CA2076997C - Apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithms - Google Patents
Apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithmsInfo
- Publication number
- CA2076997C CA2076997C CA 2076997 CA2076997A CA2076997C CA 2076997 C CA2076997 C CA 2076997C CA 2076997 CA2076997 CA 2076997 CA 2076997 A CA2076997 A CA 2076997A CA 2076997 C CA2076997 C CA 2076997C
- Authority
- CA
- Canada
- Prior art keywords
- symbols
- signal
- filtering algorithm
- time
- rls
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L25/03012—Arrangements for removing intersymbol interference operating in the time domain
- H04L25/03019—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception
- H04L25/03057—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception with a recursive structure
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L2025/0335—Arrangements for removing intersymbol interference characterised by the type of transmission
- H04L2025/03375—Passband transmission
- H04L2025/03401—PSK
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L2025/03433—Arrangements for removing intersymbol interference characterised by equaliser structure
- H04L2025/03439—Fixed structures
- H04L2025/03445—Time domain
- H04L2025/03471—Tapped delay lines
- H04L2025/03484—Tapped delay lines time-recursive
- H04L2025/0349—Tapped delay lines time-recursive as a feedback filter
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L2025/03592—Adaptation methods
- H04L2025/03726—Switching between algorithms
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Filters That Use Time-Delay Elements (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The present invention presents an apparatus and method for recovering symbols in a data packet (101) transmitted to a receiver (201) from a remote signal source (202) in a time-varying channel using multiple adaptive filtering algorithms. The symbols in the data packet (101) are recovered using an adaptive filter 203 responsive to a recursive least squares (RLS) filtering algorithm followed by a least-mean squares (LMS) filtering algorithm. The RLS filtering algorithm is used first for its fast training and fast recovery characteristics. The LMS filtering algorithm is used second for its low complexity and high stability characteristics. The switch between the RLS and the LMS filtering algorithms may be at a fixed point in time to ensure minimal filter complexity or may be adaptive responsive to the quality of an error signal (ek) processed in a coefficient determinator (307 or 309) to ensure maximum bit error rate performance and stability.
Description
Apparatus and Method for Adaptively Filtering a Time-Varying Signal Using Multiple Filtering Algorithms Related Applications This application is related to the instant assignee's co-pending Canadian patent application No. 2,097,058 filed on August 19, 1992, invented by Henry L. Kazecki and James 15 C. Baker, entitled "Apparatus and Method for Recovering a Time-Varying Signal Using Multiple Sampling Points".
Field of the Invention The present invention relates generally to information recovery, and more particulary to an apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithms.
.,. ~y Background of the Invention The rapid expansion of the number of cellular radio telephones coupled with the desire to provide additional 5 services has prompted the development of a digital standard. The standard suggests an increase in system ~p~city over the previous analog system through the use of digital modulation and speech coding techniques. The standard for the cellular system is described in detail in Electronic Industries Association, Project Number 2398, January 1991, IS-54 (Revision A), entitled Dual-Mode Mobile Station -- Base Station Compatibility Standard.
The standard describes in 1.2 a time division multiple access (TDMA) channel 40 milliseconds long divided into L6 six equally sized data packets 6.66 milliseconds long. A
data packet is a burst of information characterized by sequentially encoded consecutive pairs of bits, commonly known as symbols.
The standard describes in 2.1.3.3.1 a linear ao modulation technique known as ~/4 shifted, differentially encoded quadrature phase shift keying (11/4 DQPSK). The symbols are transmitted into one of four phase angles (i~/4. +3~/4) using differential quadrature component signals producing an eight point 25 phase constellation. The symbols are represented by a normalized magnitude vector and a phase angle. The symbols are transmitted as changes in phase rather than absolute phases.
Signal propagation in the radio frequency band, 30 such as the 800 MHz band for cellular radiotelephones, is generally characterized by tv~o types of channel-induced distortion: time dispersion distortion and multipath distortion. These types of distortion are caused by a rapid rate of change of the received data packet's amplitude over time and are predominantly affected by the frequency of the signal, how rapidly the receiver is moving through its environment and large objects in the vicinity of the receiver. When the amplitude over a 5 portion of the data packet approaches a null, the symbols can be corrupted by noise present in the channel that alters the state of the symbol causing the receiver to detect wrong information.
Time dispersion distortion is usually found in an lD environment where a large reflecting source, such as a mountain or a tall building, is present. A receiver operating in this environment receives the data packet from a fixed source transmitter and a delayed data packet from the reflecting source. The time delay ~5 between the reception of the two data packets results in time dispersion distortion.
Multipath distortion is characterized by many components of the same data packet having different energy levels reaching the receiver at the same time.
a~ As a result, the amplitude and phase of a data packet varies over time. This variance is referred to as ~Rayleigh fading" of the data packet.
The present challenge is to recover received symbols in the data packet that were transmitted in the 25 presence of the channel-induced distortion. Adaptive filtering for channel equalization is a widespread technique for removing the channel-induced distortion.
Adaptive filtering may also be used in a wide variety of applications ranging from high resolution spectrum ~0 estimation and line enhancement, speech and biomedical signal processing, adaptive differential encoding and adaptive deconvolution to interference surpression and echo cancellation. Two data sequences are available to an adaptive filter, a desired response and an input. The input sequence is applied to a linear filter with finite impulse response. The filter response is adapted so as to make its output sequence as close to as possihle in some sense to the desired response sequence. A large 5 number of equalizer adjustment algorithms are conceivable, depending on the system requirements.
One type of adaptive filtering algorithm is a recursive least squares (RLS) filtering algorithm. A
description of the RLS filtering algorithm may be found 10 in the IEEE (Institute of Electrical and Electronic Engineers) Transactions on ;Signal Processing, Volurne 39, No. 1, January 1991, entitled "Numerically Stable Fast Transversal Filters for Recursive Least Squares Adaptive Filteringn, presented by Dirk T.M. Slock and ~5 Thomas Kalith. The RLS filtering algorithm has fast training and recovery characteristics. These characteristics are important for a receiver to operate efficiently in a channel having time dispersion distortion. However, the RLS filtering algorithm is a ~) complex calculation and tends to be unstable in a time-varying channel. The arithmetic operations of the RLS
filtering algorithm require a high dynamic range that is very difficult to get in a fixed point digital signal processor and thus the RLS filtering algorithm becomes 25 unstable. The complexity increases the receiver's processing time.
