CELLULAR OFDMKDMA DOWNLINK PERFORMANCE IN THE LINK AND
`SYSTEM LEVELS
`
`Antti Toskala', Jonathan Castro2, Laurent Chalard2, Seppo Hamalainen' and Kari KalliojLvi'
`
`'Nokia Research Center, P.O.Box 407, FIN-00045, NOKIA GiROUP, Finland
`2Centre Suisse d Electronique et de Microtechnique, CSEM, Jaquet-Droz 1, CH-2007 Neuchatel, Switzerland
`
`Abstract -- In this paper the link and system level performance
`of a cellular OFDM/CDMA downlink is evaluated. A system
`level simulator is built to utilise the link level simulation
`results and to derive the actual network capacity. Three service
`transmission rates, i.e., 12 kbitds, 144 kbits/s and 2 Mbits/s
`have been studied in the evaluation of the suitability of
`OFDM/CDMA as an UMTS multiple access method. The
`work has been performed within the European Union ACTS
`program in the FRAMES project.
`
`I. INTRODUCTION
`
`the Universal Mobile
`for
`requirements
`The
`Telecommunications System (UMTS) capability set high
`demands on the air interface of the 3rd generation mobile
`networks. In Europe, within the Advanced Communications
`Technologies and Services (ACTS) programme, the FRAMES
`(Future Eadio Widebgnd Multiple Access Systems) project has
`the objective to define and specify a UMTS air interface.
`At the beginning of the FRAMES project several multiple
`access solutions were proposed. After the initial evaluations,
`two hybrids were configured, namely: Hybrid I, SMA (Slotted
`Multiple Access) based on TDMA type transmission with an
`optional CDMA component; and Hybrid 11, CATS (Code And
`Time Division Multiple Access System) using continuos
`CDMA type transmission. The two FRAMES hybrids are
`described in more detail in [ 11. In addition, other details on the
`FRAMES Multiple Access ( M A ) , can be found in this
`proceedings [2].
`CATS applied asynchronous and synchronous modes in the
`uplink. For the downlink it used singlecode, multicode as well
`as OFDMKDMA options.
`OFDM has gained a lot of interest in the field of
`communications, where it has been used for example in
`broadcast applications like DAB and various DTV solutions
`[3-41. OFDM has also been considered earlier for 3rd
`generation mobile communication systems [5]. The OFDM
`methods studied in this paper follow the latter applications
`using OFDM/CDMA modes to combat frequency selective
`multipath fading, and to offer more resistance towards co-
`channel interference typical to cellular environments. The
`model utilises MLSE type detector, although other detector
`types, as studied in [6], could be used. The interference
`
`cancellation type receiver has also bjeen applied with
`OFDM/CDMA in [7].
`The capacity of a rnultiple access option is one of the key
`criteria in the air interface evaluation process. As an input for
`capacity studies link level performance parameters, i.e. values
`such as E n o for given BER, are used. These type of link level
`measurements for different schemes have solid ground because
`commercial signal processing simulation prlograms can be used
`to obtain the link level results of various multiple access
`techniques.
`However, the system level issues are more complicated,
`because each multiple access scheme has its own demands on
`the modelling to provide capacity calculations. Thus, dedicated
`tools or programs are often designed to adequately estimate
`capacity, and to obtain other useful information on the
`strengths and weaknesses of a particular multiple access
`scheme. For the OFDMKDMA downlink model this is
`precisely the case, special adaptations were made to a system
`level program to access its performance. Therefore, in the
`following we will briefly present the link and system level
`methods utilised in the evaluation process.
`
`11. THE, LINK LEVEL MODEL
`
`In the studies done within FRAMES hybrid 11, CATS, the
`OFDMKDMA symbol was formed with 1024 subcarriers,
`using 2.5 MHz bandwidth and resulting in 409.6 ps duration.
`Guard
`intervals were
`inserted
`between
`successive
`OFDM/CDMA symbols to avoid intersymbol interference
`from the channel mudtipath. The transmitted signal on N
`subcarriers of the qth OFDM symbol can be: given as
`
`y, = cs, = [c,c 2...c,]s,
`
`where the N x U ma.trix C is the spreading code matrix,
`consisting of U length N spreading sequences as its' columns.
