a2, United States Patent
`US 6,563,892 B1
`(10) Patent No.:
`Haartsenetal.
`(45) Date of Patent:
`May13, 2003
`
`
`US006563892B1
`
`(54) METHOD AND SYSTEM FOR DETECTION
`OF BINARY INFORMATIONIN THE
`PRESENCE OF SLOWLY VARYING
`DISTURBANCES
`
`(75)
`
`Inventors: Jacobus Haartsen, Borne (NL); Paul
`Dent, Pittsboro, NC (US)
`
`(73) Assignee: Telefonaktiebolaget LM Ericsson,
`Stockholm (SE)
`:
`a
`:
`Subjectto any disclaimer, the term ofthis
`patent is extended or adjusted under 35
`US.C, 154(b) by 0 days.
`
`:
`.
`(*) Notice:
`
`(21) Appl. No.: 09/332,955
`
`Filed:
`Jun. 15, 1999
`(22)
`7
`Int. C0cc eectessesneceeetenseneees HO4B 1/10
`(51)
`.. 375/350; 375/346; 375/348
`(52) US. Ch we.
`
`
`(58) Field of Search oo... eee 375/350, 348,
`375/346, 341, 229, 230, 231, 232, 233,
`234; 708/301, 319, 320, 322, 323
`References Cited
`U.S. PATENT DOCUMENTS
`
`(56)
`
`5,142,552 A *
`5,241,702 A
`5,285,480 A
`5,659,583 A *
`6,226,323 B1 *
`
`.........-....-- 375/232
`8/1992 Tzenget al.
`8/1993 Dent oo...ccceeeeeeee 455/278.1
`2/1994 Chennakeshuetal.
`..... 375/101
`8/1997 Lane voces 375/346
`5/2001 Tametal. we 375/233
`
`6,301,298 B1 * 10/2001 Kuntz et al. oo... 375/232
`6,314,147 B1 * 11/2001 Liang et al. «0.0... 375/346
`OTHER PUBLICATIONS
`Carley, L.R., et al. “A Pipelined 16-State Generalized
`Viterbi Detector”, IEEE Transactions 0 Magnetics, vol. 34,
`No. 1, Jan. 1998, pp. 181-186, XP002132530.
`Moehrmann, K.H., et al., “Ein Wechselstromgekoppeltes
`Analogwert—Schieberegister”, Nachrichtentechnische Fach-
`berichte, De, Vde, Verlag, Berlin, Mar. 1969, pp. 353-356
`XP000763929.
`Haartsen, J., “Bluetooth—The Universal RadioInterface for
`Ad Hoc, Wireless Connectivity”, Ericsson Review, Tele-
`communications Technology Journal No. 3, 1998, pp.
`110-117.
`.
`.
`* cited by examiner
`
`Primary Examiner—Mohammad H. Ghayour
`(74) Attorney, Agent, or Firn—Burns, Doane, Swecker &
`Matthis
`
`ABSTRACT
`©”
`A radiocommunication system is described in which DC
`offset and other slowly varying disturbances whichthat may
`impact a signal are suppressed. Exemplary embodiments of
`the present invention combine a difference circuit, e.g., a
`FIR filter, with a maximum likelihood sequence estimator,
`e.g., a Viterbi detector, to implement suppression techniques
`on binary signals.
