`
`A LOW-COMPLEXITY ANTENNA DIVERSITY RECEIVER
`SUITABLE FOR TDMA HANDSET IMPLEMENTATION
`
`Chun-Ning Zhang, William K. Lam, and Curtis C. Ling
`Department of Electrical and Electronic Engineering
`The Hong Kong University of Science & Technology
`Clear Water Bay, Kowloon, Hong Kong
`
`Abstract This paper introduces a low-complexity antenna
`diversity receiver suitable for TDMA handset implementa-
`tion. The receiver employs two branches of diversity, and
`is capable of adaptively choosing among three different di-
`versity techniques which are implemented in a single design
`which we refer to as a multi-diversity receiver. It consists of a
`single conventional wireless digital receiver chain augmented
`with a few additional low-cost passive RF components and
`minor control circuits. We present algorithms for efficient
`co-phasing and equal-gain combining, as well as a novel co-
`channel interference-reduction combining algorithm. These
`are implemented in a system design simulation which con-
`forms to the PACS standard. We then present simulation link
`performance for selection diversity (SD), equal-gain combin-
`ing (EGC) and interference-reduction combining (IRC) un-
`der Raleigh fading, flat fading with co-channel interference
`and selective fading conditions. The results show that EGC
`and IRC yield a signal-to-noise ratio (SNR) improvement of
`1 dB anda signal-to-interference ratio (SIR) improvement 4
`aBv~5.5 dB, respectively, compared with SD. The receiver also
`lowers the irreducible word error rate due to multipath delay
`spread when the Doppler shift is small. These results are
`compared with the results from other research.
`
`I.
`
`INTRODUCTION
`
`NTENNAdiversity can both improve the quality of
`communications in a wireless environment, and also
`:
`yield increased system capacity. Recent work by Cox and
`Wong has shown that two-antenna optimum-combining di-
`versity (OC) produces a signal-to-interference (SIR) im-
`provement of at
`least 3dB over conventional two-antenna
`selection diversity in Personal Access Communication Sys-
`tems (PACS)[1]-[3]. Furthermore, in a line-of-sight (LOS)
`environment, selection diversity provides no gain since the
`two branches are correlated, while combining diversity can
`still cancel co-channel interference and boost the desired sig-
`nal. Qualitatively speaking, in an LOS environment the OC
`receiver adjusts the joint signal of several antennas, resulting
`in an adaptive joint antenna pattern or polarization which
`attenuates co-channel interference while amplifying the de-
`sired signal. In a multipath environment, the two antennas
`may bereceiving signals from separate paths andthis picture
`is not entirely applicable, but the concept is the same.
`System complexity, with the associated power consump-
`tion, cost and size, is a significant barrier to implementing
`diversity techniques in handsets.
`In particular, most pro-
`posed schemes require one receiver chain for each branch of
`diversity [3]. For a two-branch diversity system, this doubles
`the receiver circuits from RF to baseband, andis not an easy
`tradeoff to make where reducing system complexity is essen-
`tial. Furthermore, many of the so-called adaptive antenna
`
`array solutions also rely on algorithms requiring antenna pat-
`terns which are well-characterized. In contrast, handset an-
`tennas possess patterns which are not carefully controlled
`and quite dependent on the position of the antenna with re-
`spect to the user’s hand and head. As such, most existing
`antenna diversity research to date has been limited to either
`simple selection diversity [4], or has targeted basestation im-
`plementation [3] where complexity issues are not significant
`compared with handset implementations.
`The receiver proposed in this paper meets some impor-
`tant design constraints which make it suitable for handset
`implementation employing TDMAarchitecture. First, the
`system utilizes a single receiver chain and baseband com-
`bining processor together with standard baseband process-
`ing techniques. The additional circuit components required
`to implement the system are low-cost passive RF front-end
`components which combine the antenna signals at RF. The
`system is sufficiently robust to handle poorly-defined, user-
`dependent antenna patterns. Furthermore, the system is able
`to implement several modesof diversity without changing the
`hardware or baseband processing, and can choose the most
`appropriate mode given the type of mobile usage andsignal
`environment. We therefore refer to this as a multi-diversity
`receiver. Finally, the techniques can be applied to a larger
`number of antennas at the cost of decreased mobility and
`lower tolerance to fading.