`sq is an U vector consisting of the symbols transmitted with
`these U different spreading codes. The introduction of CDMA
`component brings frequency diversity to OFDM and allows
`better to combat the frequency selective fading on the
`subcarriers.
`
`0-7803-3659-3/97 $1 0.00 0 1 997 IEEE
`
`855
`
`Facebook's Exhibit No. 1034
`Page 1
`
`

`
`With long enough guard interval and the OFDM symbol
`being shorter than the time-coherence of the channel, the
`fading on each subcarrier will appear flat fading type and can
`be modelled with a complex attenuation term. Thus the
`received and de-multiplexed signal is then given as
`Z, = A,C,S, + W,
`
`results could be further improved to be nearer the ideal curve
`with more advanced channel estimator solutions.
`
`Lmk level results for 144 kbltds
`
`I
`
`I
`
`I
`
`where A, = diag(%,la+..%,N)
`is the channel co-efficient
`matrix, where the coefficients are time varying complex
`numbers.
`In the evaluation process, the subcarriers were grouped in
`256 sets of 4 subcarriers, each of them conveying 8 bits with
`specific codes to provide frequency diversity. Figure 1
`illustrates the frequency allocation of the sets, where each
`spread block is assigned to one specific set.
`
`lo-*
`
`a
`m w
`
`1 o3
`
`lod4
`
`4 5
`
`5
`
`5 5
`
`6
`
`6 5
`EbINo
`Fig. 2 E O , values with known channel parameters and with
`channel estimator using different amount of pilot subcarriers.
`
`7
`
`7 5
`
`8
`
`8 5
`
`9
`
`Table 1 Link level EJN, values used in the svstem simulations
`I E n 0
`Service
`I 7.9dB
`12 kbit/s. macro
`I 7,9dB
`144 kbit/s, macro
`I 6 3 dB
`2 Mbitsls, pic0
`
`Table 1 illustrates E D o values used in the system
`simulations. Notice that for the 12 and 144 kbit/s services the
`non-optimised channel estimation values were used. For the 2
`Mbit/s services the E D o level was estimated more realistically
`considering lower fading effects in the indoor environment.
`Table 2 illustrates other link level simulation parameters.
`In the preceding link level results each service was
`specified by its user data rate and its maximum end-to-end
`delay. The
`latter
`took
`into account
`the coding
`time,
`interleaving of
`the corresponding
`symbols,
`and
`the
`complementary processes at the receiver. 12 kbit/s with 40 ms
`delay represented telephone or voice services, while 144 kbit/s
`with 100 ms delay represented data services. High data rate
`transmission at 2 Mbitsls did not have delay restrictions, the
`resulting delay from interleaving in the simulations was 40 ms.
`Interleaving had higher impact on E D o values of the lower
`rate services than those of the 2 Mbits/s transmissions.
`
`fttt ... t
`
`0 1 2 3
`
`256
`
`Set 256
`
`4t
`
`512
`
`2.5MHz
`
`4t
`
`768
`
`ttt
`
`\
`1024-1
`
`Fig. 1 Frequency allocation
`
`The Maximum Likelihood Sequence Estimator (MLSE)
`method enabled detection with reasonable complexity. It
`sequence sq by
`evaluated
`the most
`likely
`transmitted
`minimising the squared Euclidean distance 6' between the
`received and all possible transmitted sequences. Other sub-
`optimal and simpler approaches can also be found in [6].
`The motivation to apply OFDWCDMA technique in this
`paper only
`to
`the downlink was
`the
`feasibility of
`implementation, and the offered flexibility in terms of service
`mapping. Also low level of intra-cell interference was ensured.
`Furthermore, service mapping benefits
`from maximum
`diversity, by means of hopping, interleaving and distribution of
`the data on several subcarrier sets. In the uplink especially the
`synchronisation is more demanding.
`
`111. LINK LEVEL RESULTS
`
`The link level simulations were done with three different
`services. In the Macro-cell environment 12 kbit/s and 144
`kbit/s were studied, and a pico-cellular channel model was
`used for 2 Mbitsls services.
