United States Patent
`
`[191
`
`Lennen
`
`USO05402450A
`
`[11]
`
`Patent Number:
`
`5,402,450
`
`[45] Date of Patent:
`
`Mar. 28, 1995
`
`[54]
`
`[75]
`
`[73]
`
`[21]
`
`[22]
`
`[5 1]
`[52]
`
`[53]
`
`[56]
`
`SIGNAL TIMING SYNCHZRONIZER
`
`5,157,695 10/1992 Westerfield et al.
`
`........... 375/116 X
`
`Inventor:
`
`Assignee:
`
`Gary R. Lennen, San Jose, Calif.
`
`Trimble Navigation, Suxmyvale,
`Calif.
`
`Primary Examiner—Tesfa1det Bocure
`Attorney, Agent, or Firm—Wil1iam E. Pelton
`
`[57]
`
`ABSTRACT
`
`Appl. No.: 823,980
`
`Filed:
`
`Jan. 22, 1992
`
`Int. Cl.5 ...................... .. H04L 27/06; H04L 7/00
`
`U.S. Cl.
`............... 375/343; 375/346;
`375/368
`Field of Search ................... .. 375/96, 116, 119, 7,
`375/115, 99, 58; 342/350, 356, 357, 358, 378;
`364/728.03
`
`References Cited
`
`U.S. PATENT DOCUMENTS '
`
`4,545,061 10/1985 Hileman ............................ 375/96 X
`
`4,829,543
`5/1989 Borth et al.
`375/96 X
`5,148,452
`9/1992 Kennedy et al.
`..................... 375/96
`
`A method and apparatus are disclosed for characteriz-
`ing multipath-induced distortions in the autocorrelation
`function of a correlation receiver in order to reduce
`effects of these multipath-induced distortions on the
`accuracy of detecting the time of arrival of a received
`signal. The magnitude of the multipath-induced errors
`adversely affecting the shape of the autocorrelation
`function is estimated in real time, for example, through
`the use of secondary scanning correlators whose time
`base is independent of a typical receiver’s detection-orb
`ented correlators. This error is subtracted from the
`
`detection-oriented correlator’s timing, thereby yielding
`a more accurate autocorrelation function.
`
`26 Claims, 15 Drawing Sheets
`
`30
`
`STAGE
`
`MIXING
`
` MICRO-
`PROCESSOR
`
`
`ANTENNA /
`LNA
`
`24
`
`LOCAL
`CODE
` CLOCK
`GENERATOR
`GENERATOR
`
`
`
`
`
`
`CODE
`CLOCK
`CODE
`GENERATOR
`GENERATOR
`
`
`
`
`
`
`
`SCANNING
`
`CORRELATORS
`
`PETITIONERS 1011-0001
`
`

`
`‘I0
`
`9
`
`OUTPUT
`
`VOLTAGE
`
`IN
`
`ARBITRARY
`
`UNITS
`
`0
`
`¥"5K"£—'N"§2'X
`
`‘
`
`13579111315171921232527293133353739
`
`DELAY TIME DIFFERENCE BETWEEN TRANSMITTED AND LOCAL CODE IN 0.05 CHIP UNITS
`
`F|G.1
`
`
`
`st10Iwas5661‘sz‘new111913,;'3'[1
`
`OS17‘Z0'I7‘S
`
`E3300-/\N|:lE|G
`
`PETITIONERS 101 1-0002
`
`

`
`
`
`?II1913cI'S'I1
`
`
`
`S66I‘SZ'-WW
`
`SIJ0Z199‘IS
`
`0sv‘z0v‘s
`
`
`
`
`LOCAL
`
`OSCILLATOR
`
`GENERATOR
`
`CODE CLOCK
`
`
`
`ATYPICAL CORRELATION RECEIVER BLOCK DIAGRAM
`
`22
`
`
`30
`
`32
`
`20
`
`
`STAGE MICROPROCESSOR
`
`
`
`ANTENNA / LNA
`
`MIXING
`
`BASEBAND CORRELATORS _
`
`
`
`LOCAL CODE
`
`OSCILLATOR
`
`MANAGEMENT
`
`FIG. 2
`
`E3300-/\N|:IE|G
`
`TIMING
`
`PETITIONERS 101 ’|-0003
`
`

`
`FROM
`WXER
`22
`
` BASEBAND
`
`MICROPl§(2)CESSOR
`
`
`
`111913d°S'fl
`
`09-p‘z()17‘gSIJO9wellsS661‘sz‘Jew
`
`€300/\N|:lE|G
`
`PETITIONERS 101 ’|-0004
`
`