`Section II introduces two-antenna diversity receiver. Sec-
`tion III briefly describes the fully digital TDMA burst de-
`modulator. Section IV briefly reviews the selection diver-
`sity, then presents an efficient co-phasing method for equal-
`gain combining and a novel co-channelinterference-reduction
`combining (IRC) algorithm. Numerical results are shown
`and compared in section V.
`
`II. Two-ANTENNA MULTI-DIVERSITY RECEIVER
`
`The multi-diversity system presented here implements se-
`lection diversity (SD), equal-gain combining (EGC) and
`interference-reduction combining (IRC) using a single re-
`ceiver chain design.
`The system marginally increases the complexity of existing
`receivers to a level comparable with selection diversity, while
`achieving the performance gains of more complex systems
`under quasi-static multipath channel and interferer condi-
`tions. The receiver consists of the following components (see
`Fig. 1):
`the two antennas may consist of any type of an-
`tenna, as long as they meet spatial or polarization diversity
`
`0-7803-3@59-3/97. $10.00 ©1997 IEEE
`1753
`uthorized licensed uselimited to: Brigham Young University. Downloaded on July 07,2023 at 03:36:11 UTC from IEEE Xplore. Restrictions apply.
`SAMSUNG 1079
`AM
`1
`SAMSUNG v. SMART MOBILE
`SAMSUNG v. SMART MOBILE
`IPR2022-01004
`IPR2022-01004
`
` 1
`
`1
`
`

`

`
`
`they are uncorrelated and with same aver-
`conditions, i.e.
`age received signal power in a multipath environment. In an
`LOS environment, selection diversity loses its effectiveness
`while combining techniques still provide substantial gain over
`single-antenna systems, since selection depends on uncorre-
`lated signals at the two antennas while combining techniques
`_ effectively yield an adaptive antenna array with an improved
`pattern.
`
`
`
`Front End
`RF
`Circuit
`
`
`
`
`
`Fully-Digital
`Demodulator Multi-diversity
`
`TDMA Burst
`
`
`Baseband Processor
`
`
`Fig. 1. Block diagram of two-branch antenna multi-diversity receiver.
`
`The most significant hardware modifications needed to
`adapt a conventionalreceiver system are restricted to simple
`passive RF front-end components:
`« Two voltage-controlled RF attenuators (0~-20 dB)
`« One voltage-controlled RF phase shifter (0~360°)
`e Au RF sigual combiner
`Each diversity branch utilizes an attenuator for amplitude
`scaling. One of the branches contains the RF phase shifter.
`The resulting signals from each branch are summed to en-
`hance the received signal quality and fed into the remaining
`receiver chain. The RF front-end converts the signal down
`to a low IF which is quantized by an analog-to-digital (A/D)
`converter. The digitized signal is processed by a fully digi-
`tal TDMA burst demodulator. The baseband signal is fed
`into the multi-diversity baseband processor, which performs
`SD, EGC and IRCalgorithmsand adaptively adjusts the at-
`tenuators and phase shifter to optimize the received signal
`quality.
`
`Ill. Futty-DieitaL TDMA Burst DEMODULATOR
`
`The receiver is designed to operate in a TDMA environ-
`ment for PCS communications, such as PACS. Wehave im-
`plemented a practical system simulation which conforms to
`the PACS standard. The PACS standard uses 7DQPSK
`modulation, 384kbps channel bit rate, 120 bits per time
`slot, 8 time slots per frame, 312.5uSec burst duration and
`2.5mSec frame duration [5][6]. A fully digital demodulation
`architecture which is suitable for VLSI implementation has
`been proposed by Chuang and Sollenberger [4][7].
`In this
`proposed implementation, {DQPSK modulation and square
`root raised cosine (a = 0.5) pulse shaping is used. The re-
`ceiver also uses two square rootraised cosine (a = 0.5)filters
`for in-phase (I) and quadrature (Q) basebandsignal to match
`
`the transmitter for optimal performance in AWGN environ-
`ments [9]. Other digital phase modulation systems can be
`treated similarly.
`We adopt the fully-digital coherent demodulation tech-
`nique proposed by Chuangand Sollenberger [4][7]. It jointly
`estimates symbol timing and carrier frequency offset by op-
`erating on an individual TDMA burst without requiring a
`training sequence. These estimates produce a signal quality
`(SQ) measurement which is a good indicator of the degree
`of signal impairment caused by noise, delay spread or inter-
`ference, which closes the eye-opening of the detected signals
`(3][4]. Unlike using maximum average eye opening as symbol
`timing in [4], we use a square-law symbol timing scheme[11]
`to estimate timing, then use the values of I and Q at the
`sampling point to calculate the SQ and carrier phase (¢). A
`novel low-complexity diversity combining processor is added
`into the receiver to control the combining circuits. At the
`sametime, the signal strength is measured through received
`signal strength indicator (RSSJ) circuits.
