US007426175B2
`
`(12) United States Patent
`Zhuang et al.
`
`(10) Patent No.:
`(45) Date of Patent:
`
`US 7,426,175 B2
`Sep. 16, 2008
`
`(54) METHOD AND APPARATUS FOR PILOT
`
`(56)
`
`References Cited
`
`U.S. PATENT DOCUMENTS
`
`(75)
`
`SIGNAL TRANSMISSION
`Inventors: Xiangyang Zhuang, Hoffman Estates,
`.
`IL (US); Kevin L. Baum, Rolling
`__
`_
`Meadows:1L(US)9V1JaYN*"1g1as
`Scha11mb11rg,1L (US); Frederick W-
`Vook, Schaumburg, IL (US)
`(73) Assignee: Motorola, Inc., Schaumburg, IL (US)
`(*) Notice:
`Subject to any disclaimer, the term ofthis
`patent is extended or adjusted under 35
`U .S.C. 154(b) by Odays.
`
`1.
`
`
`
`.......... .. 370/332
`ki
`2/1996 R
`5,493,563 A *
`et a
`Ozans
`370/320
`8/2002 Bejjanietal.
`6,430,166 131*
`455/164.1
`3/2004 Matsumoto
`6,704,552 131*
`. . . .. 375/140
`. . . . .
`6/2004 Linde etal.
`6,744,807 B1*
`. . . . .. 375/299
`6,804,307 131 * 10/2004 Popovic . . . . . .
`0
`6 £1.
`. . . . . . . . . . . . . . . ..
`‘ ‘‘‘
`‘ ‘‘‘ "
`8/2003 R0m:n ..................... .. 375/140
`7/2007 McCoy .................. .. 370/344
`9/2007 Hosseinian etal.
`....... .. 375/260
`
`2003/0152136 A1 *
`2007/0165588 A1 *
`2007/0217530 A1*
`
`(21) Appl. No.: 10/813,476
`
`(22)
`
`Filed:
`
`Mar. 30, 2004
`
`(65)
`
`Prior Publicatiml Data
`
`US 2005/0226140 A1
`
`Oct. 13, 2005
`
`,1. Cited by examiner
`
`Primary Examiner—Chi H. Pham
`Assistant Examiner—Shicl< Hom
`
`(51)
`
`Int_C1_
`(2006.01)
`H04J11/00
`(52) U.s. Cl.
`..................... .. 370/203; 370/331; 370/342;
`375/134; 375/137; 375/140; 375/148; 455/436
`(58) Field of Classification Search ............... .. 370/203,
`370/252, 310, 331, 332, 334, 335, 342, 350;
`375/134, 137, 140, 142, 143, 144, 145, 1483
`375/149, 299
`See application file for complete search history.
`
`ABSTRACT
`
`(57)
`.
`.
`.
`Pilet sequences are constructed. from distinct “classes” of
`0111113 5§9ueQCe5 that l1aVe an 0131111131 CFP55 C01Te1aU011 Prop‘
`e1"Y~U“11Za“0n°f°h11‘PSequencesf°rP11°‘Se‘1uen°eSresults
`in pilot ‘sequences that have optimal or nearly-optimal cross
`Correlation and auto-correlation properties.
`
`28 Claims, 1 Drawing Sheet
`
`301
`
`303
`
`305
`
`DETERMINE NUMBER OF
`PILOT SEQUENCES NEEDED
`
`COMPUTE PILOT SEQUENCES
`
`ASSIGN PILOT SEQUENCES
`T0 BASE UNITS
`
`1
`
`APPLE 1012
`
`APPLE 1012
`
`1
`
`

`
`U.S. Patent
`
`Sep. 16,2008
`
`US 7,426,175 B2
`
`E?