`
`31 Claims, 11 Drawing Sheets
`
`
`
`VITERBIEQUALIZER
`
`FIR FILTER,
`
`
`
`INPUT
`SIGNAL
`
`OUTPUT
`SIGNAL
`
`
`
`OPTIMALIZATION
`TEST VECTORS
`\
`
`
`
`
`10
`
`SONY Exhibit 1006 - 0001
`
`SONY Exhibit 1006 - 0001
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 1 of 11
`
`US 6,563,892 B1
`
`
`
`|MOWLWLS
`
`ISWG
`
`ISVG
`
`NOILVIS
`
`GOtlDOT00}
`
`
`
`IWOHdITULTie0NOt
`
`FHALOOMIHILIMS
`
`
`
`LA
`
`(O)fBly
`
`(LUVYOldd)
`
`
`
`volt008}=O
`
`NOILVLS
`
`ISVG
`
`
`
`OFHILIMSI1T8Nd
`
`
`
`MYOMLINTNOHATTAL
`
`SONY Exhibit 1006 - 0002
`
`SONY Exhibit 1006 - 0002
`
`
`
`
`
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 2 of 11
`
`US 6,563,892 B1
`
`240
`
`220
`
`
`DETECTOR
`WHITENING
`FILTER
`
`
`VITERBI
`
`
`
`
`
`CHANNEL
`
`ESTIMATOR
`
`
`
`Fig. (b)
`(PRIOR ART)
`
`SONY Exhibit 1006 - 0003
`
`SONY Exhibit 1006 - 0003
`
`

`

`Sheet 3 of 11
`
`US 6,563,892 B1
`
`U.S. Patent
`
`May13, 2003
`
`-—- swTHRESHOLD
`
`SONY Exhibit 1006 - 0004
`
`SONY Exhibit 1006 - 0004
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 4 of 11
`
`US 6,563,892 B1
`
`300
`
`f
`
`310
`
`
`
`320
`
`Fig. 3(a)
`
`SONY Exhibit 1006 - 0005
`
`SONY Exhibit 1006 - 0005
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 5 of 11
`
`US 6,563,892 B1
`
`toLe
`
`380
`
`a
`
`xO
`ee
`
`360
`
`390
`
`a
`
`370
`
`360
`
`Fig. 3(b)
`
`SONY Exhibit 1006 - 0006
`
`SONY Exhibit 1006 - 0006
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 6 of 11
`
`US 6,563,892 B1
`
`394x
`CATIA
`
`392x“
`
`396a
`
`Fig. 3(c)
`
`SONY Exhibit 1006 - 0007
`
`SONY Exhibit 1006 - 0007
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 7 of 11
`
`US 6,563,892 B1
`
`WLLLLU
`
`SONY Exhibit 1006 - 0008
`
`SONY Exhibit 1006 - 0008
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 8 of 11
`
`US 6,563,892 B1
`
`yl y2 yF ya yo yé
`09
`
`07
`
`A
`
`E(0/0)
`
`At]
`
`kt2
`
`~.
`
`oN EVN), ©
`\
`uo
`a
`so
`UL EVO) a
`
`7
`‘.
`><
`yoo
`
`EUI/1)
`
`ao i a
`SY
`7
`~~
`Yo Sy
`
`‘. a
`
`~.
`
`SONY Exhibit 1006 - 0009
`
`SONY Exhibit 1006 - 0009
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 9 of 11
`
`US 6,563,892 B1
`
`FIR FILTER
`
`VITERBI EQUALIZER
`
`
`INPUT.
`
`SIGNAL
`
`60
`62
`
`OUTPUT
`SIGNAL
`
`VITERBIEQUALIZER
`
`
`
`INPUT
`SIGNAL
`
`OUTPUT
`SIGNAL
`
`
`OPTIMAL IZATION
`TEST VECTORS
`
`
`Fig. 7
`
`10
`
`SONY Exhibit 1006 - 0010
`
`SONY Exhibit 1006 - 0010
`
`

`

`
`
`(4)PINdLnodWVd024
`
`U.S. Patent
`
`May13, 2003
`
`-|||tyhy
`
`(Z-4)X98p8 (f-¥JX
`
`
`
` TTTs7CTTTTT“7P||Polo}ybi—<roe::Longe--4Loy---4Lee4
`
`Sheet 10 of 11
`
`US 6,563,892 B1
`
`ey+|
`
`8bly
`
`SONY Exhibit 1006 - 0011
`
`SONY Exhibit 1006 - 0011
`
`

`

`U.S. Patent
`
`May13, 2003
`
`Sheet 11 of 11
`
`US 6,563,892 B1
`
`901
`
`
`
`
`Receive a signal comprising a
`Wanted signal and aninterfering
`signal
`
`
`
`Filter the received signalto reject
`Spectral components of the interfering
`
`Signal, thereby causing intersymbol
`interference
`
`903
`
`
`
`905
`
`907
`
`
`
`
`
`Determine a number of channel
`Coefficients descriptive of the
`Intersymbol interference caused
`By the filtering
`
`
`
`
`
`Decode the wanted information using
`an equalizer adapted using the channel
`
`
`Coefficients to compensatefor the
`
`Intersymbol interference
`
`
`FIG. 9
`
`SONY Exhibit 1006 - 0012
`
`SONY Exhibit 1006 - 0012
`
`

`

`US 6,563,892 B1
`
`1
`METHOD AND SYSTEM FOR DETECTION
`OF BINARY INFORMATIONIN THE
`PRESENCE OF SLOWLY VARYING
`DISTURBANCES
`
`BACKGROUND
`
`The present invention pertains to systems and methods
`involved in radiocommunication systems and, more
`particularly, to systems that employ binary signal streams of
`known amplitude which are disturbed by, for example, DC
`offsets, drifts, and other slowly changing disturbances super-
`imposed on the desired binary signal. The techniques
`described herein are particularly well-suited for the detec-
`tion of binary FM or binary FSK modulated signals in the
`presence of such disturbing signals, but can also be used in
`conjunction with other types of modulation.