`A low IF bandpass signal at 768kHz (4 times the sym-
`bol rate) is sampled with an A/D at 3.072MHz, resulting in
`an oversample of 16 samples per symbol. This is required
`to achieve symbol timing recovery, signal quality measure,
`frequency offset estimation and carrier phase recovery with-
`out overhead [4]. Using the same downconverter architec-
`ture, coherent and differential detection can be achieved for
`5DQPSK. Frequency offset. estimation is not addressed in
`this implementation. It can be removed either through RF
`frequency synthesizer or baseband frequency estimation [4].
`Since PACS bursts are very short relative to channel varia-
`tion, a quasi-static channel approximation which meansthe
`channel is static during a burst period is realistic applied to
`a flat fading study. All of the following simulation work. for
`this receiver is based on these assumptions.
`
`Frame = 2.5a
` B4 B3 B2 aBO
`heen — time
`rime slot=312.5u>——]
`[4fof0sTi) bits
`Synch SC
`FC Data
`CRC PC
`Synch: synchronous bits
`SC: system control andsupervisory bits
`CRC: error detection
`PC: power control
`
`FC: fast channel data
`
`Fig. 2. PACS downlink frame structure.
`
`IV. ANTENNA DIVERSITY METHODS
`
`The PACS downlinkutilizes continuous time division mul-
`tiplex (TDM) transmission. As such, we take advantage of
`bursts B1 ~ B4 preceding the desired burst BO (shown in
`Fig. 2) to measure the channel and set the combining di-
`versity parameters A;, A2, and 6, refer to Fig. 1. SD, EGC
`and IRC can be implemented via the same combiningcircuit
`and receiver chain. The system can then select the appropri-
`ate diversity technique most suited to the prevailing channel
`conditions. A,, Ag, and § are amplitudes of attenuators and
`
`1754
`Authorized licensed uselimited to: Brigham Young University. Downloaded on July 07,2023 at 03:36:11 UTC from IEEE Xplore. Restrictions apply.
`
`
`
`2
`
`

`

`
`
`phase of phase shifter, respectively. P;,S@Q;,¢; are the re-
`ceived signal power, signal quality and carrier phase of i’th
`burst, respectively. R = aa is the ratio of attenuation fac-
`tors. We can set the larger of A; and Az to be equal to 1
`and set the other equal to R or 1/R,limiting the value to
`the range 0.1 ~ 1.
`
`A. Selection Diversity (SD)
`SD uses two bursts (B2, B1) to calculate the signal quality
`of two branch andselect the better one for demodulating the
`consecutive desired burst (BO). Thefollowing steps are used:
`¢ B2: Select Ant.1 by setting Ay = 1, Ay = 0.1,48 = 0°
`Receive the burst and calculate SQ;.
`¢ BL: Select Ant.2 by setting A; = 0.1,Ay, = 1,6 = 0°
`Receive the burst and calculate SQe.
`¢ BO: Select the antenna with larger SQ, receive and de-
`modulate desired burst.
`
`B. Equal-Gain Combining (EGC)
`Because attenuators rather than variable gain amplifiers
`(VGA) are used, EGC is applied in order to minimize noise
`figure. Therefore, A, and Ag are set to 1, and the key pro-
`cessing which remains is to cophase the two branches.
`
`B.1 Co-phasing
`
`The local crystal oscillator has very high short-term stabil-
`ity. As such, we can assumeits frequency and phase are con-
`stant during several bursts, and use it as a reference. First,
`the system selects antenna 1 to receive a burst (B3) and re-
`cover phase ¢;. The system then selects antenna 2 during
`burst (B2) and obtains phase ¢2. ¢, and ¢2 have an ambigu-
`ity equal to an integer multiple of 90° introduced during the
`phase recovery process. This ambiguity causes no problems
`during coherent detection because differential decoding can
`remove it. However, we need to obtain the absolute phase
`difference between the two branchesin order to cophase them
`properly. Let ¢ = ¢, — $2 (with ambiguity), and @ equals
`the desired phase difference, with no ambiguity. To remove
`the ambiguity, the system employs an additional burst (B1)
`which is divided into four equal periods. During these peri-
`ods, the four possible phases are tested for combined signal
`power using the RSSI circuit. The phase yielding the small-
`est. power is then phase-inverted (+180°) to yield cophasing.