Another type of adaptive filtering algorithm is a least mean squares (LMS) filtering algorithm. A
description of the LMS filtering algorithm may be found 30 in The Bell System Technical Journal, October 1981, pages 1905-1925, entitled "Least-Squares Algorithms for Adaptive Equalizers~ by M.S. Mueller. The LMS
filtering algorithm has simple and very stable characteristics. These characteristics are important 207~997 for reducing the receiver's processing time and providing reliable data recovery. Ilowever, the LMS
filtering algorithm has slow training and recovery characteristics that do not operate efficiently in a 5 time-varying channel. The LMS filtering algorithm does not quickly converge for a rapidly changing channel.
For many situations, of which a cellular radiotelephone is merely an example, the prior art has not produced an apparatus or method of symbol recovery 10 in a time-varying signal to meet the difficult requirement of providing an adaptive filter with simple and stable characteristics while maintaining fast training and recovery characteristics.
Field of the Invention The present invention relates generally to information recovery, and more particulary to an apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithms.
.,. ~y Background of the Invention The rapid expansion of the number of cellular radio telephones coupled with the desire to provide additional 5 services has prompted the development of a digital standard. The standard suggests an increase in system ~p~city over the previous analog system through the use of digital modulation and speech coding techniques. The standard for the cellular system is described in detail in Electronic Industries Association, Project Number 2398, January 1991, IS-54 (Revision A), entitled Dual-Mode Mobile Station -- Base Station Compatibility Standard.
The standard describes in 1.2 a time division multiple access (TDMA) channel 40 milliseconds long divided into L6 six equally sized data packets 6.66 milliseconds long. A
data packet is a burst of information characterized by sequentially encoded consecutive pairs of bits, commonly known as symbols.
The standard describes in 2.1.3.3.1 a linear ao modulation technique known as ~/4 shifted, differentially encoded quadrature phase shift keying (11/4 DQPSK). The symbols are transmitted into one of four phase angles (i~/4. +3~/4) using differential quadrature component signals producing an eight point 25 phase constellation. The symbols are represented by a normalized magnitude vector and a phase angle. The symbols are transmitted as changes in phase rather than absolute phases.
Signal propagation in the radio frequency band, 30 such as the 800 MHz band for cellular radiotelephones, is generally characterized by tv~o types of channel-induced distortion: time dispersion distortion and multipath distortion. These types of distortion are caused by a rapid rate of change of the received data packet's amplitude over time and are predominantly affected by the frequency of the signal, how rapidly the receiver is moving through its environment and large objects in the vicinity of the receiver. When the amplitude over a 5 portion of the data packet approaches a null, the symbols can be corrupted by noise present in the channel that alters the state of the symbol causing the receiver to detect wrong information.
Time dispersion distortion is usually found in an lD environment where a large reflecting source, such as a mountain or a tall building, is present. A receiver operating in this environment receives the data packet from a fixed source transmitter and a delayed data packet from the reflecting source. The time delay ~5 between the reception of the two data packets results in time dispersion distortion.
Multipath distortion is characterized by many components of the same data packet having different energy levels reaching the receiver at the same time.
a~ As a result, the amplitude and phase of a data packet varies over time. This variance is referred to as ~Rayleigh fading" of the data packet.
The present challenge is to recover received symbols in the data packet that were transmitted in the 25 presence of the channel-induced distortion. Adaptive filtering for channel equalization is a widespread technique for removing the channel-induced distortion.
Adaptive filtering may also be used in a wide variety of applications ranging from high resolution spectrum ~0 estimation and line enhancement, speech and biomedical signal processing, adaptive differential encoding and adaptive deconvolution to interference surpression and echo cancellation. Two data sequences are available to an adaptive filter, a desired response and an input. The input sequence is applied to a linear filter with finite impulse response. The filter response is adapted so as to make its output sequence as close to as possihle in some sense to the desired response sequence. A large 5 number of equalizer adjustment algorithms are conceivable, depending on the system requirements.
One type of adaptive filtering algorithm is a recursive least squares (RLS) filtering algorithm. A
description of the RLS filtering algorithm may be found 10 in the IEEE (Institute of Electrical and Electronic Engineers) Transactions on ;Signal Processing, Volurne 39, No. 1, January 1991, entitled "Numerically Stable Fast Transversal Filters for Recursive Least Squares Adaptive Filteringn, presented by Dirk T.M. Slock and ~5 Thomas Kalith. The RLS filtering algorithm has fast training and recovery characteristics. These characteristics are important for a receiver to operate efficiently in a channel having time dispersion distortion. However, the RLS filtering algorithm is a ~) complex calculation and tends to be unstable in a time-varying channel. The arithmetic operations of the RLS
filtering algorithm require a high dynamic range that is very difficult to get in a fixed point digital signal processor and thus the RLS filtering algorithm becomes 25 unstable. The complexity increases the receiver's processing time.
Another type of adaptive filtering algorithm is a least mean squares (LMS) filtering algorithm. A
description of the LMS filtering algorithm may be found 30 in The Bell System Technical Journal, October 1981, pages 1905-1925, entitled "Least-Squares Algorithms for Adaptive Equalizers~ by M.S. Mueller. The LMS
filtering algorithm has simple and very stable characteristics. These characteristics are important 207~997 for reducing the receiver's processing time and providing reliable data recovery. Ilowever, the LMS
filtering algorithm has slow training and recovery characteristics that do not operate efficiently in a 5 time-varying channel. The LMS filtering algorithm does not quickly converge for a rapidly changing channel.
For many situations, of which a cellular radiotelephone is merely an example, the prior art has not produced an apparatus or method of symbol recovery 10 in a time-varying signal to meet the difficult requirement of providing an adaptive filter with simple and stable characteristics while maintaining fast training and recovery characteristics.
Summary of the Invention According to the present invention, there is provided a method for adaptively equalizing a signal received by a receiver, wherein the signal includes at leas~ one predetermined codeword and data. The method comprises the steps of storing the signal; adaptively filtering the signal in a forward and a reverse direction in time responsive to a predetermined location of the at least one predetermined codeword in the signal and adaptive filter coefficients to produce an equalized signal; updating the adaptive filter coefficients using a first algorithm while adaptively filtering at least the at least one predetermined codeword; and updating the adaptive filter coefficients using a second algorithm while adaptively filtering the data.