`The cellular telephone industry has made phenomenal
`strides in commercial operations in the United States as well
`as the rest of the world. Growth in major metropolitan areas
`has far exceeded expectations and is rapidly outstripping
`system capacity. If this trend continues, the effects of this
`industry’s growth will soon reach even the smallest markets.
`Innovative solutions are required to meet these increasing
`capacity needs as well as to maintain high quality service
`and avoid rising prices.
`FIG. 1 illustrates an example of a conventional cellular
`radio communication system 100 in which the present
`invention can be implemented. The radio communication
`system 100 includesa plurality of radio base stations 170a—n
`connected to a plurality of corresponding antennas 130a—7.
`The radio base stations 170a—n in conjunction with the
`antennas 130a—n communicate with a plurality of mobile
`terminals (e.g. terminals 120a, 1206 and 120m) within a
`plurality of cells 110a—m. Communication from a base
`station to a mobile terminal is referred to as the downlink,
`whereas communication from a mobile terminal to the base
`
`station is referred to as the uplink.
`The base stations are connected to a mobile telephone
`switching office (MSC) 150. Among other tasks, the MSC
`coordinates the activities of the base stations, such as during
`the handoff of a mobile terminal from one cell to another.
`
`The MSC,in turn, can be connected to a public switched
`telephone network 160, which services various communi-
`cation devices 180a, 180b and 180c.
`In conventional cellular radiocommunication systems
`such as that illustrated in FIG. 1, the signal transmitted over
`the air interface does not travel along a single, straight path.
`Instead,
`the radiated energy reflects and travels in many
`directions so that different portions of the radiated energy
`arrive at the receiver(i.e., either that of terminals 120 or base
`stations 170) at different times. As a result,
`the receiver
`receives a distorted signal that is very different from the
`original signal. This distortion problem, which is commonly
`referred to as multipath fading, can be viewed as a smearing
`of the transmitted pulses.
`the effects of the radio
`In such conventional systems,
`channel are measured and taken into accountin the receiver
`
`when attempting to correctly determine the originally trans-
`mitted information. Channel estimates are calculated based
`
`upon known information which is periodically transmitted
`over the radio channel to the receiver. Since radio channels
`
`may change rapidly, e.g., due to movement of the terminals
`120, the channel estimate can be regularly updated.
`Channel estimation can be used in conjunction with an
`application of the Viterbi algorithm to determine the origi-
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`2
`nally transmitted information as shown in FIG. 1(6).
`Therein,
`the received signal is used to produce channel
`estimates at block 200. The channel estimates are provided
`to the Viterbi detector 220, wherein they are employed to
`determine metrics associated with the likelihood of various
`state transitions. Those skilled in the art will readily under-
`stand the operation of Viterbi detector 220 and, therefore, a
`fuller discussion of this device is not provided here. A filter
`240 mayalso be provided upstream of the Viterbi detector
`220 to whiten the noise associated with the earlier process-
`ing (not shown) on the received signal, since it has been
`shown that Viterbi detectors provide optimal results in the
`presence of white, rather than colored, noise.