`
`B.2 Algorithm
`
`« B3: Select Ant.1 by setting A, = 1,A_q = 0.1,¢ = 0°.
`Receive the burst and calculate SQi, 41.
`e B2: Select Ant.2 by setting A, = 0.1, Az = 1,60 = 0°.
`Receive the burst and calculate SQe, de.
`e B1: Combine the two branches and cophase to get 6.
`« BO: Set A, = 1,A2 — 1, and @, thenreceive and de-
`modulate desired burst.
`
`C. Interference-Reduction Combining (IRC)
`In a high capacity PCS, for a given bandwidth, co-channel
`interference (CCI) limits the system capacity [1]. Usually,
`CCI is dominated by one co-channel interferer because of
`
`| Low IF Signal after A/D}‘
`
`Initial Setup (B3, B2, BI)
`Estimate A,,A,,6
`
`Search for A,
`
`, A.
`
`(B4, B3)
`
`
`
`
`
`
`
`
`Search for @ (B2, B1)
`
` .
`
`Coherent Demodulation (BQ)
`
`°
`
`‘
`wee mew ee ee ¢
`Output Received Data }
`
`Fig. 3.
`
`IRC algorithm with two antennas,
`
`the shadowing phenomenon, which has a log-normally dis-
`tributed local mean of received signal power.
`In order to
`cancel the primary source of CCI, the system must adjust
`A, and A» to make the received interference J; and Jp from
`each of the branches equal in amplitude, and then add them
`out of phase. Through system simulation, we obtain an ap-
`proximate relationship between SJR and SQ whenSIR is
`between 7~13 dB.. Thatis,
`
`SIR = 2 443889
`
`(1)
`
`This curve is similar to the result presented in [3]. Then
`SIR, and SIR» can be calculate from (1).
`
`Si
`SIR. = 7PHS th =h(l+ STR)
`1
`
`S
`SIR, = 5-1 Pa = Sot In = 1n(1 + SIR2)
`2
`After attenuating, the interference becomes,
`
`I, = Ayh, I, = Aah
`
`(2)
`
`(3)
`
`(4)
`
`1755
`Authorized licensed uselimited to: Brigham Young University. Downloaded on July 07,2023 at 03:36:11 UTC from IEEE Xplore. Restrictions apply.
`
`3
`
`

`

`
`
`To make Ii = J5, we get
`« A single dominant interferer is assumed in co-channel
`interference environment [8]
`_ Ay _ fe _ PltSiR)6
`« Delay spread effects are accounted for by using two-ray
`~ Ay ~ q, ~ P,(1+ S/Rg)
`model[3][4]
`« Quantization is not considered but 40 dB SNR is used
`as noise floor.
`The system is simulated with uniform distributed symbol
`timing and a maximum Doppler shift of fd=6 Hz which is
`corresponds to a traveling speed of 2 mph at 2GHz. RSSI
`and SQ provide a useful indicator of the channel conditions,
`and are used by the multi-diversity receiver to switch between
`the two combining schemes automatically. Below a defined
`threshold, the environmentis noise-limited and EGCis used.
`Otherwise the environmentis assumed to be CCI-limited and
`TRC is employed.
`In voice and data communication, word
`error rate (WER) is good performance indicator than bit
`error rate (BER) in a bursty error environment. All the
`simulation will be given in WER.
`
`The IRC algorithm is divided into three majorsteps: ini-
`tial setup, optimizing search, and coherent demodulation.
`Initial setup utilizes the received signal power P; and signal
`quality SQ; to estimate the amplitudes A; of attenuators
`from (1)~(5) then cophase. This provides the system with
`a very good starting point from which to begin interference-
`reduction. The optimizing search determines parameters for
`interference-reduction and tracks channel variations. Coher-
`ent detection recovers the received data and check the cyclic
`redundancy check (CRC) bits and co-channel interference
`control (CCIC) bits to determine how to next proceed. This
`is now described in detail (see Fig. 3):
`
`C.1 Initial Setup
`Define two SQ thresholds, Vi and V2, which are used to
`detect certain conditions in the algorithm flow.
`e B3: Select Ant.1 by setting A; = 1, A2 = 0.1,0 = 0°
`Receive the burst and calculate P;,5Q1, 41.