`
`I04
`
`MOBILE UNIT
`
`ffl
`
`201
`
`202
`
`PILOT SEQUENCE
`
`REMAINING TRANSMISSION/ DATA
`
`FIG. 2
`
`
`
`2
`
`

`
`US 7,426,175 B2
`
`1
`METHOD AND APPARATUS FOR PILOT
`SIGNAL TRANSMISSION
`
`FIELD OF THE INVENTION
`
`The present invention relates generally to pilot signal trans-
`mission, and in particular to a method and apparatus for pilot
`signal transmission in a communication system.
`
`BACKGROUND OF THE INVENTION
`
`A pilot signal (or preamble) is commonly used for com-
`munication systems to enable the receiver to perform a num-
`ber of critical functions, including but not limited to, the
`acquisition and tracking of timing and frequency synchroni-
`zation, the estimation and tracking of desired channels for
`subsequent demodulation and decoding of the information
`data, the estimation and monitoring of the characteristics of
`other charmels for handoff,
`interference suppression, etc.
`Several pilot schemes can be utilized by communication sys-
`tems, and typically comprise the transmission of a known
`sequence at known time intervals. A receiver, knowing the
`sequence and time interval in advance, utilizes this informa-
`tion to perform the above-mentioned functions.
`Several criteria are important when determining pilot
`sequences for communication systems. Among these criteria
`is the ability to have good auto -correlation for each ofthe pilot
`sequences utilized, and at the same time the ability to have
`good cross-correlation between any two different pilot
`sequences. Auto- and cross-correlation are sequences them-
`selves corresponding to different shifts. Auto-correlation at
`shift-d is defined as the result of summing over all entries after
`an element-wise multiplication between the sequence and its
`conjugated replica after shifting it by d entries (d can be
`positive or negative for right or left shift). Cross-correlation at
`shift-d is defined as the result of summing over all entries after
`an element-wise multiplication between a sequence and
`another sequence that is conjugated and shifted by d entries
`with respect to the first sequence. “Good” auto-correlation
`results in each pilot sequence having a minimal auto-correla-
`tion value at all shifts of interest (i.e., a range of d, except for
`d:0). “Good” cross-correlation results in the pilot sequence
`having a minimal cross-correlation value at all shifts of inter-
`est. When the auto-correlation is zero at all d, except for d:0,
`it is referred to as “ideal” auto-correlation. Since the cross-
`
`correlation of two sequences that have ideal auto-correlation
`cannot be zero at all d, the minimum of the maximum cross-
`correlation values at all shifts can be reached only when the
`cross-correlation at all d is equal in amplitude, which is
`referred to as having “optimal” cross-correlation.
`Since the received signal after propagation consists of rep-
`licas of the delayed pilot sequence after some scaling factors,
`the ideal auto-correlation property of the pilot makes the
`estimation of the channel scaling factors at different delays
`possible. The optimal cross-correlation property between any
`two pilot sequences will minimize the interference effect seen
`at the receiver that is caused by any pilot sequences other than
`the desired one (i.e., one that the receiver is tuned to). Good
`cross-correlation makes the detection of the desired pilot
`signal and the estimation of the desired channel characteris-
`tics more reliable, which enables the receiver to perform
`synchronization and channel estimation more reliably.
`Various techniques have been used in the past to design
`systems with efiicient pilot sequences. For example, in the
`current CDMA-based cellular system, the pilot sequence in a
`cell is a Walsh code that is scrambled by a cell-specific scram-
`bling code (long code). This effectively randomizes the pilot
`
`10
`
`15
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`2
`
`sequence for each cell. Charmel estimation ofthe neighboring
`base stations, when required during a soft ha11doff, is simply
`performed by correlating the received signal with the neigh-
`boring base station’s long code scrambled pilot sequences.
`But
`the cross-correlation property of two random pilot
`sequences is not optimal, and thus a larger charmel estimation
`error can be expected. Therefore, a need exists for a method
`and apparatus for pilot signal or preamble transmission that
`optimizes both the cross correlation between pilot signals, as
`well as optimizing each pilot signal’s auto correlation.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`
`FIG. 1 is a block diagram of a communication system.
`FIG. 2 illustrates pilot signal transmission for the commu-
`nication system of FIG. 1.