`Although channel effects are a dominant disturbance in
`conventional cellular systems, in other types of systems the
`dominant disturbance to transmitted signals may arise from
`other sources. For example, a new relatively low cost, low
`range wireless transmission system (defined by the recently
`developed “Bluetooth” technology) has been proposed for
`localized two-waydata transmissions. Bluetooth systems are
`envisioned as a universal radio interface in the 2.45 GHz
`frequency band that enable portable electronic devices to
`connect and communicate wirelessly via short-range, ad hoc
`networks. Readers interested in various details regarding the
`Bluetooth technology are referred to the article entitled
`“BLUETOOTH—Theuniversal radio interface for ad hoc,
`wireless connectivity” authored by Jaap Haartsen and found
`in the Ericsson Review, Telecommunications Technology
`Journal No. 3, 1998, the disclosure of which is incorporated
`here by reference. Of particularinterest for this discussion is
`the fact that channel effects associated with the Bluetooth air
`
`interface may not be the dominant disturbanceto transmitted
`signals in such systems, due to the short-range nature of the
`air interface links. Accordingly, other slowly varying dis-
`turbances may be more problematic than channel effects in
`such systems.
`Such disturbances can have several origins. In many
`instances, the disturbance cannot be filtered out when the
`desired signal
`itself has low-frequency components.
`Examples of such disturbances include DC offset in homo-
`dyne receivers, offset in FM discriminators due to inaccu-
`racies in the local oscillator frequency, drift (in otherwise
`presumably constant signal levels) due to temperature and
`aging, all of which represent situations where special atten-
`tion has to be given to obtain error-free recovery of the
`desired signal.
`There are several methods for performing DC offset
`suppression. The simplest methods use DC blocking capaci-
`tors to high-passfilter the signal. However, these filters have
`long response times whichresult in long settling times after
`turning on the receiver. Such longsettling times are unac-
`ceptable in TDMAreceivers where the receiver is switched
`on and off repetitively. Another technique for performing
`DC offset suppression is differentiation followed by inte-
`gration. The differentiation removes all DC components
`since it has a zero at DC. Integration inverse filters the
`differentiated signal. The differentiation and integration can
`conveniently be carried out using adaptive delta modulation
`(ADM)techniques, e.g., as described in U.S. patent appli-
`cation Ser. No. 07/578,251, entitled “DC Offset
`Compensation”, filed in Septemberof 1990 to Paul W. Dent.
`However, this technique requires considerable oversampling
`and can only be used to suppress DC offset. Drifts and other
`slowly varying, unwanted signals cannot be suppressed.
`Other suppression techniques are carried out in the digital
`domain, but require a high dynamic range of the A-to-D
`converter since no suppression has taken place prior to the
`digital processing.
`
`SONY Exhibit 1006 - 0013
`
`SONY Exhibit 1006 - 0013
`
`

`

`US 6,563,892 B1
`
`3
`Accordingly, it would be desirable to provide a solution to
`address the problems associated with slowly varying
`disturbances, such as DC offset, drift, etc.
`SUMMARY
`
`These, and other, drawbacks, limitations and problems
`associated with conventional techniques for compensating
`for slowly varying disturbances are overcomebythe present
`invention which removes slowly varying disturbances
`superimposed on streams of binary symbols with known
`separation. According to one exemplary method, the signal
`is first sampled at the symbol rate. The symbols are then fed
`into a discrete, finite-impulse-response (FIR)
`filter
`that
`removes the disturbance. The effects of the filter on the
`desired binary signal (e.g.,
`intersymbol
`interference) are
`undonein a decoderthat applies a Viterbi algorithm and acts
`like an equalizer. The Viterbi algorithm uses the knowledge
`of the constant amplitude separation of the binary signals,
`and forms estimates of the filter response for different
`possible desired input sequences.
`The complexity of the exemplary systems according to
`the present invention depends on the length of the FIRfilter.
`The longer the FIR filter, the more states are required in the
`Viterbi algorithm. The performance of the system depends
`on the coefficients of the FIR filter. Exemplary embodiments
`of the present invention describe how totailor the trade-off
`that can be made between the noise performance and the
`suppression performance of the FIR filter.