`» B2: Select Ant.2 by setting A; = 0.1,A2 = 1,6 = 0°
`Receive the burst and calculate P;, SQz, ¢2.
`e Using Pi, Po, $Q1, SQ2 to determine A,, As
`e B1: Combine the two branches and cophase to get 9.
`« When SQ > ¥V,, selection diversity will be used.
`e Proceed to detection step.
`
`C.2 Optimizing Search
`e B4: Ry = R+3dB, set A; and Ag to get SQ1.
`e B3: Re = R—3dB, set A, and Az to get SQe.
`e Select one of A, Ri, and Rz with max SQ.
`« B2: 6, = 6+ 45°, set 0, to get SQ1.
`«¢ B1: #2 = 6 — 45°, set G2 to get SQzo.
`« Select one of 6,1, and 92 with max SQ.
`
`C.3 Coherent Detection
`
`AverageWEA
`
`Equai Gain Combining 107 a $d L
`
`NoDiversity (Diff)
`No Diversity (Coh)
`Selection Diversity
`
`10
`
`12
`
`14
`
`16
`
`22
`20
`18
`Average Eb/No (dB)
`
`24
`
`26
`
`28
`
`30
`
`Fig. 4. Link performance under flat fading
`
`BO: Coherent detection and check CRC & CCIC.
`If CRC check fails and SQ < V, then restart a initial
`step else goto search step
`e When CRCcheck is passed and CCIC check fails, this
`means the combiner locks to a stronger inteferer. If the
`Aj or Ag equal 0.1 (only one antenna being used for
`receiving), a initial setup step will be executed and the
`phase6 will be inverted to cancel the strongerinterferer.
`Otherwise, simply invert the phase # and do searching.
`« If CRC & CCICare all passed, the receiver outputs re-
`covered data.
`
`V. SIMULATION RESULTS
`
`A. Flat Fading Environment
`
`Whenthere is no diversity, the coherent detection is about
`2.5 dB better than differential detection, see Fig. 4. There-
`fore coherent detection is adopted for remaining simulation.
`Underflat fading condition, the system uses EGC since pro-
`grammable attenuators, not VGA, are used. Noise is dom-
`inant and at a constant level, so Ai = Az = 1 yields the
`best noise figure. The two branches are then cophased. Sim-
`ulation results are presented in Fig. 4, and show that EGC
`provides a 1 dB improvernent over SD at WER=0.01. The
`results for SD performanceagree closely with those presented
`in [8].
`
`Thefollowing conditions, in addition to those described in
`the previous sections, are used in the computer simulation:
`« Coherent or differential detection
`¢ Jakes model [10] is used for generating correlated slow
`Rayleigh fading channel.
`
`B. Flat Fading with CCI Environment
`The IRC is employed to combat CCI. Fig. 5 shows that
`the performance of SD and EGC are close, and a substantial
`improvement of 4 dB over SD and EGC at a WERof 0.01
`with a Doppler frequency fd=6 Hz. The IRC is sensitive
`
`1756
`Authorized licensed uselimited to: Brigham Young University. Downloaded on July 07,2023 at 03:36:11 UTC from IEEE Xplore. Restrictions apply.
`
`4
`
`

`

`
`
`No Diversity
`Selection Diversity
`Equal Gain Combining
`Smart Combining
`——— fd=6H2z ---- fd=1Hz
`
`AverageWER §
`
`10
`
`26
`
`30
`
`This paper introduces a low-complexity antenna diversity
`receiver suitable for TDMA handset implementation. The
`receiver employs two branches of diversity, and is capable
`of adaptively choosing among selection, equal gain combin-
`ing and interference reduction combining techniques, which
`are implemented in a single multi-diversity recever design.
`The receiver employs a single conventional wireless digital
`receiver chain augmented with a few additional low-cost pas-
`sive RF components and minorcontrol circuits. Further-
`more, the algorithm presented is insensitive to antenna pat-
`tern variations. In this paper we have presented simulation
`link performance of the receiver under Raleigh fading, flat
`20
`15
`fading with co-channel interference and selective fading con-
`Average SIR (dB}
`ditions. The results show that EGC and IRC yieldasignal-
`Fig. 5. Link performance underflat fading with co-channelinterference
`to-interference ratio improvement of 1 dB and 4 dB~5.5 dB,
`respectively, compared with selection diversity. The receiver
`also lowers the irreducible word error rate due to multipath
`delay spread, especially when the Dopplershift is small. The
`performance and robust,
`low-complexity design of this re-
`ceiver makes it an excellent candidate for practical handset
`implementations.
