`
`979
`
`A Combined OFDM-CsDMA Approach to Cellular Mobile Communications
`Li Ping, Member, IEEE
`
`frequency-
`Abstract— We present a combined orthogonal
`division multiplexing (OFDM) and code-division multiple-access
`cellular system. It
`is derived from the OFDM model by
`introducing an extra code-shift division multiple access layer.
`This mitigates intercell interference without affecting intracell
`orthogonality and such property can be maintained in multipath
`environments. The new scheme is suitable for both up and
`down links. It retains the low-receiver complexity property of
`the OFDM system.
`
`Index Terms—Code-division multiple-access, land mobile radio
`cellular systems, mixed TDMA/FDMA/CDMA systems, personal
`communication networks.
`
`I. INTRODUCTION
`
`THE CAPACITY of a code-division multiple-access
`
`system can be enhanced by suppressing
`(CDMA)
`interference
`(MAI). Maximum-likelihood
`multiple-access
`(ML) estimation technique can be used for this purpose,
`but the complexity involved is generally high [1]–[3].
`In this paper we present a new CDMA scheme free from
`intracell MAI. Starting from an orthogonal frequency-division
`multiplexing (OFDM) model [4], we derive the proposed
`scheme by introducing an extra code-shift division multiple
`access (CsDMA) layer [5]. The new layer does not affect the
`normal transmission of the original system, but it provides
`a spreading effect to alleviate the intercell MAI problem.
`The orthogonality among same-cell users is maintained in a
`multipath environment. The receiver is realized by a simple
`quasi-coherent detection technique without the necessity of
`high-cost ML estimation.
`When used for the up links, frame synchronization among
`same-cell users is assumed in the proposed scheme. However,
`the accuracy requirement is very flexible as frame synchro-
`nization error can be treated in essentially the same way as
`random delay.
`
`II. THE CYCLIC PREFIX TECHNIQUE
`and
`be two discrete sequences
`. Define sliding convolution
`
`modulo
`
`Let
`and
`
`:
`
`Fig. 1. The cyclic prefix transforms a sliding convolution into a circular one.
`
`-point cyclic convolution
`
`:
`
`(1b)
`
`The multipath effect in a digital system is usually modeled
`where
`is the transmitted
`by a sliding convolution
`sequence and
`is a vector of the path reflection coefficients.
`Assume that has a limited nonzero span in (0,
`), i.e.,
`for
`and
`(see Fig. 1). Given
`, the prefix of
`is its cyclic extension defined
`by
`. Assuming
`, the prefix
`technique transforms a sliding convolution into a circular one
`as
`
`(2)
`
`with the understanding that (2) is
`or simply
`defined only in [0,
`]. This relationship is the basis for
`OFDM [4] as well as for the new system introduced below.
`
`III. THE SYSTEM MODEL
`
`(1a)
`
`Paper approved by E. S. Sousa, the Editor for CDMA Systems of the IEEE
`Communications Society. Manuscript received April 8, 1997; revised August
`6, 1998 and November 6, 1998. This work was supported by City University
`of Hong Kong Strategic Grant 7000804.
`The author is with the Department of Electronics Engineering, City Univer-
`sity of Hong Kong, Kowloon, Hong Kong (e-mail: eeliping@cityu.edu.hk).
`Publisher Item Identifier S 0090-6778(99)05231-9.
`
`1999 IEEE
`
`A. The OFDM System
`are
`and
`Consider the system model in Fig. 2 where
`a pair of discrete Fourier transformation (DFT) and inverse
`discrete Fourier transform (IDFT) operators. We first ignore
`the effect of sequence
`by setting
`.
`Then
`and the system
`reduces to a common OFDM [4]. A prefix is padded to
`before transmitting. Let the underlying quadrature amplitude
`modulator (QAM) layer be described by the well-known model
`[6], where
`and
`represent
`the intersymbol interference and additive noise, respectively.
`0090–6778/99$10.00 ª
`
`Facebook's Exhibit No. 1040
`Page 1
`
`
`
`980
`
`IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 47, NO. 7, JULY 1999
`
`Fig. 2. Overview of the proposed OFDM-CsDMA system. The QAM layer
`is based on the well-known quadrature amplitude modulation principle [6].
`
`Provided that the prefix length is longer than the maximum-
`delay dispersion, from (2)
`
`or
`
`(3)
`
`(4)
`
`when
`. Applying DFT to
`and
`the end-to-end relationship
`
`in (4), we have
`
`(5)
`
`where
`. Equation (5) implies
`and
`that the orthogonality among the OFDM frequency carriers is
`maintained in a multipath environment [4].
`
`B. The OFDM-CsDMA System and Orthogonality
`Among Same-Cell Users
`We now derive the proposed system by relaxing the con-
`straint on
`as
`
`where
`is the conjugate of
`Based on (1b)
`
`. Now
`
`(6)
`
`in Fig. 2.
`
`(7)
`
`with
`positions.