`BRIEF DESCRIPTION OF THE DRAWINGS
`
`The foregoing objects, features and advantages of the
`present invention, as well as other features, will be more
`readily understood upon reading the following detailed
`description in conjunction with the drawings in which:
`FIG. 1(a) shows a conventional radio communication
`system including plural base stations and a mobile telephone
`switching office;
`FIG. 1(b) depicts a portion of a conventional receiver
`including a channel estimator, Viterbi detector andfilter;
`FIG. 2 showsa signal representation of binary signals in
`an arbitrary noisy environment;
`FIGS. 3(a)-(d) show various types of disturbed informa-
`tion signals which can be found in communications
`receivers, more specifically;
`FIG. 3(a) illustrates step responses in an information
`signal caused by on-off events;
`FIG. 3(b) illustrates an additional step due to an adjacent
`interferer in a homodyne detector;
`FIG. 3(c) depicts the effects caused by a continuous wave
`interferer on an information signal;
`FIG. 3(d) illustrates the effects caused by a drifting local
`oscillator on an information signal;
`FIG. 4 is an example of sampled, binary sequence in
`presence of noise;
`FIG. 5 is a trellis diagram for a two-state Viterbi decoder
`employing a 2-tap FIR filter;
`FIG. 6 is an exemplary embodimentof a receiver accord-
`ing to the present invention;
`FIG. 7 depicts an exemplary embodimentof the present
`invention in which an adaptive arrangement is provided to
`suppress disturbances;
`FIG. 8 is an exampleof a discrete FIR filter in the analog
`domain; and
`FIG. 9 is a flow diagram of a method of decoding wanted
`information symbols from a wanted signal according to a
`exemplary edmbodiment of a the invention.
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`4
`DETAILED DESCRIPTION
`
`In the following description, for purposes of explanation
`and not limitation, specific details are set forth, such as
`particular circuits, circuit components, techniques, etc.
`in
`order to provide a thorough understanding of the present
`invention. However,it will be apparent to one skilled in the
`art that the present invention may be practiced in other
`embodiments that depart from these specific details. In other
`instances, detailed descriptions of well-known methods,
`devices, and circuits are omitted so as not to obscure the
`description of the present invention.
`The exemplary radio communication systems discussed
`herein are described as useful in systems employing the
`Bluetooth technology, e.g., having operating characteristics
`such as frequency hopped CDMA,low transmit power,etc.
`However, those skilled in the art will appreciate that the
`concepts disclosed herein find use in other protocols and
`systems, including, but not limited to, frequency division
`multiple access (FDMA),
`time division multiple access
`(TDMA), code division multiple access (CDMA), or some
`hybrid of any of the above protocols.
`In particular, exemplary embodiments of the present
`invention provide techniques for handling disturbances
`associated with binary signals. A general binary signal
`representation is shown in FIG. 2. Therein, there are two
`possible symbols, +A and —-A. Dueto noise, there is a signal
`probability distribution around the noise-free values +A and
`-A. A bit decision can be made by providing a threshold
`between the two possible values, and then making a bit
`decision according to the location of the detected sample
`with respect to the threshold. The optimal threshold location
`is at
`the intersection of the two probability distribution
`functions as shown in FIG. 2. Often, there is mirror sym-
`metry in the functions between +A and —-A, and then the
`threshold can be placed halfway between the twosignal
`values.
`
`If a disturbance is superimposed on the binary signal, the
`signal values +A and -A are translated in the X-direction
`according to the disturbance. FIGS. 3(a)-3(d) show some
`examples of signals disturbed by DC offset or other slowly
`varying signals. Each of these figures depicts the original
`signal as the top function, followed by one or more disturb-
`ing signals, with the resultant combination of the original
`signal plus the disturbing signal illustrated at the bottom of
`eachfigure. For example, in FIG. 3(a), an original signal 300
`is disturbed by DC offsets 310 resulting in the composite
`signal 320. These DC offsets are generated as soon as the
`electronic circuitry of the receiver is switched on. Thus, DC
`blocking is not possible for this condition, since a fast
`receiver response is needed so that the receiver can detect,
`e.g., the beginning of a signal burst associated with that
`receiver’s channel.
`
`In FIG. 3(b), an additional DC step response 360 in the
`middle of the burst is seen in the composite signal 370,
`which additional DC step response can be experienced, for
`example, in homodynereceivers due to intermodulation in
`the receive chain. As in FIG. 3(q),the original signal 380 in
`FIG. 3(b) is also disturbed by DC offset 390. FIG. 3(c)
`showsa continuous wave interference signal 392 superim-
`posed on the desired signal 394 to result in a composite
`signal 396. Additionally, FIG. 3(d) showsa drift 398 in the
`detected signal 399, which drift can result from temperature
`variations, aging or imbalance problems.