`
`
`
`1H9113cI'S'I1
`
`E
`-
`:33
`Id
`§
`
`C/2
`E‘
`3
`9..
`5
`
`U.
`‘in-
`S
`uh
`U1
`®
`
`ith SAMPLE INTERVAL
`
`
`
`jth SAMPLE
`INTERVAL
`
`kth SAMPLE
`
`INTERVAL
`
`
`
`
`.
`8 910111213141516171I3192o2122232425262728 29303132 3334353637 6639
`
`6
`
`2
`
`4
`
`IN
`ARBITRARY
`UNITS
`
`OUTPUT 7
`VOLTAGE
`
`1o
`
`9
`
`8
`
`6
`
`5
`
`4
`
`3
`
`2
`1
`o
`
`DELAY TIME DIFFERENCE BETWEEN TRANSMITTED AND LOCAL CODE IN 0.1 CHIP UNITS
`
`F|G.4
`
`€300/\N|:lE|G
`
`PETITIONERS 101 ’|-0005
`
`

`
`'S°fl
`
`111919cI
`
`z
`gs:
`N
`‘co
`E
`
`/
`-x-at-"X.-x-at-)I—x
`—-x—-/<
`1 2 3 ,1 5 3 7 3 91311121314151317131930 122232425332723 29333132 33343533373339
`
`. X/X
`
`82 /
`
`x
`
`so
`
`/"
`*
`
`’
`
`:‘=’:’
`E
`c"V:
`
`G
`
`U1
`
`*5
`3n
`4:
`UI
`G
`
`PETITIONERS 101 ’|-0006
`
`
`
`(TRUE PEAK IS AT 21 TIME UNITS)
`
`DELAY TIME DIFFERENCE IN 0.1 CODE CHIP UNITS
`
`F|G.5
`
`OUTPUT
`
`VOLTAGE IN
`
`ARBITRARY
`UNITS
`
`1o
`
`8
`
`5
`
`4
`2
`0
`
`- 2
`
`— 4
`
`- 6
`
`- 8
`
`-10
`
`
`
`-€300/\N|:lE|G
`
`

`
`U.S. Patent
`
`Mar. 28, 1995
`
`Sheet 6 of 15
`
`5,402,450
`
`PETITIONERS 101 ’|-0007
`
`

`
`
`
`0S17‘Z0‘l7‘S913°L1°°‘ISS661‘sz‘mmguamtl-S-fl
`
`PETITIONERS 1011-0008
`
`14.000
`
`12.000
`
`110
`
`108
`
`
`
`10.000
`
`OUTPUT
`
`VOLTAGE IN
`
`ARBITRARY 8000
`UNlTS
`
`6.000
`
`4.000
`
`2.000
`
`DELAY TIME DIFFERENCE BETWEEN TRANSMITTED & LOCAL CODE IN
`
`0.1 CHIP UNIT
`
`F|G.7
`
`
`
`lvb(,UU/\N|:l:|U
`
`

`
`
`
`Wailgd'S'fl
`
`3 9
`
`3
`
`5,;
`id
`
`33UI
`
`3
`_,
`°°
`9»
`7-
`
`9'
`-9-
`S
`
`3c
`
`OUTPUT VOLTAGE
`IN ARBITRARY UNITS
`
`15.00
`
`5.00
`
`0.00
`
` 10.00
`
`11
`
`13
`
`15
`
`17
`
`19
`
`23
`
`25
`
`27
`
`29
`
`31
`
`33
`
`35
`
`37
`
`39
`
`500
`_
`
`_
`
`-10.00
`
`-15.00
`
`'
`
`I
`7
`
`|
`.
`/3 21
`[j
`U/
`
`.
`( FALSE PEAK INDICATED AT 22 TIME UNITS )
`
`K [KD/
`
`DELAY TIME DIFFERENCE IN 0.1 CODE CHIP UNITS
`F|G.8
`
`3300-/\N|:lE|G
`
`PETITIONERS 101 ’|-0009
`
`

`
`OUTPUT
`
`VOLTAGE IN
`
`ARBITRARY
`
`UNITS
`
`1.2
`
`1.0
`
`0
`
`€300/\N|:lE|G
`
`121 ’
`
`116
`
`I
`
`I
`
`II
`
`7‘
`
`1
`
`2
`
`3 4 5 6 7 8 910111213141516171819202122232425262728
`
`DELAY TIME DIFFERENCE BETVVEEN TRANSMITTED & LOCAL CODE IN
`
`0.1 CHIP UNIT
`
`F|G.9
`
`
`
`9661‘sz‘mm1II91B([‘S’[1
`
`SI.-I06199‘IS
`
`0sv‘z0v‘s
`
`PETITIONERS 101 ’|-0010
`
`