`toward right by
`the cyclic shift of
`Equation (7) can be regarded as spreading
`, which
`by
`is equivalent to the CsDMA principle in [5] and, hence, the
`name for the middle layer.
`Suppose that a cyclic prefix is padded to
`still holds. Substituting the relationships
`and
`into (3), we have
`
`and so that (3)
`
`with
`. The similarity between (4) and (8) indicates
`that the upper layer transfer function of the wanted signal is not
`affected by the CsDMA layer. In particular, the orthogonality
`relationship implied by (5) still holds.
`
`(8)
`
`Fig. 3. A general signal structure of vvv. The dark signals represent prefixes.
`
`. Assume that the
`Fig. 3 is a more general structure for
`input
`is partitioned into sequential frames. The IDFT of
`each frame of
`is referred to as an OFDM frame. In Fig. 3,
`OFDM frames together with their prefixes are multiplexed
`in , which are convoluted with
`together. An extra prefix is
`padded to the resultant
`. In this way, it can be verified that
`(5) is again valid. Let the OFDM frame length be
`. Each
`frame contains
`orthogonal carriers distinguished by their
`discrete frequencies [4]. A total of
`orthogonal carriers
`can be established within
`in this way and they can carry
`complex symbols.
`Consider to share these
`carriers among users in a cell
`for the up-link. Assume that same cell users are approximately
`synchronized with certain tolerance for frame synchronization
`error that has essentially the same effect as multipath delay.
`Provided that the combined effect of frame synchronization
`error and multipath delay is covered by the prefix length,
`the orthogonality among all the users can be maintained. This
`distinguishes the proposed system from a common CDMA.
`
`C. Intercell MAI Mitigation
`We assign a set of different
`sequences, referred to as
`master codes, to the neighboring cells. Suppose that the master
`codes are properly designed to be approximately random to
`each other. For intercell MAI, the convolution involving
`at
`the receiver is applied to the signal generated from different
`master codes. The distortion is mitigated by the randomness
`among master codes. The effect can also be seen as to evenly
`distribute the intercell MAI among all the users, which avoids
`the worst case interference between OFDM carriers of the
`same discrete frequency. This property will be demonstrated
`by the simulation results later.
`
`D. Transceiver Complexity
`the main difference
`So far as transceiver is concerned,
`between OFDM and OFDM-CsDMA is the convolutions in-
`volving
`and . These operations incur very modest costs
`if
`is carefully chosen. For example,
`let
`be a length-
`-sequence,
`an all-1 sequence,
`and
`. Then
`satisfies (6)
`is trivial,
`. Since
`, which can be rewritten in a
`
`[7]. Now
`we will concentrate on
`matrix form as
`
`...
`...
`
`...
`
`(9)
`
`Facebook's Exhibit No. 1040
`Page 2
`
`
`
`PING: A COMBINED OFDM-CSDMA APPROACH TO CELLULAR MOBILE COMMUNICATIONS
`
`981
`
`Augment
`
`by one zero row and one zero column to give
`
`(10)
`
`form the
`. The columns of
`Then
`codeword set of a length
`augmented
`-
`sequence, which is equivalent to a Hadamard code up to an
`and
`such
`interleaving. There exist permutation matrices
`that
`, where
`is a Hadamard matrix. Hence
`
`(11)
`
`Equation (11) can be implemented very efficiently by a fast
`additions,
`Hadamard transform (FHT), costing
`and two interleavers. For detailed discussion, see [5].
`
`IV. SIMULATION STUDY
`
`A. System Model
`The underlying propagation is based on COST-207 (typical
`urbane) models [8] with maximum-delay spread
`5
`s,
`maximum-frame alignment error
`5
`s, chip duration
`0.9977
`s, bandwidth
`1.002 MHz, and carrier frequency
`1 GHz. The prefix length is
`. Rectangular pulse
`shaping and simple mixing-integration demodulator are used
`for the QAM layer.
`Input information bits are encoded by a rate 1/2, constraint
`length 9 convolutional code (IS-95 rate-1/2 code) with 192 bits
`in a frame, producing a length 400 binary sequence (including
`16 tailing bits), from which a length-200 complex sequence
`is formed in a quantenary phase-shift keying manner. It is
`convolutionally interleaved to generate the input vector
`in
`Fig. 2.
`The master codes
`-
`are constructed from length 511
`sequences (different cells using different
`-sequences). The
`parameters for
`(see Fig. 3) are
`, and
`. One OFDM frame is used to estimate
`phase reference, which avoids the complicated pilot scheme.
`The remaining ten frames are used to carry information. We
`observed that the performance can be improved by setting
`the power level of the reference frame to about 5 dB above
`the information frames, which are used in all of the results
`presented below. The information symbols, prefixes and ref-
`erence frame occupy a total of 473 positions in , leaving
`38 positions unused. Within one frame of
`, every user is
`assigned with one fixed carrier, distinguished by a unique
`discrete OFDM frequency,
`in all 11 OFDM frames. This
`carrier changes for every frame of
`to average out the fading
`effect in a frequency-hopping manner. Leaving one carrier as
`reference, the remaining ten carriers carry a total of 20 coded
`bits. With a gating factor of 1/2, we obtain user rate
`kbps.