Brief Description of the Drawings FIG. 1 illustrates one data packet in a time division multiple access (TDMA) channel.
FIG. 2 is a block diagram for a radio telephone transceiver constructed in accordance with the present invention for receiving the data packet of FIG.1.
FIG.3 is a block diagram of the ~ecision feedback equalizer of FIG. 2 constructed in accordance with the present invention.
FIGS. 4A and 4B describe the ~ecision process carried out in the digital signal processor (DSP) of FIG. 2 for adaptively filt~ring the symbols in the data packet of FIG. 1 using multiple filtering algorithms.
Brief Description of the Drawings FIG. 1 illustrates one data packet in a time division multiple access (TDMA) channel.
FIG. 2 is a block diagram for a radio telephone transceiver constructed in accordance with the present invention for receiving the data packet of FIG.1.
FIG.3 is a block diagram of the ~ecision feedback equalizer of FIG. 2 constructed in accordance with the present invention.
FIGS. 4A and 4B describe the ~ecision process carried out in the digital signal processor (DSP) of FIG. 2 for adaptively filt~ring the symbols in the data packet of FIG. 1 using multiple filtering algorithms.
Detailed Description of a Preferred Embodiment FIG. 1 illustrates the format for a data packet 101 in a TDMA system. A receiver 201 (shown in FIG. 2) 5 selectively receives digital information transmitted from a remote signal source 202 (shown in FIG. 2). The digital information includes a plurality of sequential symbols. A predetermined number of sequential symbols forms a data packet 101 having information for the 10 receiver 201. A primary feature of the preferred embodiment of the present invention is that the data packet 101 is adaptively filtered using multiple filtering algorithms rather than only a conventional single filtering algorithm to remove channel-induced 15 distortion.
Adaptively filtering the data packet 101 using multiple filtering algorithms is advantageous for receiving long data packets transmitted in a time-varying channel. In the preferred embodiment of the present invention, the RLS filtering algorithm is used over a portion of the data packet 101 for its fast training and recovery characteristics and the LMS
filtering algorithm is used over a portion of the data packet 101 for its stability and low complexity 25 characteristics. Thus, the combined filtering algorithm's complexity is reduced and the stability is improved while maintaining the fast training and recovery characteristics. In the preferred embodiment of the present invention, the bit error rate performance 30 for symbol recovery using the RLS and the LMS filtering algorithms in combination is substantially the same as using only the RLS filtering algorithm. A receiver 201 employing the present invention may offer improved audio quality, receiver control operation or data reception for the received data packet transmitted in a time-varying channel.
In the TDMA system, the data packet 101 includes a codeword providing synchronization, a desired slot sync 5 word (DSSW), and a coded digital voice color code (CDVCC). A second data packet 102 adjacent to the data packet 101 also has a codeword providing synchronization designated as the ~dj~cent slot sync word (ASSW) bec~us~ of its location A~ cent to the data lD packet 101. In accordance with the preferred embodiment of the present invention, the data packet 101 is divided into four regions (A, B, C and D) for symbol training and recovery, each region (A, B, C and D) is adjacent to one of the three codewords. The symbols 15 in each region (A, B, C and;E)) are adaptively filtered using multiple filtering algorithms.
The data packet 101 includes in sequential order:
the DSSW having 14 symbols, a slow associated control channel (SACCH) having 6 symbols, 65 symbols of data, ~) the CDVCC having 6 symbols, another 65 symbols of data, and 6 symbols reserved (RSV) for future use.
Sequentially following the RSV symbols is the ASSW
having 14 symbols residing in the adjacent data packet 102. The DSSW and the ASSW are typically used for 25 synchronization, equalizer retraining and time slot verification of the data packet 101 and the adjacent data packet 102, respectively, as described in the standard per 1.2.4. The DSSW and the ASSW have good autocorrelation properties to facilitate synchronization 30 and training. The CDVCC described in the standard 1.2.5 provides the receiver 201 with channel control information.
Region A includes the fourteen DSSW symbols, six SACCH symbols and twenty nine data symbols 113 and 115. Region B includes the six CDVCC symbols and thlrty six data symbols 117 and 119 on the left side of the CDVCC. Re~ion C includes th~ six CDVCC symbols and thirty six data symbols 121 and 123 on the ri~ht side of 5 ttle CDVCC. Region D includes the fourteen ASSW
symbols, six RSV symbols and twent,y nine data symbols 125 and 127.
The symbols in region A are recovered in a fonvard direction in time using a first samplin~ point determined 1~ from correlatin~ the receiver 201 to the DSSW. The symbols in region B are recovered in a reverse direction in time using a second samplin~ point determined from correlating the receiver 201 to the CDVCC. The symbols in re~ion C are recovered in a forward direction in time 15 using the second sampling point determined from correlatin~ the receiver 201 to the CDVCC. The symbols in region D are recovered in a reverse direction in time using a third samplin~ point determined from correlating the receiver to the ASSW. Data recovery in ttle forward and backward direction in tirne is described in the instant assignee's co-pending Canadian patent application No. 2,073,182 filed on December 20, 1991, invented by Henry L. Kazecki and James C. Baker, entitled "Apparatus and Method for Recovering a Time-Varying Symbol in a 25 Serial Data System". Data recovery using multiple sampling points is described in the instant assignee's co-pending Canadian patent application No. 2,097,058 filed on August 19, 1992, invented by Henry L. Kazecki and James C.
saker~ entitied ~Apparatus and Method for Recovering a 30 Time-Varying Signal Using Multiple Sampling Points".
In the preferred embodiment of the present invention the symbols in region A are filtered first with the RLS
filtering algorithm for the fourteen DSSW
, ~
~,~ ., ~ .. ..
symbols, the six SACCH symbols and the eleven data symbols 113 then switching to the LMS filtering algorithm for the following eighteen data symbols 115.
The symbols in region B are filtered first with the RLS
5 filtering algorithm for the six CDVCC symbols and the eighteen data symbols 117 then switching to the LMS
filtering algorithm for the following eighteen data symbols 119. The symbols in region C are filtered first with the RLS filtering algorithm for the six CDVCC.