`
`
`
`
`
`
`
`SIJ0OI1991189661‘SE‘NWQIIGJBJ'S°fl
`
`//)~G~o\K
`
`._A
`
`120
`
`118
`
`
`
`ooooo,,I
`:IluIn.,,..
`141516171819201 3‘ 2728
`1234567 '.I
`L
`'‘-..,..-I' 542526
`
`-2
`
`DELAY TIME DIFFERENCE BETWEEN TRANSMITTED & LOCAL CODE IN
`
`122
`
`0.1 CHIPUNIT
`
`F|G.10
`
`OUTPUT
`VOLTAGE IN
`ARBITRARY
`
`UNITS
`
`1.0
`
`.8
`
`.6
`
`.4
`
`.2
`
`Z00-/\N|:|E|G
`
`PETITIONERS 1011-0011
`
`

`
`I
`
`so
`
`32
`
`
`
`
`
`
`MICRO-
`I (BASEBAND) CORRELATORS
`PROCESSOR
`
`
`
`
`CODE
`CLOCK
`GENERATOR
`
`
`
`LOCAL
`CODE
`
`GENERATOR
`
` ANTENNA I
`
`LNA
`
`LOCAL
`OSCILLATOR
`
`24
`
`
`
`05-p‘z()-p‘gSIJ0IImus$661‘sz‘mm11191125’S‘f1
`
`
`
`
`CONTROL
`
`
`
`
`CODE
`LOCAL
`CLOCK
`CODE
`GENERATOR
`GENERATOR
`
`
`
`
`
`
`CONTROL
`
`
`
`
`SCANNING
`CORRELATORS
`
`FlG.11
`
`€300/\N|:lE|G
`
`PETITIONERS 101 1-0012
`
`

`
`U.S. Patent
`
`Mar. 28, 1995
`
`Sheet 12 of 15
`
`5,402,450
`
`FIG.12
`
`
`
`
`
`TIME~——_.25MINUTES(300-5SEC.MEASUREMENTS)
`
`PETITIONERS 101 ’|-0013
`
`

`
`U.S. Patent
`
`Mar. 28, 1995
`
`Sheet 13 of 15
`
`5,402,450
`
`wn_D._..._n=2<
`
`PETITIONERS 101 ’|-0014
`
`

`
`U.S. Patent
`
`Mar. 28, 1995
`
`Sheet 14 of 15
`
`5,402,450
`
`150
`
`152
`
`154
`
`READ CORRELATORS
`E,P_I,L + E‘ IN I
`
`READ CORRELATORS
`
`Pa IN 0
`
`CLOSE TRACKING LOOP
`
`usme P1 + F0
`
`
`
`156
`
`LOCKED LOOP +
`
`CORRELATORS E - L
`
`CLOSE COST TRACKING
`
`LOOP USING STD DELAY
`
`158
`
`160
`
`152
`
`
`
`
`CLOSE EXTRA CODE TRACKING
`LOOP usme P; + E‘:
`LOOP oruves
`KPI - 5' = ¢’
`
`
`
`
`COMPARE TIME DIFFERENCE
`BETWEEN PI + E‘ FROM STEP158
`
`T0 N0 - MULTIPATH CASE:
`ESTIMATE THE MULTIPATH
`
`
`
`WITH MULTIPATH ESTIMATE FROM
`STEP 160:
`
`
`COMBINE CODE PHASE
`MEASUREMENT FROM STOP156
`
`
`
`
`
`PRODUCE MULTIPATH FREE
`
`PSEUDO RANGE
`
`
`
`
`
`
`
`FIG.14
`
`PETITIONERS 1011-0015
`
`

`
`U.S. Patent
`
`Mar. 28, 1995
`
`Sheet 15 of 15
`
`5,402,450
`
`CLOSE CODE TRACKING
`
`LOOP IN NORMAL WAY
`
`WITH E-L-P GATE
`
`METHOD AND DELAY
`
`LOCKED LOOP AND
`
`CALCULATE Tp
`
`MEASURE E’ AND L‘ AT
`
`PRECISE TIME
`
`INTERVALS GIVEN BY
`
`TnI2 AND ESTIMATE OF Tp
`FROM STEP 1
`
`LOOK UP CORRECTION
`
`TO Tp IN TABLE AND
`APPLY TO Tp TO GET Tp',
`CORRECTED TIME OF
`
`ARRIVAL
`
`MULTIPATH FREE MEASUREMENT
`
`FIG. 15
`
`PETITIONERS 101 ’|-0016
`
`