`Despite these types of variations and disturbances expe-
`rienced by the signal, the threshold for the bit value deter-
`mination has conventionally remained fixed (assuming no a
`
`SONY Exhibit 1006 - 0014
`
`SONY Exhibit 1006 - 0014
`
`

`

`US 6,563,892 B1
`
`6
`5
`the interference present, which
`priori knowledge about
`Consider the binary symbols with noise as shown in FIG.
`would allow an adaptation of the threshold). Asaresult, bit
`4. For those readers more familiar with the binary alphabet
`errors are introduced since the fixed threshold does not
`being {+1, -1}, a 0 corresponds to a -1 value whereas a 1
`remain located at the optimal position midway between the
`correspondsto a +1 value. The signal separation A is 2 in this
`example and the desired sequence is x={1 000 0 1}.
`signal values. However, since the amplitude of the binary
`signal is fixed (e.g., as in binary FM signals, wherein the
`However, due to noise, the sampled values y__1 to y_6are,
`in this example, {0.9, -1.2, -0.1, -0.5, -0.7, 0.7} as shown
`modulation index represents the amplitude of the final
`in FIG. 4.
`detection signal at the output of the FM detector and is
`fixed), the separation A between the signal symbols remains
`If the detector only takes the difference between adjacent
`samples, the difference signal is then {-2.1, 1.1, -0.4, -0.2,
`fixed regardless of the disturbance superimposed on the
`1.4} for this example. Assumingthat the first bit x_1 was a
`signal.
`1, a conventional DFSK detector would give an output
`Therefore, a better detection technique for binary signals
`sequence of {101111}, using the frameworkthat a 1 is
`disturbed by slowly varying signals is to abandon the
`decided whenthe difference is bigger than 1.0, a 0 is decided
`threshold technique, and instead use the difference A
`whenthe difference is smaller than -1.0, and no change from
`between the two possible symbols. To be able to use this
`the previous bit is decided when the difference is between
`technique, the signal separation between the two symbols
`-1.0 and 1.0.
`must be fixed. Therefore, this difference technique can only
`However, Applicants have recognized that a more intel-
`be used in systems where this separation is constant and is
`ligent scheme would recognize other characteristics within
`not affected by propagation effects. Such difference tech-
`the sampled values to morecorrectly detect the received bits.
`niques include, for example, binary phase or frequency
`For example, noting the magnitude of the last difference in
`modulation schemes(e.g. continuous phase frequency shift
`the difference sequence, 1e., 1.4, makes it highly probable
`keying, CPFSK) which are widely used in wireless commu-
`that a transition from 0 to 1 took place. Therefore, x__6 must
`nication because of the property that the signal variation is
`be 1 and x_5 must be 0. Taking into accountthe relatively
`hardly affected by propagation effects.
`small differences between the sampled values associated
`A known technique that uses the difference between two
`with x_5, x_4, x_3, and x_2,it can be recognized that all
`adjacent symbols is differential keying.
`In differential
`these bits should have the same value as bit x_5, ie., 0.
`keying, a one is represented by a change between two
`Between x_2 and x_1, the difference showsthat the tran-
`adjacent symbols, whereas a zero is represented by no
`sition from 1 to 0 is highly probable. Since x_1 was
`change between adjacent symbols (or the other way around).
`assumed to be 1, x__2 must then be zero which matches with
`Differential keying is primarily found in phase modulation
`the derivation of the value of x__2 described above based on
`schemes (DPSK) but can be used in other modulation
`its similarity to bits x_5, x_4 and x_3. Therefore, the
`schemes as well. For example, frequency modulation could
`decoded sequence according to this empirical detection
`also be implemented in this manner, e.g., a DFSK scheme
`technique is {1 0 0 0 0 1}, which is identical with the
`(Differential Frequency Shift Keying) wherein for a binary
`original signal.