`FIG. 3 is a flow chart showing pilot sequence assignment
`for the communication system of FIG. 1.
`
`DETAILED DESCRIPTION OF THE DRAWINGS
`
`To address the above-mentioned need, a method and appa-
`ratus for pilot signal transmission is disclosed herein. In par-
`ticular, pilot
`sequences are constructed from distinct
`“classes” of chirp sequences that have an optimal cyclic cross
`correlation property while satisfying the ideal cyclic auto-
`correlation requirement. Utilization of chirp sequences for
`pilot sequences results in pilot channels that have good cross
`correlation as well as having good auto-correlation.
`The present invention encompasses a method for assigning
`a pilot sequence to communication units within a communi-
`cation system. The method comprises the steps of assigning a
`first communication unit a first pilot sequence, wherein the
`first pilot sequence is selected from a group ofpilot sequences
`constructed from a set of Generalized Chirp-Like (GCL)
`sequences, and then assigning a second cor11r11unicatior1 ur1it a
`secondpilot sequence taken from the group ofpilot sequences
`constructed from the set of GCL sequences.
`The present invention additionally encompasses a method
`comprising the steps of receiving a pilot sequence as part of
`an over-tl1e air transmission, wherein the pilot sequence is
`constructed from a set of Generalized Chirp-Like (GCL)
`sequences, and utilizing the pilot sequence for at least acqui-
`sition and tracking of timing and frequency synchronization,
`estimation and tracking of desired channels for subsequent
`demodulation and decoding, estimation and monitoring of
`characteristics of other charmels for handoff purposes, and
`interference suppression.
`Finally, the present invention encompasses a communica-
`tion unit comprising pilot charmel circuitry for transmitting or
`receiving a pilot charmel sequence, wherein the pilot channel
`sequence comprises a sequence unique to the communication
`unit and is constructed from a GCL sequence.
`Turning now to the drawings, where like numerals desig-
`nate like components, FIG. 1 is a block diagram of commu-
`nication system 100 that utilizes pilot transmissions. Com-
`munication system utilizes
`an Orthogonal Frequency
`Division Multiplexing (OFDM) protocol; however in alter-
`nate embodiments communication system 100 may utilize
`other digital cellular communication system protocols such
`as a Code Division Multiple Access (CDMA) system proto-
`col, a Frequency Division Multiple Access (FDMA) system
`protocol, a Spatial Division Multiple Access (SDMA) system
`protocol or a Time Division Multiple Access (TDMA) system
`protocol, or various combinations thereof.
`As shown, communication system 100 includes base unit
`101 and 102, and remote unit 103. A base unit comprises a
`
`3
`
`

`
`US 7,426,175 B2
`
`3
`transmit and receive unit that serves a number of remote units
`
`within a sector. As known in the art, the entire physical area
`served by the communication network may be divided into
`cells, and each cell may comprise one or more sectors. When
`multiple antennas are used to serve each sector to provide
`various advanced communication modes (e.g., adaptive
`beamforming, transmit diversity, transmit SDMA, and mul-
`tiple stream transmission, etc.), multiple base units can be
`deployed. These base units within a sector may be highly
`integrated and may share various hardware and software
`components. For example, all base units co-located together
`to serve a cell can constitute what is traditionally known as a
`base station. Base units 101 and 102 transmit downlink com-
`
`munication signals 104 and 105 to serving remote units o11 at
`least a portion of the same resources (time, frequency, or
`both). Remote unit 103 communicates with one or more base
`units 101 and 102 via uplink communication signal 106.
`It should be noted that while only two base units and a
`single remote unit are illustrated in FIG. 1, one of ordinary
`skill in the art will recognize that typical communication
`systems comprise many base units in simultaneous commu-
`nication with many remote units. It should also be noted that
`while the present invention is described primarily for the case
`ofdownlink transmission from multiple base units to multiple
`remote units for simplicity, the invention is also applicable to
`uplink transmissions from multiple remote units to multiple
`base units. A base unit or a remote unit may be referred to
`more generally as a communication unit.