`
`1
`
`SIGNAL TIIVIING SYNCHRONIZER
`
`FIELD OF THE INVENTION
`
`5,402,450
`
`2
`the satellite-based GPS system and is used to calculate
`the “pseudo-range”, which is the first-cut estimate of
`the distance between the receiver and a GPS satellite.
`
`The present invention relates generally to signal tim-
`ing synchronization for radio receivers and, more spe-
`cifically, to methods and apparatus for improving the
`accuracy of measuring the time-of-arrival of an incom-
`ing signal in, by way of example, a spread spectrum
`correlation receiver by reducing the adverse effects of
`multipath signals on the measurement. The invention
`also relates generally to any communications system
`that may suffer from multipath effects.
`BACKGROUND OF THE INVENTION
`
`The Global Positioning System (GPS) as now being
`implemented utilizes a number of satellites in precise
`orbits that broadcast navigational information that may
`be used by anyone with a proper GPS satellite receiver.
`This so-called navigational information is also useful to
`surveyors and the like because it can provide accurate
`position information concerning any point on the globe.
`Each satellite in the GPS system broadcasts with the
`same carrier frequency and each broadcast signal in-
`cludes an individual code that serves to identify the
`particular satellite. The codes are generally long and are
`made up of a pattern of 1’s and O’s that repeats over long
`time periods relative to the data rate. A complete study
`and report on GPS signals has been published by J. J.
`Spilker, Jr., “GPS Signal Structure and Performance
`Characteristics”, Navigation, 1980.
`In addition,
`the
`basic methods and techniques of GPS are also repre-
`sented by J. J. Spilker, Jr. in his book “Digital Commu-
`nications by Satellite”, Prentice Hall, Inc. 1977.
`Radio receivers for the GPS navigation data bit
`stream are commonly referred to as correlation receiv-
`ers and examples of such receivers are described in U.S.
`Pat. No. 4,754,465 (“465 patent”) to Charles Trimble,
`and assigned to the assignee of the present application.
`The disclosure of the ’465 patent is incorporated herein
`by reference. Correlation receivers are typically em-
`ployed because they are designed for situations typi-
`cally encountered in satellite broadcasting where the
`strength of the GPS signal is quite weak compared to
`the noise level. The relative signal level is low at least in
`part because the receiver must use a wide-angle an-
`tenna, which has very low gain, due to the system con-
`straint of having to listen to a number of satellites that
`might be located anywhere in the sky.
`In order to boost the weak signal without also ampli-
`fying the noise, it is the practice to use spread spectrum
`modulation in GPS satellite systems. The spread spec-
`trum technique modulates the satellite transmission by
`the individual satellite identification code, and this has
`the effect of spreading the satellite signal over a band-
`width that is determined by the reciprocal of the pulse
`width. Conversely, the receiver multiplies the signal
`received on the ground by a replica of the individual
`satellite code, and this kind of demodulation is generally
`known as correlation. A typical spread spectrum re-
`ceiver is described in U.S. Pat. No. 4,965,759 to Uchida
`et al. Spread spectrum systems in general are more fully
`described by R. C. Dixon, “Spread Spectrum Systems”,
`J. Wiley & Sons, Inc., 1976.
`A particular advantage of using spread spectrum
`modulation is that itallows the time of arrival of the
`transmitted signal to be determined by the receiver.
`This time-of-arrival measurement is the cornerstone of
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`45
`
`50
`
`55
`
`65
`
`Determining the time of arrival of a signal requires
`the recognition of at least a portion of the incoming
`signal and a comparison of the time of arrival of the
`incoming signal with the time that it was known to have
`been transmitted. This measurement is made by aligning
`the incoming code and the local code using a code
`tracking loop. Such code tracking loop adjusts the
`delay time of one code with respect to the other and
`tries to minimize the time difference between the incom-
`ing and local codes. The time-of-arrival measurement is
`then based on the current local code delay time as mea-
`sured by the local clock relative to the known time
`when the incoming signal was transmitted. When this
`delay time is multiplied by the signal propagation speed,
`which is assumed to be the speed of light, the path
`length is derived. As mentioned above, this path length
`is referred to as the pseudo-range because the local
`clock is not in perfect synchronism with the transmitter
`clock. Further, any errors affecting the code tracking
`loop will also directly affect the time-of-arrival mea-
`surement.
`
`As mentioned above, a correlation receiver typically
`demodulates the spread spectrum signal by multiplying
`the incoming signal with a locally generated replica of
`the spread spectrum code. The operation of multiplying
`the local code and the incoming signal to produce mea-
`surable signal power at the receiver requires that the
`local code and the incoming code be aligned with each
`other to be within one cycle of the code clocking rate.
`This one cycle at the clocking rate is also referred to as
`a “chip”. If the two codes are within one chip of each
`other, some measurable signal power will be observed
`at the output of the receiver correlator, and the closer
`the two codes are aligned, the greater is the power that
`will be observed. The relationship of the delay time
`between the two codes to the amount of signal power
`observed at the output of the correlation operation is
`called the autocorrelation function (AF). It will be
`appreciated that peak received power will be detected
`when the two codes are perfectly aligned in time. The
`autocorrelation function is generally observed as a se-
`ries of correlation spikes output from a matched filter in
`the correlator circuit. This type of receiver is com-
`monly known as a “matched filter” receiver. By keep-
`ing the local code phase synchronous with the code
`phase of the received signal, optimum detection of the
`modulation is accomplished, and from this optimized
`detection the time-of-arrival of the signal is determined.
`Through subsequent calculations, the latitude,
`longi-
`tude, and height of the receiver can be determined.
`The ideal autocorrelation function between two
`spread spectrum codes is shown by the spike 10 in FIG.
`1. This correlation spike represents the voltage output
`of a correlating receiver as a function of the relative
`shift in time between the two correlating codes. There-
`fore, the maximum voltage at the output of the correla-
`tor, as shown at the peak 104: in FIG. 1 will be ideally
`detected when the two codes are in perfect alignment.
`The true autocorrelation function in a real receiver is,
`however, somewhat different from the ideal and is
`shown by the curve 12 in FIG. 1. As will be noted, the
`peak of the curve 12 is not sharp, and the leading and
`trailing slopes from the peak are not straight. This
`rounding of the ideal triangular shape is caused by the
`
`’
`
`PETITIONERS 101 ’|-0017
`
`