Adaptively filtering the data packet 101 using multiple filtering algorithms is advantageous for receiving long data packets transmitted in a time-varying channel. In the preferred embodiment of the present invention, the RLS filtering algorithm is used over a portion of the data packet 101 for its fast training and recovery characteristics and the LMS
filtering algorithm is used over a portion of the data packet 101 for its stability and low complexity 25 characteristics. Thus, the combined filtering algorithm's complexity is reduced and the stability is improved while maintaining the fast training and recovery characteristics. In the preferred embodiment of the present invention, the bit error rate performance 30 for symbol recovery using the RLS and the LMS filtering algorithms in combination is substantially the same as using only the RLS filtering algorithm. A receiver 201 employing the present invention may offer improved audio quality, receiver control operation or data reception for the received data packet transmitted in a time-varying channel.
In the TDMA system, the data packet 101 includes a codeword providing synchronization, a desired slot sync 5 word (DSSW), and a coded digital voice color code (CDVCC). A second data packet 102 adjacent to the data packet 101 also has a codeword providing synchronization designated as the ~dj~cent slot sync word (ASSW) bec~us~ of its location A~ cent to the data lD packet 101. In accordance with the preferred embodiment of the present invention, the data packet 101 is divided into four regions (A, B, C and D) for symbol training and recovery, each region (A, B, C and D) is adjacent to one of the three codewords. The symbols 15 in each region (A, B, C and;E)) are adaptively filtered using multiple filtering algorithms.
The data packet 101 includes in sequential order:
the DSSW having 14 symbols, a slow associated control channel (SACCH) having 6 symbols, 65 symbols of data, ~) the CDVCC having 6 symbols, another 65 symbols of data, and 6 symbols reserved (RSV) for future use.
Sequentially following the RSV symbols is the ASSW
having 14 symbols residing in the adjacent data packet 102. The DSSW and the ASSW are typically used for 25 synchronization, equalizer retraining and time slot verification of the data packet 101 and the adjacent data packet 102, respectively, as described in the standard per 1.2.4. The DSSW and the ASSW have good autocorrelation properties to facilitate synchronization 30 and training. The CDVCC described in the standard 1.2.5 provides the receiver 201 with channel control information.
Region A includes the fourteen DSSW symbols, six SACCH symbols and twenty nine data symbols 113 and 115. Region B includes the six CDVCC symbols and thlrty six data symbols 117 and 119 on the left side of the CDVCC. Re~ion C includes th~ six CDVCC symbols and thirty six data symbols 121 and 123 on the ri~ht side of 5 ttle CDVCC. Region D includes the fourteen ASSW
symbols, six RSV symbols and twent,y nine data symbols 125 and 127.
The symbols in region A are recovered in a fonvard direction in time using a first samplin~ point determined 1~ from correlatin~ the receiver 201 to the DSSW. The symbols in region B are recovered in a reverse direction in time using a second samplin~ point determined from correlating the receiver 201 to the CDVCC. The symbols in re~ion C are recovered in a forward direction in time 15 using the second sampling point determined from correlatin~ the receiver 201 to the CDVCC. The symbols in region D are recovered in a reverse direction in time using a third samplin~ point determined from correlating the receiver to the ASSW. Data recovery in ttle forward and backward direction in tirne is described in the instant assignee's co-pending Canadian patent application No. 2,073,182 filed on December 20, 1991, invented by Henry L. Kazecki and James C. Baker, entitled "Apparatus and Method for Recovering a Time-Varying Symbol in a 25 Serial Data System". Data recovery using multiple sampling points is described in the instant assignee's co-pending Canadian patent application No. 2,097,058 filed on August 19, 1992, invented by Henry L. Kazecki and James C.
saker~ entitied ~Apparatus and Method for Recovering a 30 Time-Varying Signal Using Multiple Sampling Points".
In the preferred embodiment of the present invention the symbols in region A are filtered first with the RLS
filtering algorithm for the fourteen DSSW
, ~
~,~ ., ~ .. ..
symbols, the six SACCH symbols and the eleven data symbols 113 then switching to the LMS filtering algorithm for the following eighteen data symbols 115.
The symbols in region B are filtered first with the RLS
5 filtering algorithm for the six CDVCC symbols and the eighteen data symbols 117 then switching to the LMS
filtering algorithm for the following eighteen data symbols 119. The symbols in region C are filtered first with the RLS filtering algorithm for the six CDVCC.
10 symbols and the eighteen data symbols 121 then switching to the LMS filtering algorithm for the following eighteen data symbols 123. The symbols in region D are filtered first with the RLS filtering algorithm for the fourteen ASSW symbols, the six RSV
~5 symbols and the eleven data symbols 125 then switching to the LMS filtering algorithm for the following eighteen data symbols 127.
The switch from the RLS filtering algorithm to the LMS filtering algorithm is made at a predetermined point in time determined by counting the symbols in the data packet 101. The RLS filtering algorithm is used first because of its fast training and recovery characteristics. In the data packet 101 having one hundred sixty two symbols,. ninety-six symbols are 25 filtered using the RLS filtering algorithm (the six CDVCC
symbols are filtered twice for training) and seventy-two symbols are filtered using the LMS filtering algorithm.
Using the LMS filtering algorithm for a substantial number of symbols in the data packet 101 results in a 30 substantial reduction in the processing time for symbol recovery and increased filter stability.
An alternate embodiment of the present invention adaptively switches from the RLS and the LMS filtering algorithm responsive to the quality of an error signal (ek 1 1 207699~
in FIG. 3) in the adaptive filter rather than by counting the symbols. The error signal ek is determined from the difference between a desired signal rk and the filtered output signal Yk. The error signal ek is affected by the 5 quality of the received signal, i.e. the data packet 101, and the instability generated by the RLS filtering algorithm. For example, using the RLS filtering algorithm can increase the error signal level if it is used in a nondelay spread channel. Also, using the RLS
~D filtering algorithm in a delay spread channel over multiple iterations can cause instability thereby increasing the error signal. As in the preferred embodiment a fast converging filtering algorithm such as the RLS filtering algorithm is used first to quickly ~5 train the filter 203. The error signal ek is compared to a predetermined threshold signal in the filtering algorithm. If the error signal is below the threshold signal, the adaptive filter continues to use the RLS
filtering algorithm. If the error signal is above the threshold signal, the adaptive filter switches to the LMS
filtering algorithm.
The present invention is not intended to be restricted to the description of the preferred embodiment. The number and type of adaptive filtering 2Ej algorithms used in accordance need not be limited to the RLS filtering algorithm and the LMS filtering algorithm.