`As discussed above, pilot assisted modulation is com-
`monly used to aid in many functions such as charmel estima-
`tion for subsequent demodulation oftransmitted signals. With
`this in mind, base units 101 and 102 transmit known
`sequences at known time intervals as part of their downlink
`transmissions. Remote unit 103, knowing the sequence and
`time interval, utilizes this information in demodulating/de-
`coding the transmissions. Such a pilot transmission scheme is
`illustrated in FIG. 2. As shown, downlink transmissions 200
`from base units 101 and 102 typically comprise pilot
`sequence 201 followed by remaining transmission 202. The
`same or a different sequence can show up one or multiple
`times during the remaining transmission 202. Thus, each base
`unit within communication system 100 comprises pilot chan-
`nel circuitry 107 that transmits one or more pilot sequences
`along with data channel circuitry 108 transmitting data.
`It should be noted that although FIG. 2 shows pilot
`sequence 201 existing at the beginning of a transmission, in
`various embodiments ofthe present invention, the pilot chan-
`nel circuitry may include pilot sequence 201 anywhere within
`downlink transmission 200, and additionally may be trans-
`mitted on a separate channel. Remaining transmission 202
`typically comprises transmissions such as, but not limited to,
`sending information that the receiver needs to know before
`performing demodulation/decoding (so called control infor-
`mation) and actual information targeted to the user (user
`data).
`As discussed above, it is important for any pilot sequence
`to have optimal cross-correlation and ideal auto-correlation.
`With this in mind, communication system 100 utilizes pilot
`sequences constructed from distinct “classes” of chirp
`sequences with ideal cyclic auto-correlation and optimal
`cyclic cross-correlation. The construction of such pilot
`sequences is described below.
`
`Construction of a Set of Pilot Sequences to Use within a
`Communication System
`The construction ofthe pilot sequences depends on at least
`two factors, namely, a desired number of pilot sequences
`
`10
`
`15
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`4
`
`needed in a network (K) and a desired pilot length (NP) where
`K carmot exceed NP. In fact, the number of pilot sequences
`available that has the ideal cyclic auto-correlation and opti-
`mal cyclic cross-correlation is P—l where P is the smallest
`prime factor of NP other than ‘‘I’’ after factoring NP into the
`product of two or more prime numbers including “I”. For
`example, the maximum value that P can be is NP—l when NP
`is a prime number. But when NP is not a prime number, the
`number of pilot sequences often will be smaller than the
`desired number K. In order to obtain a maximum number of
`
`sequences, the pilot sequence will be constructed by starting
`with a sequence whose length N6 is a prime number and then
`performing modifications. In the preferred embodiment, one
`of the following two modifications is used:
`1. Choose NG to be the smallest prime number that is
`greater than NP and generate the sequence set. Truncate
`the sequences in the set to NP; or
`2. Choose NG to be the largest prime number that is smaller
`than NP and generate the sequence set. Repeat the begin-
`ning elements of each sequence in the set to append at
`the end to reach the desired length NP.
`The above design of requiring NG to be a prime number
`will give a set of NG—l sequences that has ideal auto corre-
`lation and optimal cross correlation. However, if only a
`smaller number of sequences is needed, NG does not need to
`be a prime number as long as the smallest prime factor of NG
`excluding “l” is larger than K.
`When a modification such as truncating or inserting is
`used, the auto-correlation will not be precisely ideal and the
`cross-correlation will not be precisely optimal anymore.
`However, the auto- and cross-correlation properties are still
`acceptable. The modified pilot sequence can be referred to as
`nearly-optimal pilot sequences that are constructed from
`GCL sequences with optimal auto- and cross-correlation.
`Further modifications to the truncated/extended sequences
`may also be applied, such as applying a unitary transform to
`them.