`
`3
`use of fmite-bandwidth filters in the receiver prior to
`correlation. This rounding has been found to be signifi-
`cant in determining and compensating for multipath
`effects, described in detail below.
`The basic elements of a typical correlation receiver
`are shown in FIG. 2. The incoming spread spectrum
`signal is received by an antenna and low noise amplifier
`20 and is mixed to baseband in a mixing stage 22 by a
`locally generated carrier signal from a local oscillator
`24. This mixing is performed in mixing stage 22 so that
`quadrature signals (Q), as well as in-phase signals (1), are
`available at baseband to facilitate carrier tracking. Both
`the quadrature and in-phase baseband signals (Q and I)
`after the mixing are still binary phase-shift keyed
`(BPSK) modulated by the spreading code and by any
`lower modulation rate information that might be in-
`cluded in the satellite transmitted signal. For example,
`in the Global Positioning System the baseband signals
`contain a coarse/acquisition (C/A) spreading code hav-
`ing a clock rate of 1.023 MHz and a precision (P)
`spreading code having a clock rate of 10.23 MHz, as
`well as a lower rate 50 baud data bit stream. The local
`oscillator 24 output signal also drives a code clock gen-
`erator 26 having an output signal fed to a local code
`generator 28 that generates the local code signal fed to
`the correlators 30. The correlators 30 then do correla-
`tion calculations between the local code and the base-
`
`band signals and the correlation output signal is fed to a
`microprocessor 32 for performing the necessary time-
`of-arrival calculations. The microprocessor 32 also con-
`trols the timing of the local oscillator 24 and code clock
`generator 26, as well as telling the local code generator
`28 which individual code to generate.
`Determination of the optimum local code delay time
`requires a feedback technique that either minimizes
`some error signal based on the difference between the
`local code and the incoming code or that maximizes the
`autocorrelation function (AF). Because the signals are
`so weak, it has generally been the case to elect to use a
`differencing technique involving the use of “early-late”
`gates instead of looking for a maximum of the autocor-
`relation function. This early-late gate approach relies
`upon the fact that the ideal, uncorrupted autocorrela-
`tion function is symmetrical around its peak, the peak
`being representative of the point in time where the
`codes are perfectly synchronized. Correlation receivers
`then have heretofore attempted to locate the peak in
`time of the autocorrelation function, because this point
`represents perfect code alignment.
`Typically, measurement of the degree of correlation
`between the incoming code and the local code is per-
`formed at three distinct points on the autocorrelation
`function. These correlation points are:
`the punctual (P) point, where voltage output is maxi-
`mized due to perfect alignment of the two codes;
`the early (E) point, which represents voltage output
`when the two codes are approaching alignment
`and are about § chip out of alignment, i.e., where
`the local code is advanced by 5 chip with respect to
`the incoming code; and
`the late (L) point, which represents voltage output
`when the two codes are receding from alignment,
`i.e., where the local code is delayed by 5 chip with
`respect to the incoming code. Only one time offset
`is used to shift all three correlators synchronously.
`In other words, the time shifts of the correlators
`are not independent of one another.
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`45
`
`50
`
`55
`
`65
`
`5,402,450
`
`4
`The early-late gate method begins by measuring the
`autocorrelation function and establishing a sample volt-
`age level, and in doing this the receiver samples the
`autocorrelation function at the beginning of the -1 bit
`time interval, shown as the ith sample interval in FIG. 4.
`The microprocessor easily can keep track of successive
`samples and so samples that are one chip interval apart
`are subtracted from one another. The later time sample,
`the “late” gate, is shown as the k”' interval in FIG. 4 and
`is subtracted from the early time sample, the “early”
`gate, resulting in a well-known control function that
`can be used to drive the local code chip time delay so
`that the local code is synchronized with the incoming
`code. This early-late gate error function is shown in
`FIG. 5 at curve 80 relative to the autocorrelation func-
`tion shown as curve 82. As the early-late gate sampling
`function 80 progresses in time, i.e. is shifted in time to
`the right on the graph of FIG. 5, the difference between
`the early and late voltages diminishes, and when that
`difference (the error function) equals zero volts the
`peak of the autocorrelation function is found and the
`local code is synchronized with the incoming code.
`This occurs at 21 time units in the specific example
`shown in FIG. 5.
`
`Thus, the early-late gate method in effect drives the
`error voltage to zero, measures the offset in code phase
`relative to a local clock, and derives an estimate of the
`time-of-arrival of the incoming signal. The punctual
`correlator output signal serves as a check on the steer-
`ing provided by the early-late gate correlators, and in an
`interference-free environment this signal can confirm
`the early-late gate derived position of the peak. Never-
`theless,
`the punctual correlator cannot provide any
`steering information on which way to shift the local
`code. With no extraneous interfering signals, this early-
`late gate method works well.
`The correlators 30 that are typically employed in a
`correlating receiver such as that of FIG. 2 are shown in
`more detail in FIG. 3. In the correlators of FIG. 3, the
`baseband signals from the mixing stage 22 are fed to
`respective sets of three mixers corresponding to early,
`punctual, and late. More specifically the in-phase base-
`band signals (I) on line 40 are fed to mixers 42, 44, 46
`that also receive the local code from the local code
`generator 28. This local code is fed on line 48 to a dis-
`tributor unit 50 that might consist of a shift register and
`that operates to sequentially distribute the local code
`input on line 48 to the three mixers 42, 44, 46. Similarly,
`the quadrature baseband signals (Q) on line 52 are fed to
`three mixers 54, 56, 58 corresponding to early, punctual,
`and late, a.nd these three mixers 54, 56, 58, respectively
`also receive the local code from the distributor unit 50.
`The extent of coincidence between the received sig-
`nal (I and Q) and the local code in the three states as
`determined by the mixers 42, 44, 46 and 54, 56, 58 is
`accumulated over a number of cycles in six accumula-
`tors 60, 62, 64, 66, 68, 70 that are connected respectively
`to the above-noted mixers. Thus, early, punctual, and
`late data for both the in-phase and quadrature signals
`are fed to the microprocessor 32 where the appropriate
`timing calculations are performed.
`The information from the early and late correlators is
`combined by the microprocessor 32 to generate a delay
`locked loop tracking signal that is used to close the code
`tracking loop. The information used to close the code
`tracking loop is taken from early and late comparisons
`of the local and incoming codes; therefore, the perfor-
`mance of this code tracking determines the accuracy of
`
`PETITIONERS 101 ’|-0018
`
`