Depending on such factors as the time duration of the data packet 101, the type of channel-induced distortion, and the requirements of the receiver 201, various types 30 and the number (two or more) of adaptive filter filtering algorithms may be used. Some other types of adaptive filtering algorithms are described in John Proakis, Digital Communications, Chapter 8 (1989).
12 20769~7 The format of the regions (A, B, C and D) are not intended to be limited to the specified number of regions or the number of symbols per region. Rather, any number of regions, including one, may be used to filter 5 the received information. Also, the data packet 101 is not restricted to be formatted in accordance with the IS-54 digital standard. Rather, any like signal format may embody the present invention.
A block diagram of a cellular subscriber radiotelephone 200 employing the present invention is shown in FIG. 2. Signal information including the data packet 101 is transmitted from the remote signal source 202, such as a cellular base station, to the radiotelephone 200. In an alternate embodiment, the 15 receiver 201 may reside in a cellular base station and the remote signal source 202 would represent a cellular subscriber radiotelephone.
The data packet 101 of FIG. 1 is coupled to a duplex filter 205 via an antenna 207. The duplex filter ao 205 separates the received and transmit band of frequencies such that a signal may be received at the same time another signal is transmitted.
An IF receiving stage 209, coupled to the duplex filter 205, comprises a filtër that is more selective 26 than the duplex filter to generate an IF signal. The IF
stage 209 separates the frequency of the data packet 101 from other radio frequency signals in the receive band presented at the antenna 207. An analog-to-digital (A/D) converter 211 converts the IF signal from the IF
30 stage 209 in an analog format to a digital format for a data buffer Z13.
The data buffer 213 stores the data packet 101 for selective processing by a digital signal processor (DSP) 255, such as DSP/556001, provided by Motorola, Inc.
The use of the DSP/56001 is described in DSP/56001 Digital Signal Processor User Manual, revision 1, available from Motorola, Inc.
A sampling point detector 215 provides an optimum sampling point for symbol recovery of the data packet 101. A controller 217 sends a slot sync signal at line 219 for the sampling point detector 215 to synchronize the receiver 201 to the data packet 101.
The sampling point detector 215 then trains the receiver 201 on the DSSW in the data packet 101 responsive to a train/recover signal at line 221 set to a training mode by the controller 217. In the training mode, the sampling point detector 215 uses a reference vector rv from the reference vector buffer 223. In the training ~5 mode, the decision feedb~ck equalizer 203 also uses the reference vector rv from the reference vector buffer 223 responsive to the train-recover signal 221 biasing a switch 225 to couple the reference vector rv to the decision feedback equalizer (DFE) 203.
aD The sampling point detector 215 couples a symbol clock signal at line 227 to the controller 217. The controller 217 keeps track of which symbol is being processed in the sampling point detector 215 via the symbol clock signal at line 227.
After training, the controller 217 switches the train/recover signal at line 221 to a recovery state which couples a recovered signal rs to the DFE 203 via the switch 225. The symbols in the data packet 101 are recovered in a forward and backward direction in time responsive to a forward/backward signal at line 229 coupled from the controller 217 to the sampling point detector 215. The symbols in the data packet 101 are recovered using either the RLS filtering algorithm or the LMS filtering algorithm responsive to an RLS/LMS select 14 207fi997 signal at line 231 coupled from the controller 217 to the DFE 203. The controller 217 decides at what point in time to switch between the forward and backward recovery modes and between the RLS and the LMS
5 filtering algorithms responsive to the symbol clock signal at line 227 from the sampling point detector 215.
A symbol locator signal at line 233 generated by the controller 217 lets the sampling point detector 215 know at which symbol to begin forward or backward lD recovery.
The sampling point detector 215 produces sampled symbols xk corresponding to the symbols in the data packet 101. The sampled symbols Xk are adaptively filtered by the decision feedback equalizer (DFE) 203 to produce corresponding filtered symbols Yk. The DFE 203 is a particular type of an adaptive filter. Other types of adaptive filters may be used in accordance with the present invention. A variety.of adaptive equalizers are described in IEEE (Institute of Electrical and Electronic ao Engineers) Communication Magazine, March 1982, in an article entitled, ~Adaptive Equalization~ by Shahid Qureshi, pages 9 through 16.
The controller 217 signals the DFE 203 via line 231 which adaptive filtering algorithm (RLS or LMS) to 25 use for filtering the sampled symbol xk responsive to the symbol clock signal at line 227. The DFE tracks the phase of the data packet 101 and cancels the distortion caused by the delayed version of the data packet 101.
DFE's are generally described in John Proakis, Digital 30 Communications, Chapter 6 (1989).
The filtered symbols Yk are coupled from DFE 203 to a symbol recovery processor 235. A symbol recovery processor 235 translates the.symbol to a position on the eight point constellation, representing a recovered symbol. The symbol position on the eight point constellation represents two bits of binary information.
The symbol recovery processor 235 is a conventional coherent detector used in digital communications for 5 symbol recovery. Coherent detection is described in Bernard Sklar, Digital Communications, Fundamentals and Applications, Chapter 3 (1988).
A binary data process.or 237 receives the recovered symbol from the symbol recovery processor lD 235 to produce corresponding formatted signals. The binary data processor 237 separates recovered voice information from recovered control information. Voice output signals are coupled to a vocoder 239 that decode the voice output signal to produce a digital 16 representation of the voice output signal. The digital voice signal is converted to an analog voice signal in a D/A converter 241. The analog voice signal is coupled to a speaker 243, providing audible voice. The control output signals from the binary data processor 237 are a~ coupled to a control unit 245 having a keypad and a display.
Voice and control information may also be transmitted by the cellular radiotelephone 200. Per the IS-54 standard, the content of the plurality of 25 sequential symbols transmitted is different from content of the plurality of sequential symbols received.
An audible voice message coupled to a microphone 247 produces an analog voice signal that is converted to a digital voice signal in a A/D converter 249. The digital 30 voice signal is encoded into symbol information by the vocoder 239. The encoded symbol is formatted into the data packet 101 with any control information from the control unit 245 via the binary data processor 237. The symbols in the formatted data packet 101 are coupled to a D/A converter 251 to produce an analog signal. The analog signal from the D/A converter 251 is transmitted by a transmitter 253 through the duplex filter 205 from the antenna 207 for reception by the remote signal 5 source 202.