`
`It should also be noted that while only sequence truncation
`and cyclic extension were described above,
`in alternate
`embodiments of the present invention there exist other ways
`to modify the GCL sequences to obtain the final sequences of
`the desired length. Such modifications include, but are not
`limited to extending with arbitrary symbols, shortening by
`puncturing, etc. Again, further modifications to the extended/
`punctured sequences may also be applied, such as applying a
`unitary transform to them.
`The length-NP sequences are assigned to base units in com-
`munication system 100 as the time-domain pilot sequence, or
`as the frequency-domain pilot sequence (i.e., the entries ofthe
`sequence or its discrete IDFT will be assigned onto a set of
`subcarriers in the frequency domain). If the sequences
`obtained are used as the time-domain pilot, option 2 will be
`preferred because the autocorrelation over a size-NG window
`is still ideal. If the sequences obtained are used as the fre-
`quency-domain pilot and the channel estimation is performed
`in the frequency domain, the autocorrelation is irrelevant (but
`the cross-correlation properties of the sequences can still be
`important in many situations). In this case, either modifica-
`tion 1 or 2 is acceptable with a preference to choosing NG as
`the closest to NP.
`The final pilot sequences transmitted in time domain can be
`cyclically extended where the cyclic extension is typically
`longer than the expected maximum delay spread of the chan-
`nel (LD). In this case, the final sequence sent has a length
`equal to the sum of NP and the cyclic extension length. The
`cyclic extension can comprise a prefix, postfix, or a combi-
`nation of a prefix and a postfix. The cyclic extension may also
`
`4
`
`

`
`US 7,426,175 B2
`
`5
`be an inherent part ofthe communicatio11 system used such as
`an Orthogonal Frequency Division Multiplexing (OFDM)
`protocol. The inserted cyclic prefix makes the ordinary auto-
`or cross-correlation appear as a cyclic correlation at any shift
`that ranges from 0 to the cyclic prefix length. If no cyclic
`prefix is inserted, the ordinary correlation is approximately
`equal to the cyclic correlation ifthe shift is much smaller than
`the pilot sequence length.
`As discussed above, in the preferred embodiment of the
`present invention Generalized Chirp-Like (GCL) sequences
`are utilized for constructing pilot sequences. There exists a
`number of “classes” of GCL sequences and if the classes are
`chosen carefully (see GCL property 3 below), sequences with
`those chosen classes will have optimal cross-correlation and
`ideal autocorrelation. Class-u GCL sequence (S) oflength NG
`are defined as:
`
`Su:(au(0)b, au(1)b, -
`
`-
`
`-
`
`, au(NG_1)b)>
`
`where b can be any complex scalar of unit amplitude and
`
`k k
`l
`2
`k
`G
`a,,(/c) = exp[—j2nu ],
`
`(1)
`
`2
`
`(
`
`)
`
`where,
`. NG—l is known as the “class” of the GCL sequence,
`u:l, .
`.
`k:0, l, .
`.
`. NG—l are the indices of the entries in a sequence,
`q:any integer.
`
`Each class of GCL sequence can have infinite number of
`sequences depending on the particular choice of q and b, but
`only one sequence out of each class is used to construct one
`pilot sequence.
`It should also be noted that if an NG-point DFT (Discrete
`Fourier Transform) or IDFT (inverse DFT) is taken on each
`GCL sequence, the member sequences of the new set also
`have optimal cyclic cross-correlation and ideal autocorrela-
`tion, regardless of whether or not the new set can be repre-
`sented in the form of (1) and (2). In fact, sequences formed by
`applying a matrix transformation on the GCL sequences also
`have optimal cyclic cross-correlation and ideal autocorrela-
`tion as long as the matrix transformation is unitary. For
`example, the NG-point DFT/IDFT operation is equivalent to a
`size-NG matrix transformation where the matrix is an NG by
`NG unitary matrix. As a result, sequences formed based on
`unitary transformations performed on the GCL sequences
`still fall within the scope of the invention, because the final
`sequences are still constructed from GCL sequences. That is,
`the final sequences are substantially based on (but are not
`necessarily equal to) the GCL sequences.