`one, two adjacent symbols use f,,+Af and f,-Af, whereas for
`To implement the evaluation of such characteristics, one
`a binary zero, the symbols use both f,4Af or both f,-Af. The
`exemplary embodiment employs algorithms which incorpo-
`original signal d(k) at time instant k is retrieved in the
`rate information associated with future bits or symbols,e.g.,
`detector by sampling the input signal at the symbolrate and
`the well known Viterbi algorithm (VA). In the VA, all
`comparing two adjacent samples. This can be achieved by
`possible transitions are investigated, and a record is kept
`subtracting the previous symbol x(k-1) from the current
`from the error between the hypothesized transitions and the
`symbol x(k):
`actual transitions experienced. One realization of possible
`dk)=x(k)-x(&-1)
`transitions form a path through thetrellis tree of the Viterbi
`decoder. The error signals are accumulated and represent the
`metric of the path. At each node, the path with the lowest
`metric survives,
`the other is eliminated. When sufficient
`future symbols are investigated, a bit decision of the symbol
`at the start of the paths can be performed. The number of
`future symbols tested is called the decision depth. Those
`skilled in the art will understand the operation of the Viterbi
`algorithm per se and, therefore, additional details regarding
`this algorithm are not described herein.
`Exemplary embodiments of the present invention feed an
`input signal received by a communications device, e.g., ina
`mobile phone or base station,
`through a prefilter which
`removesthe DC offset and low-frequency components. This
`prefilter applies the difference equation given in equation
`(1), above. The output of the prefilter is then passed to a
`Viterbi decoder(see FIG. 6). From the differencesignals, the
`Viterbi decoder retrieves the original signals but with the
`disturbances removed. For the prefilter of equation (1),
`wherein only the difference between two adjacent symbols
`is taken into account, the corresponding Viterbitrellis is
`shown in FIG. 5. Each state in the Viterbi trellis of FIG. 5
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`is represented by a circle into which the state transition
`arrowsfeed.
`In the trellis, the instantaneous error associated with state
`transitions is represented by E(c|p) where c is the current,
`
`SONY Exhibit 1006 - 0015
`
`@)
`
`It will be appreciated by those skilled in theart that this type
`of differential modulation scheme removesall DC offset. In
`addition, some low-frequency signals can be removed as
`long as the difference in the disturbance level between
`adjacent samples is less than A/2. DFSK modulation is,
`however, not used muchin practice since its signal-to-noise
`(SNR) performance is degraded compared to FSK. This
`degradation occurs because,for the determination of a single
`bit, the noise of two samples is taken into account in the
`differential process. Therefore, the performance in white
`Gaussian noise of DFSK modulation is more than 3 dB
`worse than that of FSK modulation.
`
`According to exemplary embodiments of the present
`invention, a performance improvement can be obtained by
`taking into account the difference signals from more than
`two adjacent samples. That is, whereas conventional DFSK
`modulation only uses the information from two samples
`which are adjacent to determine a bit’s value, exemplary
`embodimerits of the present invention use the information
`from more than two adjacent samples to makea bit decision.
`To perform this technique, the decision of a bit is delayed
`and the difference information of future bits is used to make
`
`a more accurate decision. An example will serve to better
`illustrate an exemplary method according to the present
`invention.
`
`SONY Exhibit 1006 - 0015
`
`

`

`US 6,563,892 B1
`
`7
`hypothesized bit and p is the previous, hypothesized bit. For
`a binary signal with distance A, the error signals at instant k
`are:
`
`E,(0/0)=d(k)*
`
`E,(1/0)=(d(k)-Ay*
`
`E,{0/V)=(d(k)+A
`
`E,A/1)=d(k)?
`
`where d(k)=x(k)-x(k-1)is the difference between the cur-
`rent sample and the previous sample. This instantaneous
`error is addedto the path metric. The hypothesized bits in the
`path form the path history. When sufficient bits are incor-
`porated in the path history, 1.e., when the decision depth is
`reached), a decision onthefirst bit is made by comparing the
`cumulative metrics of all paths and selecting the path with
`the lowest metric. The oldest bit in the path history of this
`path is selected as the decision bit.
`Using a Viterbi decoder 62 after the difference circuit 60
`instead of just a threshold detection (i.e., as in conventional
`DFSKdetection) improves the SNR performance by about
`1 dB.
`In addition,
`the performance in the presence of
`disturbance signals is much better than in case of conven-
`tional DFSK, since the step in the disturbance between
`adjacent samples can be muchbigger than A/2. However, the
`SNR performance is not yet as good as the conventional
`FSKdetector, as will be apparent to those skilled in the art
`from the following discussion.