The sampling point detector 215, the controller 217, the decision feedb~ck equalizer 203, the switch 225, and the symbol recovery processor 235 are employed in the digital signal processor 255, the data ~D buffer 213 and the reference vector buffer 223 are memory portions of a conventional random access memory (RAM).
FIG. 3 is a block diagram of the ~lecision feedb~Gk equalizer 203 of FIG. 2 constructed in accordance with 15 the present invention. The DFE 203 is a simple nonlinear equalizer that is particularly useful for channels with severe amplitude distortion. The DFE 203 uses decision feedback to cancel the interference from the symbols which have already been detected. The equalized signal a~ Yk is the sum of the outputs of feedforward and feedback parts of the equalizer. The sampled symbol xk from the sampling point detector 215 is processed through the delay line with T-second taps (where T is the symbol duration). The outputs of the taps 301 25 through 304 are linearly weighted by equalizer coefficients Co through C4, respectively, and summed in summer 305 to produce the equalized output signal Yk.
The difference between a desired reference signal rk and the output signal Yk is determined in a summer 306 to 30 produce the error signal ek. The error signal ek is coupled to the RLS and LMS coefficient determinators 307 and 309, respectively. The tap coefficients are set to subtract the effects of interference from symbols that are adjacent in time to the desired symbol.
17 2~76997 Decisions made on the equ~ erl signal Yk fed back via the transversal filter 304. The basic priciple of the DFE
is that if the value of the symbols already detected are known (past decisions are assumed to be correct), then the distortion contributed by these symbols can be cancelled exactly, by subtracting past symbol values with appropriate weighting from the e~lu~ er output.
The criterion for selecting the tap coefficients Cn is typically based on the minimization of either peak distortion or mean-square distortion. Using adaptive equalization, the coefficients Cn are continually and automatically adjusted directly from the received data.
The equ~ ed symbols Yk of the DFE are expressed in terms of the sampled input symbols xk and the tap L6 coefficients Cn as:
Yk = ~n=0 Cn Xk-n + C4 rk 1 where rk is either rv or rS.
ao The coefficients Cn for the DFE 203 are generated by either the RLS coefficient determinator 307 or the LMS coefficient determinator 309. The determination of which coefficient determinator to use is made based on the state of the RLS/LMS select signal at line 231 from the controller 217. If the state of the RLS/LMS select signal at line 231 is high, the RLS coefficient determinator is enabled and LMS coefficient determinator is disabled. An invertor 311 causes the state of the RLS/LMS select signal to be inverted for the LMS coefficient detorminator 309. Likewi~e, if the state of the RLS/LMS select signal at line 231 is low the RLS coefficient determinator 307 is disabled and the LMS coefficient determinator 309 is enabled. Both the RLS coefficient determinator- 307 and the LMS
coefficient determinator 309 generate the coefficients Cn responsive to the sampled symbol xk and the error signal ek. The RLS coefficient determinator 307 also uses the desired signal rk (rs or the reference vector rv 5 depending on whether the equalizer is in a recovery or training mode, respectively); The coefficients Cn for the RLS and the LMS filtering algorithms may be determined using equations described in John Proakis, Digital Communications, Chapter 6 (1989) FIG. 4A and 4B describe the decision flow process carried out in the DSP 255. At the block 401 in FIG. 4A, symbols are retrieved from the data packet 101 that are stored in the data buffer 213. At block 403, the sampling point is detected in the sampling point ~5 detector 215 for each symbol in the data packet. The rest of the decision flow process in FIGS. 4A and 4B may be grouped into four decision portions corresponding to the four regions A-D. Blocks 409, 411, 413, 415, and 417 are used to process region A. Blocks 419, 421, 423, 425, and 427 are used to process region B. Blocks 431, 433, 435, 437, and 439 are used to process region C.
Blocks 441, 443, 445, 447, and 449 are used to process region D.
The symbols in region A of the data packet 101 are 2~ adaptively filtered forward in time. The RLS filtering algorithm is used to train the DFE 203 on the DSSW in the data packet 101 at block 409 and then continues recovering the six SACCH symbols and the eleven data symbols forward in time at block 411. The controller 30 217 determines when to switch between the training and the recovery mode via the train/recovery signal at line 221. At block 413 a determination is made if it is time to switch from the RLS filtering algorithm to the LMS filtering algorithm. If the determination is negative the flow returns to block 411 whereby the RLS
filtering algorithm is used to continue recovsring the symbols. If the determination is positive, the eighteen data symbols 115 in the data packet 101 of FIG. 1 are 6 recovered forward in time using the LMS filtering algorithm. The controller 217 determines when to switch to the LMS filtering algorithm at block 413 responsive to the symbol clock signal at line 227 of FIG.
2. The decision to switch is made by the controller 217 lD by counting the number of symbol clock signals or by comparing the error signal to the threshold signal. The controller 217 enables the LMS filtering algorithm and disables the RLS filtering algorithm via the RLS/LMS
select signal at line 231. A determination is made at ~5 block 417 if it is time to train the DFE 203 on the CDVCC backward in time. If the determination is negative, the flow returns to block 415 to continue recovering the data symbols 115 using the LMS filtering algorithm. If the determination is positive, the flow ao continues to block 419.
The symbols in region B are adaptively filtered backward in time. The decision process continues with the CVDCC in the data packet responsive to the symbol locator signal at line 233 generated by the controller 25 217 in FIG. 2. Using the RLS filtering algorithm, the DFE
is trained on the CVDCC at block 419 and continues recovering data symbols 117 backward in time at block 421. A determination is then made at block 423 if it is time to switch to the LMS filtering algorithm. If the 30 determination is negative, the flow returns to block 421 whereby the data symbols 117 are recovered backward in time. If the determination is positive at block 423, the DFE 203 switches from using the RLS filtering algorithm to the LMS filtering algorithm responsive to 207fi997 the RLS/LMS select signal at line 231 to recover the data symbols 119 backward in time at block 425. At block 427 a determination is made if it is time to train the DFE 203 on the CDVCC forward in time. If the determination at block 427 is negative, the flow continues to block 425 wherein the data symbols 119 are recovered using the LMS filtering algorithm backward in time. If the determination at block 427 is positive, the flow continues to block 431 in FIG. 4B.