`IfNG is a prime number, the cross-correlation between any
`two sequences of distinct “class” is optimal and there will
`NG—l sequences (“classes”) in the set (see properties below).
`The original GCL sequences have the following properties:
`
`Property 1: The GCL sequence has constant amplitude, and
`its NG-point DFT has also constant amplitude.
`Note that constant amplitude in both the time and fre-
`quency domain is desired for a pilot signal. Constant ampli-
`tude of the temporal waveform is ideal for a power amplifier
`to operate at higher output power without causing clipping.
`Constant amplitude in the frequency domain means that the
`subcarriers are equally excited and hence the channel esti-
`mates will not be biased. However, for multi-carrier systems
`such as OFDM, some ofthe subcarriers (typically those at the
`edges of the band) are unoccupied to form the guard band.
`
`6
`The corresponding time-domain pilot waveform is not of
`constant modulus anymore, but is essentially the result of
`interpolating the time-domain,
`i.e., over sampling the
`sequence to obtain a longer sequence after running it through
`a “sinc” filter. The resulting waveform still enjoys low peak-
`to-average ratio (PAPR is typically <3 dB).
`
`Property 2: The GCL sequences of any length have an “ideal”
`cyclic autocorrelation (i.e., the correlation with the circularly
`shifted version of itself is a delta function)
`
`Property 3: The absolute value of the cyclic cross-correlation
`function between any two GCL sequences is constant and
`equal to 1/, SIG, when |u1—u2|, u 1, and u2 are relatively prime
`to NG.
`
`Assignment ofPilot Sequences within a Communication Sys-
`tem
`
`Each communication unit may use one or multiple pilot
`sequences any number of times in any transmission interval
`or a communication unit may use different sequences at dif-
`ferent times in a transmission frame. Additionally, each com-
`munication unit can be assigned a different pilot sequence
`from the set of K pilot sequences that were designed to have
`nearly-optimal auto correlation and cross correlation proper-
`ties. One or more communication units may also use one pilot
`sequence at the same time. For example where multiple com-
`munication units are used for multiple antennas, the same
`sequence can be used for each signal transmitted form each
`antenna. However, the actual signals may be the results of
`different functions of the same assigned sequence. Examples
`of the functions applied are circular shifting of the sequence,
`rotating the phase of the sequence elements, etc.
`
`Receiver Functions that May Benefit from the Pilot Design:
`A number of critical receiver functions are described that
`can benefit from the above-described pilot design. The
`examples given here are not exhaustive, and it will be under-
`stood by those skilled in the art that various changes in form
`and details may be made therein without departing from the
`spirit of utilizing the good auto- and/or cross-correlation of
`the designed sequence.
`1. Single Channel Estimation:
`This section shows how the charmel estimation can benefit
`from the above pilot design strategy. In essence, channel
`estimation can be performed easily by correlating the
`received data with the pilot sequence. Thanks to the ideal
`auto-correlation of GCL sequences, the output of the corre-
`lation provides the channel estimate. The charmel estimate
`can then be refined, if desired, using a “tap selection” process.
`An example tap selection process is provided below. Also,
`time synchronization with the desired base station (BS) can
`be achieved straightforwardly because the arrival path can be
`detected easily. If channel information to an interference BS
`is also needed, it can be obtained from the correlation of the
`received data with the pilot sequence of that BS. The cross-
`correlation property increases the accuracy and detection reli-
`ability of the significant channel taps and reduces the false
`detections, as will be explained here.
`The GCL sequence effectively spreads the power of each
`tap of the interference channel evenly across NG taps thanks
`to the cross-correlation properties of GCL sequences. There-
`fore, after correlating with the desired sequence, the interfer-
`ence will be more evenly distributed in the time domain. The
`significant tap of the desired charmel will be preserved better
`than the smaller taps. In comparison, if non-GCL sequences
`are used, the power of each tap of the interference will not be
`evenly distributed across NG taps. The distortion effect on the
`
`10
`
`15
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`5
`
`

`
`US 7,426,175 B2
`
`7
`desired channel varies from tap to tap with an unpredictable
`behavior. Hence, with non-GCL sequences, the detected sig-
`nificant taps are more likely to be false due to the interference,
`or the true significant taps can be distorted so much that they
`become undetectable. The interference power on each desired
`tap is P,/NG with P, being the interference power; in other
`words, the spreading factor for each interference channel tap
`is NG.