`The difference circuit can be regarded as a discrete
`high-pass filter with a frequency response G(w):
`G(o)=sin(w)-1<0<%
`
`where w is the normalized frequency. This is not an ideal
`high-pass filter, since an ideal high passfilter would have a
`flat spectrum with only a zero at w=0. Since G(«) is not an
`idealfilter, the noise outof this filter is not white but colored
`which meansthat there is a correlation between the noise in
`
`adjacent difference samples. This degrades the Viterbi
`detection, which is optimized for white noise.
`The high-pass filter characteristics can be improved by
`increasing the order of the difference equation. Considerthat
`the difference circuit can be regarded as a FIR filter with
`coefficients a(m), such that:
`d(k)=a(0)-x(k)-+a(1)-x(k-1)+a(2)-x(k-2) .. .
`
`Thefirst-order filter discussed above for implementation of
`equation (1) has only two taps with values a(0)=1 and
`a(1)=-1, Le., it provides a difference only between adjacent
`samples. Therest of the coefficients of the FIR filter are zero.
`Fora higherorderfilter, more taps are used, i.e., information
`regarding other samples is employed. To maintain the DC
`offset suppression qualities ofthis filter, the sum of the FIR
`coefficients should be zero:
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`55
`
`2)
`
`In order to correct for the intersymbol interference (ISI)
`introduced in higher order FIR filters, the numberofstates
`employed by the Viterbi decoder 62 has to increase. For an
`N-tap FIR filter, the numberof required states in the Viterbi
`decoder is 2%". The instantaneous error in the Viterbi
`decoder can be derived from thefilter response as follows:
`
`60
`
`65
`
`- PN)=(d()La()era)pit
`-
`E,(clpipoPapaPs -
`a(2)po+ . .. +a(N)pyl-A/2)
`
`3)
`
`8
`where c is the current hypothesized bit, p,p. ... py is the
`path history (py is the oldest bit) and d(k) is the current
`output of the FIR filter. E, can be determined for all 2%
`possible paths. Then at each Viterbi state, the worst path
`(ie.,
`the path with highest path metric) is rejected and
`eliminated from the scheme.
`
`To suppress DC offsets and other low frequency distur-
`bances according to exemplary embodiments of the present
`invention, the FIR filter coefficients of filter 60 should be
`chosen to provide a high-pass filter characteristic having a
`predetermined order, which characteristic will represent a
`trade-off between the SNR performance (under disturbance-
`free conditions), and the disturbance suppression capabili-
`ties. The better the SNR performance, the worse the sup-
`pression of low-frequency disturbances. If only DC offset
`suppression is required in a particular implementation of the
`present invention, as compared with other types of slowly
`varying disturbances, then apart from the requirement on the
`sum ofthe coefficients as given in equation (2), an additional
`requirement
`is to have a filter impulse response whose
`autocorrelation approaches a Dirac pulse as closely as
`possible.
`The combination of a FIR filter 60 and a Viterbi detector
`62 shown in FIG. 6 is an example of what more generally
`can be considered as a prefilter to remove unwanted signal
`components followed by an equalizer to retrieve the desired
`signal. Thus, other types of filters (e.g.,
`infinite impulse
`filters) and other types of equalizers(e.g., linear, non-linear,
`decision feedback,etc.) can also be used in conjunction with
`the present invention. However, introduction of the FIR filter
`60, havingfilter coefficients which are set to suppress inband
`interference (e.g., as opposed to adjacent channel
`interference), colors the noise included in the output from
`the filter and also introduces intersymbolinterference in the
`desired signal. Thus, the coefficients of the equalizer(or the
`metrics used in the Viterbi detector 62) should be determined
`taking into accountthe filter coefficients employed in filter
`60 to optimize the detection of symbols in the desired signal.
`Since the FIR filter 60 of FIG. 6 has fixed coefficients, the
`Viterbi equalizer parametersare fixed as well. The foregoing
`exemplary embodiments have focused on suppression of DC
`offsets and low-frequency disturbances. However, the FIR
`filter 60 can also be configured to suppress other kind of
`disturbances, e.g., using equation (3),
`to recalculate the
`metrics used in the VA and provide adaptive suppression as
`illustrated in FIG. 7. In this example,
`the detector 70 is
`periodically provided with a test sequence, e.g., a sequence
`of known symbols. Thefilter 72 is adaptive and the coeffi-
`cients are adjusted by an optimizing function 74 such th