`
`5,402,450
`
`6
`either add to the desired direct signal or subtract from
`it.
`
`5
`the time-of-arrival measurement, which is used to gen-
`erate the pseudo-range. As described hereinabove, the
`time-of-arrival measurement is typically performed by
`comparing in time the local code, which ideally is track-
`ing in perfect alignment with the transmitted code, with
`the time reference of the receiver.
`As indicated below, such prior attempts to locate the
`peak in time of the autocorrelation function have not
`yielded entirely satisfactory results. Certain types of
`interfering signals can distort the autocorrelation func-
`tion in a way that transfers errors into the tracking loop.
`One troublesome kind of interfering signal is known
`as multipath. Multipath refers to the phenomenon in
`radio wave propagation wherein a receiver system is
`able to collect a so-called primary signal, representing
`the direct path of radio wave propagation between the
`source and the receiver, and also a plurality of second-
`ary delayed versions of the direct signal, representing
`reflections of the direct signal from objects adjacent the
`direct path. This phenomenon is particularly acute in
`receiver systems with large coverage-area antennas,
`such as are commonly found in GPS systems. The mag-
`nitude of multipath error induced in GPS systems has
`been reported by J. M. Tranquilla et a1., “GPS Multi-
`path Field Observations at Land and Water Sites”,
`Navigation Journal of the Institute of Navigation, Vol.
`37, No. 4, 1990-91.
`Signal reception at moving vehicles suffers from this
`phenomenon to an even greater extent. Multipath ad-
`versely affects FM reception, cellular mobile telephony,
`and other voice/data radio systems, whether or not
`they use spread spectrum digital modulations. In mini-
`mizing the adverse affects of multipath, the present
`invention is not limited to GPS systems and is com-
`pletely applicable to these other applications, particu-
`larly where a spread spectrum technique is employed.
`An example of a typical receiver system with multi-
`path signals is shown in FIG. 6, in which a GPS patch
`antenna 90 receives not only direct path signals 92 from
`the satellite but also multipath reflected signals 94, 96.
`The multipath signals 94, 96 represent the signal from
`the satellite being reflected by a building 98 or some
`other large object 100, respectively, in the vicinity of
`the antenna 90.
`
`These secondary signals 94, 96 have been found to
`have several important characteristics in relation to the
`primary signal 92. For example, the secondary signals
`always have a delayed time-of-arrival compared to the
`primary signal 92, because the secondary signals 94, 96
`travel a slightly longer path than the primary signal 92.
`Furthermore, the respective amplitudes of the second-
`ary signals 94, 96 are nearly always less than that of the
`primary signal 92, because the reflections are specular
`and attenuate the signal. In addition, the sense of polar-
`ization is reversed by the reflection, and the receiving
`antenna is not as sensitive to these cross-polarized multi-
`path signals as to the primary signal. For correlation
`receivers using digital modulation, moreover, multipath
`code phase signals with delays greater than one chip are
`completely uncorrelated, and so can be ignored. Fi-
`nally, the multipath signal distance, that is, the differen-
`tial path length variation from the direct signal path,
`varies over the wavelength of both the carrier phase
`and the code phase. For example, in GPS the carrier
`phase wavelength is 19 cm (}r=c/1575 MHz, where
`c=the speed of light), but the code phase wavelength is
`much longer, because the code frequency is as low as
`1.023 MHz. As a result, the multipath carrier signal can
`