`Results were derived with both known channel parameters
`and with actual channel estimation having different amounts of
`overhead for pilot subcarriers. Fig. 2 illustrates non-optimised
`E D o values with the channel estimator used together with the
`results from the ideal estimation with known channel state. The
`
`856
`
`Facebook's Exhibit No. 1034
`Page 2
`
`

`
`Table 2 Link level simulation parameters
`12.5 % pilot overhead,
`Channel estimation
`20 % total overhead
`
`Spreading
`
`Walsh-Hadamard,
`
`Bandwidth
`
`Modulation
`Detector
`Coding
`
`spacing 3 MHz.
`QPSK
`256 state MLSE
`1/2 Convolutional with
`constraint length 9
`Pulse shaping
`Raised cosine
`OFDM symbol duration I 0.4096 ms
`I 0.004 ms
`Guard Interval
`
`Finally, it should be mentioned that the link level results
`presented in this paper are initial values with models simulated
`at full loads.
`
`IV. SYSTEM LEVEL SIMULATOR
`
`The system level performance, i.e. cellular capacity was
`evaluated in two different environments. The environments
`chosen for OFDMKDMA for system level studies were macro
`and micro-cellular environments. The indoor environment at
`the system level is not covered in this paper. The simulation
`approach was static, i.e. independent snapshots were observed.
`The basic equation for the C/I of a user and a carrier with
`OFDMKDMA in the case of synchronously arriving signals
`is:
`
`( C l I ) i =
`
`j=O
`j t i
`
`P i x , i G p
`M
`
`k=O
`
`In the above equation, PRr,i is the receiver power of the
`carrier i. P2fk is the total received power from BTS k. Gp is
`the processing gain, aj is the orthogonality factor for intra-cell
`interference, plc models the orthogonality loss due to un-ideal
`channel estimation and due fading multipath channel, and No
`models the thermal noise. Since the system level simulations
`model the interference limited case, No can be neglected. The
`orthogonality factor 9 models how much the jth subcarrier that
`was received from the same BTS interferes the observed
`carrier. The parameter y models the orthogonality between the
`signals from different BTS.
`We can assume orthogonality inside the cell, since intra-
`cell interference was already included in the E n o values from
`the link level simulations. Thus, the term including its own cell
`interference will be
`
`j = O
`j t i
`
`(4)
`
`For inter-cell interference DS-CDMA type interference
`where the interference spreads over the whole bandwidth was
`assumed as signals come via different channels with different
`propagation delays and, multipaths and thus do not remain
`orthogonal and as the ;subcarrier set allocation was random.
`Because the network is (also asynchronous, both plc and y will
`have value 1. Since the interference is not always spread over
`the whole bandwidth the value for y would be then less than
`one and the bandwidth considered for the interference less than
`the total bandwidth. Also the use of Dynamic Channel
`Allocation (DCA) requires modifications to the used notation.
`The equation for single carrier is now given as
`
`The equation for a singlle user will be now
`
`L k=O
`
`1
`
`where
`
`(7)
`
`The power control concept is such that all the subcarriers
`(all the users) have the same transmission power within a BTS.
`The transmission powex used depends on the measurements
`performed by all the mobiles. An approach where parts of the
`OFDM/CDMA symbo1:s have slightly higher power, was also
`tested, but resulted
`in
`lower orthogonality
`inside
`the
`OFDWCDMA symbol in the link level.
`Thus, in the power control algorithm the C/I was measured
`for each mobile and the power balancing was made without
`command errors. The power control error dlue to the errors in
`measurements, channel state changes and errors in signalling,
`was modelled by adding a lognormal vaxiable with 3 dB
`variance to the transmission power level.
`In the macro-cellular environment, the following pathloss
`notation was used, with the assumption for the carrier
`frequency to be around 2.0 GHz.
`L = 29 + 3610g,,(.R) + 3llog,,(f)
`
`(8)
`
`The shadow fading in the macro-cellular environment was
`modelled as a lognormally distributed variable with standard
`
`857
`
`Facebook's Exhibit No. 1034
`Page 3
`
`

`
`deviation 10 dB. The macro-cellular base stations were placed
`on a hexagonal grid
`In the system level micro-cell simulator the base stations
`were located in every second street corner as illustrated in
`Fig. 3. The shadowing in the micro-cell environment was a
`lognormally distributed variable with 4 dB standard deviation.