The symbols in region C are adaptively filtered forward in time. Using the RLS filtering algorithm, the DFE is trained on the CDVCC at block 431 and the data symbols 121 are recovered forward in time at block 433.
At block 435, a determination is made if it is time to switch to the LMS filtering algorithm. If the determination is negative at block 435, the flow returns to block 433 whereby the data symbols 121 are recovered forward in time. If the determination at block 435 is positive, the flow continues to block 437 wherein aD the data symbols 123 are recovered forward in time using the LMS filtering algorithm. At block 439 a determination is made if it is time to train the DFE 203 on the ASSW backward in time. If the determination is negative at block 439, the flow returns to block 437 - 26 wherein the data symbols continue to be recovered using the LMS filtering algorithm forward in time. If the determination at block 439 is positive, the flow continues to block 441.
The symbols in region D are adaptively filtered backward in time. Using the RLS filtering algorithm, the DFE 203 is trained on the ASSW at block 441 and continues recovering the six RSV symbols and the data symbols 125 backward in time. At block 445, a determination is made if it is time to switch from the RLS filtering algorithm to the LMS filtering algorithm.
If the determination at block 445 is negative, the flow returns to block 443 wherein the symbols continue to be recovered using the RLS filtering algorithm. If the 5 determination at block 445;is positive, the flow continues to block 447 whereby the data symbols 127 are recovered backward in time using the LMS filtering algorithm. While each of the 18 data symbols 127 are recovered, a determination is made at block 449 if the lD last symbol in the data packet has been recovered. If the determination is negative at block 449, the flow returns to block 447 whereby the data symbols 127 continue to be recovered using the LMS filtering algorithm backward in time. If the determination at 15 block 449 is positive, the flow continues to block 451 wherein other receiver functions are processed such as mobile assisted handoff (MAHO) and voice or data determination.
Thus, a receiver 201 having an adaptive filter uses aD multiple filtering algorithms; to train and recover information in a time-varying channel. The switch from one filtering algorithm to the other may be responsive to a predetermined location of a symbol in the data packet or the quality of the error signal in the adaptive 25 filter. In the preferred embodiment of the present invention, multiple filtering algorithms were used to minimize processing time in the DSP while providing substantially equivalent bit error rate performance. In other applications where processing time is not as 30 critical, other advantages may be gained such as improving the bit error rate performance and improving the stability of the adaptive filter.
What is claimed is:
~5 symbols and the eleven data symbols 125 then switching to the LMS filtering algorithm for the following eighteen data symbols 127.
The switch from the RLS filtering algorithm to the LMS filtering algorithm is made at a predetermined point in time determined by counting the symbols in the data packet 101. The RLS filtering algorithm is used first because of its fast training and recovery characteristics. In the data packet 101 having one hundred sixty two symbols,. ninety-six symbols are 25 filtered using the RLS filtering algorithm (the six CDVCC
symbols are filtered twice for training) and seventy-two symbols are filtered using the LMS filtering algorithm.
Using the LMS filtering algorithm for a substantial number of symbols in the data packet 101 results in a 30 substantial reduction in the processing time for symbol recovery and increased filter stability.
An alternate embodiment of the present invention adaptively switches from the RLS and the LMS filtering algorithm responsive to the quality of an error signal (ek 1 1 207699~
in FIG. 3) in the adaptive filter rather than by counting the symbols. The error signal ek is determined from the difference between a desired signal rk and the filtered output signal Yk. The error signal ek is affected by the 5 quality of the received signal, i.e. the data packet 101, and the instability generated by the RLS filtering algorithm. For example, using the RLS filtering algorithm can increase the error signal level if it is used in a nondelay spread channel. Also, using the RLS
~D filtering algorithm in a delay spread channel over multiple iterations can cause instability thereby increasing the error signal. As in the preferred embodiment a fast converging filtering algorithm such as the RLS filtering algorithm is used first to quickly ~5 train the filter 203. The error signal ek is compared to a predetermined threshold signal in the filtering algorithm. If the error signal is below the threshold signal, the adaptive filter continues to use the RLS
filtering algorithm. If the error signal is above the threshold signal, the adaptive filter switches to the LMS
filtering algorithm.
The present invention is not intended to be restricted to the description of the preferred embodiment. The number and type of adaptive filtering 2Ej algorithms used in accordance need not be limited to the RLS filtering algorithm and the LMS filtering algorithm.
Depending on such factors as the time duration of the data packet 101, the type of channel-induced distortion, and the requirements of the receiver 201, various types 30 and the number (two or more) of adaptive filter filtering algorithms may be used. Some other types of adaptive filtering algorithms are described in John Proakis, Digital Communications, Chapter 8 (1989).
12 20769~7 The format of the regions (A, B, C and D) are not intended to be limited to the specified number of regions or the number of symbols per region. Rather, any number of regions, including one, may be used to filter 5 the received information. Also, the data packet 101 is not restricted to be formatted in accordance with the IS-54 digital standard. Rather, any like signal format may embody the present invention.
A block diagram of a cellular subscriber radiotelephone 200 employing the present invention is shown in FIG. 2. Signal information including the data packet 101 is transmitted from the remote signal source 202, such as a cellular base station, to the radiotelephone 200. In an alternate embodiment, the 15 receiver 201 may reside in a cellular base station and the remote signal source 202 would represent a cellular subscriber radiotelephone.
The data packet 101 of FIG. 1 is coupled to a duplex filter 205 via an antenna 207. The duplex filter ao 205 separates the received and transmit band of frequencies such that a signal may be received at the same time another signal is transmitted.
An IF receiving stage 209, coupled to the duplex filter 205, comprises a filtër that is more selective 26 than the duplex filter to generate an IF signal. The IF
stage 209 separates the frequency of the data packet 101 from other radio frequency signals in the receive band presented at the antenna 207. An analog-to-digital (A/D) converter 211 converts the IF signal from the IF
30 stage 209 in an analog format to a digital format for a data buffer Z13.
The data buffer 213 stores the data packet 101 for selective processing by a digital signal processor (DSP) 255, such as DSP/556001, provided by Motorola, Inc.
The use of the DSP/56001 is described in DSP/56001 Digital Signal Processor User Manual, revision 1, available from Motorola, Inc.
A sampling point detector 215 provides an optimum sampling point for symbol recovery of the data packet 101. A controller 217 sends a slot sync signal at line 219 for the sampling point detector 215 to synchronize the receiver 201 to the data packet 101.
