEL S IMULATION PARAMETERS Parameter Value Channel model SCME-C (Urban Macro) Center frequency 2.0 GHz Bandwidth 30.72 MHz No. of Carriers 512 No. of BSs 7 No. of sectors per BS 3 BS height 32 m MS height 2m BS distance 1000 m Antenna Kathrein 14dBi, dual polarized Tilt 3.7Â◦ Transmission Power 30 dBm Receiver Sensitivity -95 dBm
The SCME simulation environment is set up to randomly place 10000 users in between the three BSs illustrated in Fig. 8. A 2x2 polarized channel realization H is calculated for each user. The SCM channel coefficients are then used to perform a BER simulation for the example in fig. 1. The simulated transmission chain uses 512 OFDM carriers. 320 resources are assigned to layer 1 with BPSK modulation. Layer 2 gets 100 resources with 16QAM and a power multiplier of 1.25. The last layer gets 68 resources with a multiplier of 0.66. The channel H is assumed to be constant for at least 50 realizations whereas the noise is changing in each run. The noise floor is set to -95 dBm and the transmission power is 30dBm. This low power ensures that none of the evaluated transmission schemes is able to achieve 100% coverage to allow a fair comparison of the results. All in all, 66000 random bits are transmitted. From these bits, a BER calculation is performed for all SVC layers. If the resulting BER is pe < 2 · 10−4 , the layer is assumed to be error-free. No additional channel code is used. Results are illustrated in figure 9.
Fig. 9 shows the relative number of users that can be reached with a certain SVC layer and transmission scheme. The two considered cases are zero forcing (ZF)-OFDM (dark grey bars) and MMSE-DFT-SC-FDM (light grey bars). The graph can be interpreted in the following ways: 1) MMSE-DFT-SCFDM always outperforms ZFOFDM. Nevertheless, these gains decrease when a lot of spatial diversity is available (see cases 3, 7, 8 and 10). This comes from the fact that ZF-OFDM is not able to achieve additional frequency diversity. When increasing the antenna diversity (MRC, STC), the performance of ZF-OFDM also increases. 2) SFN transmission increases the performance of all transmission schemes. In SFN mode, the received power of the cell edge users is rising due to the additional channel filter taps. The DFT-MMSE equalizer is able to combine these taps constructively whereas ZFOFDM only uses the array gain but not the increased diversity. 3) Macrodiversity achieves no substantial gains. Cases 9 and 10 were precoded with an orthogonal 3x1 STC from [6]. Case 2 (SISO-SFN) and case 9 (3x1 STC) show the same antenna constellation of three transmitters and one receiver. The only difference is that in case 9, diversity is combined in the spatial domain whereas case 2 uses the additional transmitters for frequency diversity. The same constellation holds for case 10 (3x2 STC) and 4 (SIMO-SFN-MRC). The macrodiversity scheme outperforms the SFN scheme only for the ZF-OFDM system. This comes, again, from the fact that ZF is not aware of frequency diversity. When including frequency diversity due to DFT precoding and MMSE equalization, macrodiversity only enhances the coverage by 1-4%. If one considers that a 3x1 STC only achieves rate 43 and that there exists no orthogonal full rate STC for more then two transmit antennas, one must conclude that macrodiversity is not necessary.
1500 6 1000
4 5
y-position [m]
21
9 19
500
7
20
8 3 1
0 2 18 -500
12 16
10
17
11 15 13
-1000 14 -1500 -1500
-1000
-500
0 500 x-position [m]
1000
1500
Figure 8. BS arrangements for the system level evaluations: User channels in the dark shaded areas are evaluated.
According to these results, two transmission schemes may be used in a broadcast system: Alamouti precoded SFN transmission with two transmitters per BS and two receivers (case 8) achieves the best coverage results. With the reduced transmission power of 1 Watt, 83% of the users are able to decode the first SVC layer and 47% are able to decode the entire stream. When switching to MIMO multiplexing with SFN (case 6), two streams can be transmitted at the same time. This increases the data rate by a factor of two but the coverage decreases substantially. With 46 dB (40W) transmission power, we observe coverage results of 99.8 | 95 | 79% for the Alamouti-SFN mode and 94 | 68 | 49% for the spatial multiplexing mode. One can expect, that these values rise even further when the inter-site distance is decreased. The MIMO multiplexing mode might therefore be an option for small cells.
SISO SFN
MRC
MIMO Mux SFN
Alamouti
Alamouti SFN 83
60 55
37 27
22
3
2
4 5
45
41
41
14 9
6
32
8
7
20 14 8
47 45
37
57
52
35
31
30 18
Layer 3 - 200 kBit/s Layer 2 - 200 kBit/s Layer 1 - 160 kBit/s
25
10
9
transmission scheme / layer
MMSE, SCFDM, DFT precoding ZF, OFDM
Sector Coverage with different Transmission Schemes at 30dBm Transmission Power
Figure 10.