`The buildings along the streets formed a heavy separation
`between micro-cells on different streets. The attenuation as a
`function of distance was modeled with a three slope model.
`Slopes are non-line-of-sight slope, and line-of-sight slopes for
`short distances and for long distances. If the connection
`between transmitter and receiver was a line-of-sight link, the
`attenuation is calculated as
`
`L,.,s =
`
`82 + 20 log(-),
`X
`300
`182 + 40log(-),
`300
`
`if x I 300meters
`
`i f x > 300meters
`
`(9)
`
`MICROCELL MODEL
`
`Fig. 3 The base station deployment in micro-cellular system
`level simulator.
`
`At a distance of 300 meters a breakpoint marks the
`separation between two line-of-sight segments. Turning round
`a corner causes an additional loss, L,,,,,,.
`Attenuation between
`a transmitter and a receiver
`that have non-line-of-sight
`connection constitutes a line-of-sight segment, a non-line-of-
`sight segment, and an additional corner attenuation, as seen in
`(10).
`
`Line-of-sight attenuation is calculated between a corner
`and receiver, and non-line-of-sight connection between a
`corner and transmitter.
`
`V. CAPACITY RESULTS
`
`Table 3 illustrates the values from the cellular capacity
`simulations for 12 kbits/s, 144 kbits/s and 2 Mbits/s without
`Dynamic Channel Allocation (DCA) techniques.
`
`Table 3 OFDM/CDMA spectral efficiency values for 5 %
`outage with reuse order of 3.
`I Service
`I Spectral efficiency [kbit/s/celVMHz] I
`I 58 (macro-cell)
`I 60 (macro-cell)
`I 240 (micro-cell)
`
`12 kbits/s
`144 kbits/s
`I 2 Mbits/s
`
`I
`
`including DCA, showed potential for
`Other results
`improvement up to 45-50 % with the lower rate services.
`In the spectral efficiency values scaling corresponding 3
`MHz carrier spacing was assumed. The signal -3 dB
`bandwidth is 2.75 MHz, but the actual carrier spacing to be
`used depends on the requirements set to the spectral mask and
`also from the spectral spreading due to the power amplifier. In
`the base station transmitter the use of highly linear amplifier is
`not considered to be that critical when compared to the mobile
`terminal. Thus moderate spectral spreading can be expected
`and the required carrier spacing is supposed to be 3 MHz or
`less.
`Based on the studies it could be noted that reuse 1 with
`OFDM/CDMA does not provide optimum results, higher reuse
`order is clearly needed for efficient operation.
`The power control algorithm used had problems since the
`users share the same OFDMKDMA symbol. Thus, setting
`different power level for different users was difficult. This
`resulted in that for part of the users the received signal level
`was too high, and thus generating unnecessary interference to
`the surrounding cells. This was the main reason for the lower
`capacity with 12 kbits/s and 144 kbits/s when compared to the
`2 Mbits/s service.
`With 2 Mbits/s the problem with OFDMKDMA power
`control was not relevant as a single 2.5 MHz carrier can
`support only a single user and thus power level can be adjusted
`optimally accordingly. The cell isolation was also much better
`in the micro-cellular environment than in the macro-cellular
`environment.
`
`VI. CONCLUSIONS
`
`level performance of cellular
`The link and system
`OFDWCDMA downlink was evaluated in this paper. As a
`multiple access method the OFDMKDMA can fulfil UMTS
`requirements in providing the required range of services up to
`2 Mbits/s.
`From the system level results can be concluded that for
`cellular applications OFDMKDMA works better with higher
`reuse values than 1, when the power control mechanism is not
`optimised. Also the case of having several users sharing the
`same OFDMKDMA symbol in the environments with low cell
`
`858
`
`Facebook's Exhibit No. 1034
`Page 4
`
`

`
`T. Ojanpcra et al., “Comparison of Multiple Access
`Schemes for UMTS”, VTC’97, Phoenix, Arizona,
`USA, In Proceedings, May 1997.