`These secondary signals have a deleterious effect on
`the accuracy of the correlation receiver. For example,
`because multipath signals are replicas of the incoming
`direct signal and its code, and the principal or primary
`autocorrelation function is generated by examining the
`incoming direct signal and the locally generated signal,
`each multipath signal generates its own secondary auto-
`correlation function with respect to the locally gener-
`ated code. The secondary autocorrelation functions will
`always have lower maximum amplitudes than the corre-
`lated primary signal and will always be delayed relative
`to the primary signal. This phenomenon is illustrated by
`way of example in FIG. 7. The secondary or multipath
`signals generate a series of secondary autocorrelation
`functions 102, 104 and 106 that are smaller than and are
`delayed relative to the primary autocorrelation function
`108. The primary and secondary autocorrelation func-
`tions add by superposition, and the resulting net auto-
`correlation function is shown at 110 in FIG. 7. It should
`be noted that because the carrier multipath signals can
`also subtract from the primary signal rather dramatic
`distortions can be induced in the direct signal autocorre-
`lation function. Also, because of the rounded autocorre-
`lation function curve due to the finite-bandwidth filters,
`the indicated peak of the composite autocorrelation
`function may move in time. Therefore, when using the
`conventional early-late gate method, such distortions
`translate directly into errors in the steering control or
`error function for the delay-locked loop and result in
`erroneous estimates of local code phase and therefore
`erroneous estimates of pseudo-range.
`The adverse effects caused by these multipath signals
`is readily appreciated by comparing curves 80’ and 82’
`of FIG. 8 with the respective curves 80 and 82 of FIG.
`5. Such comparison will reveal that in FIG. 5 the true
`autocorrelation frmction peak occurs when the error is
`zero at 21 time units, whereas in FIG. 8 the peak is
`falsely indicated to occur by the zero crossing of the
`error function 80' at 22 time units. This error is intro-
`
`duced because the composite autocorrelation function
`has experienced a change in the apparent time that a
`predetermined and fixed amplitude level is intercepted
`by the early-late gate correlation method.
`FIG. 9 illustrates a composite autocorrelation func-
`tion with additive multipath distortions. In FIG. 9, the
`uncorrupted autocorrelation function has its arbitrary
`amplitude, early- and late-gate times identified at E and
`L, respectively. The delayed multipath signals cause the
`composite autocorrelation function 112 to have a larger
`voltage level at the true late-gate time point L (17 time
`units) as shown at 121. It takes a little time (about one
`time unit) for the composite autocorrelation function
`amplitude 112 to fall to the prescribed voltage at level
`L’ normally associated with the true late time point L.
`As can be seen from FIG. 9, the corrupted autocorrela-
`tion function 112 falls to the predetermined level 0.6 at
`18 time units. The timing error At is the time difference
`between L and L’ which in this example is one time unit.
`While the magnitude of this error is small, it is still
`significant, and can therefore cause an error in the de-
`termination of the true time of signal arrival.
`The absolute positioning accuracy of a GPS naviga-
`tion solution is essentially lirnited by systemic errors in
`the satellites and other signal degrading influences such
`as doppler offsets, ionospheric effects and other causes.
`The GPS specification is such that users heretofore can
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`45
`
`50
`
`55
`
`65
`
`PETITIONERS 101 ’|-0019
`
`