`
`B. Results
`Using dual antenna diversity and equal gain combining,
`BER = 10
`can be achieved at
`dB. Leaving
`a 1-dB implementation margin, we set required
`dB. Substitute this into the formula of [9] and consider the
`
`Fig. 4. Simulated up-link performance comparison with worst-case interfer-
`ence. Every interfering signal has the same mean arrival power as the wanted
`signal. The comparable random waveform CDMA employs spreading ratio
`105, a rate 1/3, constraint length 9 convolutional code, dual receiving antenna
`and a four-finger Rake receiver. The OFDM system is obtained by setting
`ppp = and keeping everything else unchanged.
`
`elimination of the intracell MAI, the proposed system can
`achieve an uplink capacity of 54 user/cell. This is compared
`with about 24 user/cell for a random waveform CDMA system
`with similar spreading ratio of about 105 [9].
`We observed that the system is not sensitive to vehicle
`speed up to about 400 km/h. Similar to the standard OFDM,
`the signal envelope of the proposed scheme is not constant.
`When peak-to-mean power ratio is clipped to 3 and 0 dB,
`respectively, the corresponding performance losses are about
`0.2 and 0.8 dB at BER = 10
`.
`A distinguished feature of CDMA is its robustness against
`worst case intercell MAI. This is demonstrated for the pro-
`posed scheme in Fig. 4 (vehicle speed = 50 km/h). Perfect
`power control is used. Every user is allocated to the cell
`that results in minimum-transmission power. Consequently,
`the maximum-arrival power level (averaged over fast fading)
`of an interfering signal cannot exceed that of the wanted
`signal. Let the service quality be BER
`and
`be the maximum number of worst case interferers that can be
`tolerated. It is seen that the values of
`are, respectively,
`3 for OFDM, 38 for OFDM-CsDMA, and 44 for a comparable
`random waveform CDMA. Recall that
`for a random
`waveform CDMA includes all
`the same-cell users. As a
`fair comparison, a fully loaded random waveform CDMA
`system with 24 users/cell can tolerate only about 20 worst-
`case other-cell interferers, which is considerably lower than
`OFDM-CsDMA. This clearly demonstrates the advantage of
`the proposed scheme.
`
`V. DISCUSSIONS AND CONCLUSIONS
`Eliminating intracell MAI may potentially lead to 2.8 times
`of CDMA capacity increase [9] relative to a random wave-
`form CDMA. The example in Section IV achieves a capac-
`ity increase of about 2.2 times. The difference is due to
`the overheads (prefix and reference) involved. The proposed
`scheme is most suitable for applications with relatively low-
`delay dispersion, such as indoor systems. The low-receiver
`complexity of the proposed scheme makes it suitable to
`handle relatively high-data rates typically required in such
`applications.
`
`Facebook's Exhibit No. 1040
`Page 3
`
`
`
`982
`
`IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 47, NO. 7, JULY 1999
`
`REFERENCES
`
`[1] S. Moshavi, “Multiuser detection for DS-CDMA communications,”
`IEEE Commun. Mag., vol. 34, pp. 124–136, Oct. 1996.
`[2] J. Ruprecht, F. D. Neeser, and M. Hufschmid, “Code time division
`multiple access: An indoor cellular system,” in Proc. IEEE Vehicle
`Technology Conf., 1992, pp. 736–739.
`[3] K. Fazel, “Performance of CDMA-OFDM for mobile communications,”
`in Proc. IEEE Int. Conf. Universal Personal Communications, 1993, pp.
`975–979.
`[4] M. Alard and R. Lassalle, “Principle of modulation and channel coding
`for digital broadcasting for mobile receivers,” EBU Rev., no. 224, pp.
`168–190, Aug. 1987.
`
`[5] L. Ping, “Code-shift division multiple access for cellular mobile,” in
`Proc. IEEE Vehicle Technology Conf., 1997, pp. 377–381.
`[6] J. G. Proakis, Digital Communications. New York: McGraw-Hill,
`1995.
`[7] H. Harada, G. Wu, K. Taira, Y. Hase, and H. Sasaoka, “A new multicode
`high-speed mobile radio transmission scheme using cyclic modified
`m-sequence,” in Proc. IEEE Vehicular Technology Conf., 1997, pp.
`1709–1713.
`[8] ETSI/TC, CEPT/CCH/GSM, GSM Recommendation 05.05: Transmis-
`sion and Reception, Draft 3.1.0, Feb. 1988.
`[9] A. Viterbi, “The orthogonal random waveform dichotomy for digital
`mobile personal communication,” IEEE Personal Commun. Mag., vol.
`1, pp. 18–24, 1994.
`
`Facebook's Exhibit No. 1040
`Page 4