`
`icant factor even when signal power is high. Under such
`conditions, IRC is used, treating multipath as CCI.
`
`VI. CoNcLUSION
`
`ACKNOWLEDGMENTS
`
`The authors would like to acknowledge J. C-I Chuang and
`P. B. Wong for their discussions during the course of this
`work.
`
`REFERENCES
`
`“Universal digital portable radio communications”,
`[1] D. C. Cox,
`IEEE Proc., vol. 75, no. 4, pp. 436-477, Apr. 1987.
`{2] P. B. Wong and D. C. Cox, “Low-complexity co-channelinterference
`cancellation and macro scopic diversity for high capacity PCS”,
`Conf. Record IEEE I[0'0°95, Seattle, WA, pp.852-857, Jun. 18-22,
`1995.
`[3] P. B. Wong and D. C. Cox, “Low-complexity diversity combining
`algorithmsandcircuit archi tectures for co-channel interference can-
`cellation and frequency selective fading mitigation”, [EEE Trans.
`Commun., vol. 44, no. 9, pp.L107-1116, Sep. 1996.
`[4] J. C-I Chuang and N.R.Sollenberger, “Burst coherent detection
`with combined symbol timing, frequency offset estimation, and di-
`versity selection”, [EEE Trans. Commun., vol. 39, no. 7, pp. 1157-
`1164, July 1991.
`[5] N. R. Sollenberger, J. C-I. Chuang, L. F. Chang, S. Ariyavisitakul,
`and H. W. Arnold, “Architecture and implementationof an efficient
`and robust TDMAframestructure for digital portable communica-
`tions”, IEEE Veh. Technol., vol.40, no.1, pp250-60, Feb. 1991,
`Motorola, Personal Access Communications System Air Interface
`Standard J-STD-014, SP-3418. June 1995.
`[7] N. R. Sollenberger and J. C-I Chuang, “Low-overhead symboltim-
`ing and carrier recovery for TDMA portable radio systems”, JEEE
`Trans. Commun., vol. 38, no. 10, pp. 1886-1892, Oct. 1990.
`[8] A. Afrashteh and N. R. Sollenberger, J. C-I Chuang, and D.
`Chukurov,
`“Performance of a TDM/TDMA portable radio link
`for interference, niose, and delay spread impairments”, IEEE Veh.
`Technol., vol. 43, no. 1, pp. 1-7, Feb. 1994.
`[9] K. Feher, “MODEMsfor emerging digital cellular-mobile radio sys-
`tem”, IEEE Veh. Technol., vol. 40, no. 2, pp. 1-7, May 1991.
`[10] W. C. Jakes, Ed., Microwave Mobile Communications,
`IEEE
`press, 1994.
`{11] J. G. Proakis, Digital Communications,
`McGraw-Hill, 1995.
`
`3rd ed., New York:
`
`[6=
`
`to Doppler shift because of the delay introduced by chan-
`nel measure. When the Doppler frequency is 1 Hz, the SIR
`improvement increases to 5.5 dB compared with EGC.
`A key feature of the IRC techniqueis its insensitivity to the
`accuracy of attenuators and phase shifter. The attenuator
`should have a range of 0~20 dB with + 1 dB accuracy, and
`the phase shifter should have a range of 0~ 360° range with
`+10° accuracy.
`
` 10
`
`x No Diversity
`©
`Selection Diversity
`+
`Equal Gain Combining
`%
`Smart Combining
`ome fd=6HZ --- fd=1Hz
`
`
`
`AverageIrreducibleWER
`
`Normalized delay spread by symbol period
`
`Fig. 6. Link performance under frequencyselective fading
`
`C. Frequency Selective Fading Environment
`
`Because we focus on cancelling interference, the IRC may
`be not suitable for dealing with multipath delay spread, just
`as cophasing is ineffective in dealing with CCI and delay
`spread. Fig. 6 shows that the IRC’s irreducible WER per-
`formancein frequency selective fading environmentis better
`than SD and EGC. Normally, delay spread can bea signif-
`
`757
`oa
`Authorizedlicensed uselimited to: Brigham Young University. Downloaded on July 07,2023 at 03:36:11 UTC from IEEE Xplore. Restrictions apply.
`
`5
`
`