Pilot insertion
STC coded DFT Data
DFT Data
Space-time Coder
DFT Spreading
Mapped Symbol Stream
SC-FDM Mapper
IFFT IFFT SFN distribution
Overview of the System Concept
H.264 / SVC Dencoder
CRC based Rate Adaption
H
scaled SVC Video Stream
Mobile Channel
FFT
Binary Data Stream
Power scaled Symbol Stream
sampled Data
Error detection and correction
DFT Data
FFT
Binary Data Stream
MMSE Equalizer
Time D omain Transm ission S ignals
IQ Demodulator
[1] ITU-T Recommendation H.264 and ISO/IEC 14496-10 (MPEG-4 AVC), Advanced video coding for generic audiovisual services - Version 8 (including SVC extension), ITU-T and ISO/IEC JTC 1, 07 2007. [2] H. Schwarz, D. Marpe, and T. Wiegand, “Overview of the scalable video coding extension of the H.264/AVC standard,” IEEE Trans. Circuits Syst. Video Technol., vol. 17, no. 9, pp. 1103–1120, 2007. [3] T. Schierl, T. Stockhammer, and T. Wiegand, “Mobile video transmission using scalable video coding,” IEEE Trans. Circuits Syst. Video Technol., vol. 17, no. 9, pp. 1204–1217, 2007.
Video Source
Pilot Tones
Inverse Power Multiplication
R EFERENCES
Channel Estimator Channel Matrix
DFT Despreading
Many thanks to Thomas Schierl, Cornelius Hellge, Thomas Wiegand (all with HHI) and Thomas Haustein (NSN) for the stimulating discussions and the German Ministry of Education and Research (BMBF) for the financial support in the project 3GeT (MIMO-SVC).
Channel Coding FEC
ACKNOWLEDGMENTS
Raw Video Stream Symbol Stream
H.264 / SVC Encoder SVC Stream
Power Multiplication
In this paper, we have explored potential system concepts for multimedia multicasting services over next generation cellular networks, such as WiMAX and 3G LTE. We have shown by numerical simulations that single frequency networking, space-time coding, DFT spreading and SC-FDM mapping can be combined in order to achieve a high link reliability and to maximize the throughput in wireless multimedia broadcast scenarios where no or little channel feedback is available. Our results indicate that: • One should use DFT spreading in the downlink broadcast mode since it takes optimal advantage of the multipath diversity and thus, realizes single frequency networking gains. • The DFT size should be matched to the full system bandwidth, and the inputs of the DFT should be handled as primary resources. • The different layers of the scalable video stream should be separated by different modulation and coding schemes and power scaling should be applied to optimize the PAPR and to provide an unequal error protection. • It is recommended to use a diversity mode based on space-time coding across all antennas of one sector. Different sectors of one BS as well as adjacent sites should be synchronized in a single-frequency network. • For small cells, the spatial multiplexing mode should be considered as an option.
[4] ETSI EN 300 744 v1.5.1, “Digital video broadcasting (DVB); framing structure, channel coding and modulation for digital terrestrial television,” European Standard, 11 2004. [5] I. E. Telatar, “Capacity of multi-antenna gaussian channels,” Europ. Trans. Telecommun., vol. 10, no. 6, pp. 585–596, 1999. [6] B. Hassibi and B. M. Hochwald, “High-rate codes that are linear in space and time,” IEEE Trans. Inf. Theory, vol. 48, no. 7, pp. 1804–1824, 2002. [7] S. Shamai and B. Zaidel, “Enhancing the cellular downlink capacity via co-processing at the transmitting end,” Proc. IEEE VTC ’01 Spring, vol. 3, pp. 1745–1749, 2001. [8] V. Jungnickel, S. J. amd L. Thiele, U. Krueger, A. Brylka, and C. Helmolt, “Capacity measurements in a multicell mimo system,” Proc. IEEE Globecom ’06, 2006. [9] S. Alamouti, “A simple transmit diversity technique for wireless communications,” IEEE J. Sel. Areas Commun., vol. 16, no. 8, pp. 1451–1458, 1998. [10] A. van Zelst, “Space division multiplexing algorithms,” Proc. IEEE MEleCon ’00, vol. 3, pp. 1218–1221, 2000. [11] S. Jaeckel and V. Jungnickel, “On the optimality of frequency-domain equalization in DFT-spread MIMO-OFDM systems,” Proc. IEEE WCNC ’08, pp. 1172–1177, 2008. [12] V. Jungnickel, T. Hindelang, T. Haustein, and W. Zirwas, “SC-FDMA waveform design, performance, power dynamics and evolution to MIMO,” Proc. IEEE PORTABLE ’07, 2007. [13] S. H. Han and J. H. Lee, “An overview of peak-to-average power ratio reduction techniques for multicarrier transmission,” IEEE Wireless Commun., vol. 12, no. 2, pp. 56–65, 2005. [14] 3GPP TR 25.996 v7.0.0, “Spatial channel model for multiple input multiple output (MIMO) simulations,” Tech. Rep., 6 2007. [15] L. Jiang, L. Thiele, and V. Jungnickel, “On the modelling of polarized MIMO channel,” Proc. EW ’07, 2007. [16] L. Thiele, M. Schellmann, W. Zirwas, and V. Jungnickel, “Capacity scaling of multiuser MIMO with limited feedback in a multicell environment,” Proc. ACSSC ’07, 2007.
IQ Modulation
V. C ONCLUSIONS
Power scaled Symbol Stream
Figure 9.
61 59
59 46
45
32
29
25 20
45 40
35
SCFDM Demapper
1
44 35
73
48
Binary Data Stream
38
47
49
46 40
MacroSTC 3x2 83
62
63
59
MacroSTC 3x1
78
78 66
70
64 57
30
20
MIMO Mux
81 69 55
40
MRC SFN
Mapped Symbol Stream
60
coverage (%)
80
SISO