`M. Alard and I?.. Lassalle, “Principles of Modulation
`and Channel Coding for Digital Broadcasting for
`Mobile Receivers,” EBU Review, Tech. No. 224, Aug.
`1987, pp. 47-69.
`K. Fazel, S. Kaiser, P. Robertson and M.J.Ruf, “A
`Concept
`of
`Digital
`Terrestrial
`Television
`Broadcasting,” Wireless Personal Communications,
`vol. 2, NO. 1 pp. 9-27, 1995.
`K. Fazel, S. Kaiser and M. Schnell, “A Flexible and
`High Performance Cellular Mobile Communications
`System Based on Orthogonal Multi-Carrier SSMA, “
`Wireless Personal Communications, vol. 2, No. 1, pp.
`121-144, 1995.
`S. Kaiser, “On tlhe Performance of Different Detection
`Techniques for OFDM-CDMA in Fading Channels,”
`Globecom’95, Singapore, In Proceedings, pp. 2059-
`2063.
`K. Kalliojarvi, “Interference Cancellation in OFDM-
`CDMA Systems”, IEEE Nordic Signal Processing
`In Proceedings,
`Symposium, E!spoo,
`Finland.
`September 24-25’. 1996, Espoo, Finland
`J. Mikkonen, J. ICruys, “The Magic WAND: a Wireless
`ATM
`Access
`System”,
`ACTS
`Mobile
`Telecommunications Summit, Granada, Spain, In
`Proceedings, pp. 535-542, November 1996.
`
`isolation causes degradations in capacity as the power control
`implementation for several low rate users is difficult.
`The power control commands needed when coupled with
`DS-CDMA in the uplink caused some limitations since the
`OFDMKDMA symbols used in the present model were too
`long for the fast power control mechanisms used. Thus
`command rates in the order of 1 or 2 kHz could not be
`implemented in the downlink transmission for the mobile
`direction. However, it should be noted that the symbol length
`could be reduced with other configurations.
`Therefore, when using adaptable power control and shorter
`symbols, OFDWCDMA downlink coupled with DS-CDMA
`detection techniques in the uplink, would be an interesting
`alternative to balance restrictions seen in typical DS-CDMA
`systems. For example, OFDMICDMA based systems tend to
`be less susceptible to traffic loads. Changes in the length of the
`symbol would affect
`the diversity properties of
`the
`OFDWCDMA.
`On the other hand, the use of present OFDMKDMA model
`could also perform well when put together with FRAMES
`Multiple Access (FMA) Mode 1. There the transmission is
`naturally slotted, and applicability is not be restricted to the
`use of fast power control and reuse factor of 1, expedient in
`DS-CDMA based systems to obtain high capacity gains. In the
`further work of the FRAMES project the OFDMKDMA will
`be studied as a modulation option within the FMA Mode 1
`work.
`isolation,
`environment having higher cell
`an
`In
`OFDMKDMA achieves comparable capacity with 2 Mbit/s
`with a minimum power control when compared to DS-CDMA
`downlink utilising RAKE receivers in the mobile terminals.
`This kind of examples of the utilisation of OFDM can also be
`found for example in the ACTS project WAND [8] where
`OFDM is used for indoor wireless LAN type high bit rate
`applications.
`
`ACKNOWLEDGEMENT
`
`This work has been partially funded by the European
`Community in the ACTS program under the ACTS AC090
`FRAMES project. The authors acknowledge the contributions
`of colleagues from Siemens AG, Roke Manor Research
`Limited, Ericsson Radio Systems AB, Nokia Corporation,
`Technical University of Delft, University of Oulu, France
`Telecom CNET, Centre Suisse d’Electronique et de
`Microtechnique SA, Swiss Federal Institute of Technology
`Zurich, University of Kaiserslautern, Chalmers University of
`Technology, The Royal Institute of Technology, Instituto
`Superior Tkcnico.
`
`REFERENCES
`
`[ 11
`
`T. Ojanpera et al.,”FRAMES - Hybrid Multiple Access
`Technology”,
`ISSSTA’96 Mainz, Germany.
`In
`Proceedings, pp. 320-324, September 1996.
`
`859
`
`Facebook's Exhibit No. 1034
`Page 5