`The correlation is typically performed in the time domain.
`But correlation can also be performed in the frequency
`domain as will be described below. Frequency domain esti-
`mation may be more computationally eflicient because ofthe
`FFT operation and is preferred for multicarrier systems such
`as OFDM systems. The example below is for an OFDM
`system.
`First, assume the frequency domain received data is Y(m)
`where m is a pilot subcarrier. Assume SG(m) is the pilot when
`m is a pilot subcarrier and zero otherwise, then a “noisy”
`channel estimate at the pilot subcarriers can be obtained as:
`
`8
`
`tap (denoted as 02) can be easily derived after accounting for
`the zeros at those null-subcarrier positions.
`Finally hwy will be transformed back to the frequency
`domain with a DFT to obtain the frequency channel response,
`and the windowing effect of (5) is preferably “de-empha-
`sized”, i.e.,
`
`13’m(m):13’w,L(m)/W(m)-
`
`(7)
`
`10
`
`15
`
`2. Multiple Charmel Estimation:
`When multiple channels corresponding to different pilot
`sequences are needed, the above single channel estimation
`process is conducted for the different pilot sequences one at a
`time or simultaneously. The characteristics learned about
`other charmels can be useful to improve the speed and per-
`formance of handoff, to perform interference suppression at
`the receiver for better demodulation and decoding, to enable
`the base unit to intelligently schedule transmission to avoid
`interference, etc.
`3. Synchronization Acquisition and Tracking:
`One way to achieve good initial acquisition of synchroni-
`zation to the desired base unit is to first correlate the received
`
`(3)
`
`signal with the pilot sequence candidates. The results will be
`assessed to find the desired base unit (e.g., the strongest one).
`The characteristics learned from the correlation with the
`
`Y(m)
`
`H,,(m) = { Sc(m)0
`
`if SG(m) i O
`
`if SG(m) = 0
`
`The noisy estimates will be transformed to the time domain
`through an IDFT as
`
`fiw:IDFT ({Hy.(m)W(m)}),
`
`(4)
`
`where w is a weighting window applied onto the noisy fre-
`quency response. The window is to reduce the power leakage
`problem caused by the discontinuity from the edge to null
`subcarriers (since zeros are inserted in place of the null sub-
`carriers before the IDFT). A “Harming” window can be used,
`1.e.,
`
`2mm
`w(m) = (0.5 + 0.5cos?],
`
`40
`
`(5)
`
`where the parameter F controls the shape of the window (an
`infinite F means a flat window).
`The resulting hw will then be truncated to length-LD to
`obtain hW,L. Furthermore, only the “significant” charmel taps
`in hw,L should be included before being DFT’d back to get the
`frequency domain response, i.e.,
`
`HM —1DFTN(hM).
`
`(6)
`
`The tap selection procedure is important, as described ear-
`lier, for exploiting the cross-correlation property of the pilot
`sequences. Tap selection also tries to enforce the frequency
`correlation according to the instantaneous charmel delay pro-
`file, which can improve the charmel estimation especially in
`the case of a sparse charmel.