`
`7
`get position fixes to a three-dimensional accuracy of
`about 100 meters. The magnitude of the multipath effect
`has a maximum of around 10 meters for coarse acquisi-
`tion (C/A) code and 5 meters for precision (P) code.
`Therefore, in the navigation mode the multipath error is
`largely masked by other system errors. Differential and
`survey applications of the GPS remove satellite-based
`systemic effects by differencing between two co-
`located receivers, in which two receivers are located
`located at any arbitrary distance apart in order to get
`another set of equations relating to satellite uncertain-
`ties. In these applications the largest error source affect-
`ing the pseudo-range, after receiver and ionospheric
`effects have been removed, is caused by multipath sig-
`nals. The objects that are the source of multipath error
`are dependent on the environment in which the receiver
`has to operate and are therefore difficult to predict.
`Previous attempts at substantially reducing the adverse
`effects of multipath by altering the antenna characteris-
`tics of a receiver have not provided a consistent solution
`and often are very expensive.
`The differential survey method achieves a relatively
`high accuracy from measurements based on the carrier
`phase of the received signal. Because the wavelengths
`of the two GPS frequencies are small, 19 and 24 cm,
`respectively, accuracies of much less than 1 m are possi-
`ble. The problem with a time-of-arrival measurement
`based on carrier phase is that the time-of-arrival of each
`carrier phase cycle is ambiguous, that is, one carrier
`cycle cannot be distinguished from any other carrier
`cycle. Previously proposed post-processing techniques
`typically have required that the two receivers remain
`stationary for about one hour, so that this carrier cycle
`ambiguity can be resolved. Thus, multipath-induced
`errors of 3 to 10 meters are a major stumbling block to
`decreasing measurement time.
`It has also been proposed to use the pseudo-range to
`resolve the carrier phase ambiguity, thereby allowing
`the ambiguity to be resolved much faster. The ability of
`a receiver to utilize pseudo-range to resolve the carrier
`cycle ambiguity, however, is based on the quality of the
`pseudo-range measurement. Unfortunately, multipath
`signals bias the pseudo-range so that the wrong carrier
`cycle is often chosen when resolving the carrier cycle
`ambiguity. One possible technique to prevent choosing
`the wrong carrier cycle is to rely on the fact that the
`multipath may average out over a period of time. This
`technique, however, requires the multipath magnitude
`and direction to change substantially during this period
`of time. That is, it relies on extensive satellite or user
`movement.
`
`Other proposed techniques to compensate multipath
`effects in receivers rely on more commonly understood
`channel equalization techniques, however, none of
`these techniques makes use of measurements from the
`autocorrelation function. Instead, these techniques rely
`on a less accurate form of charmel characterization. For
`example, U.S. Pat. No. 4,829,543 describes a technique
`for correcting the multipath effects applicable to digital
`data demodulation in a Time-Division Multiple-‘Access
`('I'DMA) data transmission system. TDMA systems
`operate with regularly repeated short-duration bursts of
`data and differ in that respect from the Global Position-
`ing System, however, the method of the above-men-
`tioned patent discloses a correlation receiver for this
`burst start-up sequence. As described in the above-men-
`tioned patent, a known sequence of data bits is transmit-
`ted as part of a TDMA burst preamble from which the
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`5,402,450
`
`8
`receiver derives an estimate of the correct time to start
`demodulation of the data. The system then re-estimates
`the time to start demodulation in the TDMA system at
`each time slot of TDMA transmission. While that ap-
`proach has merit in improving demodulation by obtain-
`ing a new and better estimate of the carrier phase timing
`for each time slot of transmission, it does not address
`multipath effects on the burst preamble itself.
`It can be recognized that removing multipath effects
`would avoid the delays inherent in the need for averag-
`ing periods and would essentially allow signal ambigui-
`ties to be reso