`A threshold (denoted as 11) used in tap selection should be
`determined according the noise-plus-interference power esti-
`mated previously, or the total noise-plus-interference power
`over the used bandwidth can be estimated from the samples in
`hw that will be discarded (after LD+l). Note that compensa-
`tion for the windowing effect is recommended during noise
`power estimation. Based on the above noise-plus-interfer-
`ence power over the occupied bandwidth, the corresponding
`time-domain reference noise-plus interference power at each
`
`45
`
`50
`
`55
`
`60
`
`65
`
`25
`
`30
`
`35
`
`desired pilot sequence may be used to adjust the timing and
`frequency of the receiver to achieve synchronization. For
`example, the channel knowledge will give a good indication
`ofthe arrival time ofthe propagation paths and their strength,
`so the sample timing can be adjusted accordingly. The corre-
`lation results may also be used to adjust the frequency offset
`of the receiver. For example, correlation results from pilot
`sequences received at nearby but different times can be com-
`pared to identify the frequency offset. In another example,
`when the pilot sequence is mapped onto a set of OFDM
`subcarriers, a frequency domain correlation can identify fre-
`quency offsets to the nearest integer number of subcarriers.
`The tracking of the synchronization to the desired signal
`can also be accomplished through the correlation results
`where only correlation with the desired pilot is required. The
`fine tuning ofthe timing and frequency offset can be achieved
`as in the initial acquisition step.
`Another type of synchronization required is frame syn-
`chronization. Since a frame consists of many symbols, the
`information content at different locations within a frame may
`be different. The ability to detect the frame boundary is a
`prerequisite for decoding the information. The pilot
`sequences can be used to support this function as well. For
`example, if multiple pilot sequences are assigned in a frame
`and the location ofeach sequence relative to the frame bound-
`ary is designed to be fixed, when a certain pilot sequence is
`detected, the frame boundary can then be determined.
`FIG. 3 is a flow chart showing the assignment ofpilot codes
`to various base units within communication system 100. The
`logic flow begins at step 301 where a number ofneeded pilots
`(K), desired pilot length (NP) and a candidate length (N6) of
`each pilot sequence are determined. Based on NP and NG, the
`pilot sequences are computed (step 303). As discussed above,
`the pilot sequences are constructed from the Generalized
`Chirp—Like (GCL) sequences of length NP, with each GCL
`sequence being defined as shown in equation (1). Finally, at
`step 305, the pilot sequences are assigned to base units within
`communication system 100. It should be noted that each base
`unit may receive more than one pilot sequence from the K
`available pilot sequences. However, at a minimum a first base
`unit is assigned a first pilot sequence taken from a group of
`GCL sequences while a second base unit is assigned a differ-
`ing pilot sequence from the group of GCL sequences. During
`
`6
`
`

`
`US 7,426,175 B2
`
`9
`operation, pilot channel circuitry within each base unit will
`transmit the pilot sequence as part of an overall strategy for
`coherent demodulation. Particularly, each remote unit within
`communication with the base units will receive the pilot
`sequence and utilize the pilot sequence for many functions,
`such as channel estimation as part of a strategy for coherent
`demodulation of the received signal.
`As described above, the pilot sequences of the present
`invention have a low peak-to-average ratio (PAPR). As a
`result, the PAPR of a pilot signal/sequence of the present
`invention is lower than the PAPR of data signals that are also
`transmitted by a communication unit. The low PAPR property
`ofthe pilot signal enables pilot charmel circuitry 107 to trans-
`mit the pilot signal with a higher power than the data in order
`to provide improved signal-to-noise/interference ratio on the
`pilot signal received by another communication unit, thereby
`providing improved channel estimation, synchronization,
`etc.
`
`While the invention has been particularly shown and
`described with reference to a particular embodiment, it will
`be understood by those skilled in the art that various changes
`in form and details may be made therein without departing
`from the spirit and scope of the invention. For example,
`although the above discussion was related to assignment of
`pilot sequences to base units, it would be obvious to one of
`ordinary skill in the art that such pilot sequences may be
`assigned to other forms of transmitters/systems, such as but
`not limited to remote units, in which case the base station tries
`to detect the desired remote unit and estimate its channel. It is
`
`intended that such changes come within the scope of the
`following claims.
`The invention claimed is:
`
`1. A method for assigning a pilot sequence to communica-
`tion units within a communication system, the method com-
`prising the steps of:
`assigning a first communication unit a first pilot sequence,
`wherein the first pilot sequence is selected from a group
`o