The sampling point detector 215 then trains the receiver 201 on the DSSW in the data packet 101 responsive to a train/recover signal at line 221 set to a training mode by the controller 217. In the training mode, the sampling point detector 215 uses a reference vector rv from the reference vector buffer 223. In the training ~5 mode, the decision feedb~ck equalizer 203 also uses the reference vector rv from the reference vector buffer 223 responsive to the train-recover signal 221 biasing a switch 225 to couple the reference vector rv to the decision feedback equalizer (DFE) 203.
aD The sampling point detector 215 couples a symbol clock signal at line 227 to the controller 217. The controller 217 keeps track of which symbol is being processed in the sampling point detector 215 via the symbol clock signal at line 227.
After training, the controller 217 switches the train/recover signal at line 221 to a recovery state which couples a recovered signal rs to the DFE 203 via the switch 225. The symbols in the data packet 101 are recovered in a forward and backward direction in time responsive to a forward/backward signal at line 229 coupled from the controller 217 to the sampling point detector 215. The symbols in the data packet 101 are recovered using either the RLS filtering algorithm or the LMS filtering algorithm responsive to an RLS/LMS select 14 207fi997 signal at line 231 coupled from the controller 217 to the DFE 203. The controller 217 decides at what point in time to switch between the forward and backward recovery modes and between the RLS and the LMS
5 filtering algorithms responsive to the symbol clock signal at line 227 from the sampling point detector 215.
A symbol locator signal at line 233 generated by the controller 217 lets the sampling point detector 215 know at which symbol to begin forward or backward lD recovery.
The sampling point detector 215 produces sampled symbols xk corresponding to the symbols in the data packet 101. The sampled symbols Xk are adaptively filtered by the decision feedback equalizer (DFE) 203 to produce corresponding filtered symbols Yk. The DFE 203 is a particular type of an adaptive filter. Other types of adaptive filters may be used in accordance with the present invention. A variety.of adaptive equalizers are described in IEEE (Institute of Electrical and Electronic ao Engineers) Communication Magazine, March 1982, in an article entitled, ~Adaptive Equalization~ by Shahid Qureshi, pages 9 through 16.
The controller 217 signals the DFE 203 via line 231 which adaptive filtering algorithm (RLS or LMS) to 25 use for filtering the sampled symbol xk responsive to the symbol clock signal at line 227. The DFE tracks the phase of the data packet 101 and cancels the distortion caused by the delayed version of the data packet 101.
DFE's are generally described in John Proakis, Digital 30 Communications, Chapter 6 (1989).
The filtered symbols Yk are coupled from DFE 203 to a symbol recovery processor 235. A symbol recovery processor 235 translates the.symbol to a position on the eight point constellation, representing a recovered symbol. The symbol position on the eight point constellation represents two bits of binary information.
The symbol recovery processor 235 is a conventional coherent detector used in digital communications for 5 symbol recovery. Coherent detection is described in Bernard Sklar, Digital Communications, Fundamentals and Applications, Chapter 3 (1988).
A binary data process.or 237 receives the recovered symbol from the symbol recovery processor lD 235 to produce corresponding formatted signals. The binary data processor 237 separates recovered voice information from recovered control information. Voice output signals are coupled to a vocoder 239 that decode the voice output signal to produce a digital 16 representation of the voice output signal. The digital voice signal is converted to an analog voice signal in a D/A converter 241. The analog voice signal is coupled to a speaker 243, providing audible voice. The control output signals from the binary data processor 237 are a~ coupled to a control unit 245 having a keypad and a display.
Voice and control information may also be transmitted by the cellular radiotelephone 200. Per the IS-54 standard, the content of the plurality of 25 sequential symbols transmitted is different from content of the plurality of sequential symbols received.
An audible voice message coupled to a microphone 247 produces an analog voice signal that is converted to a digital voice signal in a A/D converter 249. The digital 30 voice signal is encoded into symbol information by the vocoder 239. The encoded symbol is formatted into the data packet 101 with any control information from the control unit 245 via the binary data processor 237. The symbols in the formatted data packet 101 are coupled to a D/A converter 251 to produce an analog signal. The analog signal from the D/A converter 251 is transmitted by a transmitter 253 through the duplex filter 205 from the antenna 207 for reception by the remote signal 5 source 202.
The sampling point detector 215, the controller 217, the decision feedb~ck equalizer 203, the switch 225, and the symbol recovery processor 235 are employed in the digital signal processor 255, the data ~D buffer 213 and the reference vector buffer 223 are memory portions of a conventional random access memory (RAM).
FIG. 3 is a block diagram of the ~lecision feedb~Gk equalizer 203 of FIG. 2 constructed in accordance with 15 the present invention. The DFE 203 is a simple nonlinear equalizer that is particularly useful for channels with severe amplitude distortion. The DFE 203 uses decision feedback to cancel the interference from the symbols which have already been detected. The equalized signal a~ Yk is the sum of the outputs of feedforward and feedback parts of the equalizer. The sampled symbol xk from the sampling point detector 215 is processed through the delay line with T-second taps (where T is the symbol duration). The outputs of the taps 301 25 through 304 are linearly weighted by equalizer coefficients Co through C4, respectively, and summed in summer 305 to produce the equalized output signal Yk.
The difference between a desired reference signal rk and the output signal Yk is determined in a summer 306 to 30 produce the error signal ek. The error signal ek is coupled to the RLS and LMS coefficient determinators 307 and 309, respectively. The tap coefficients are set to subtract the effects of interference from symbols that are adjacent in time to the desired symbol.
17 2~76997 Decisions made on the equ~ erl signal Yk fed back via the transversal filter 304. The basic priciple of the DFE
is that if the value of the symbols already detected are known (past decisions are assumed to be correct), then the distortion contributed by these symbols can be cancelled exactly, by subtracting past symbol values with appropriate weighting from the e~lu~ er output.
The criterion for selecting the tap coefficients Cn is typically based on the minimization of either peak distortion or mean-square distortion. Using adaptive equalization, the coefficients Cn are continually and automatically adjusted directly from the received data.
The equ~ ed symbols Yk of the DFE are expressed in terms of the sampled input symbols xk and the tap L6 coefficients Cn as:
Yk = ~n=0 Cn Xk-n + C4 rk 1 where rk is either rv or rS.
ao The coefficients Cn for the DFE 203 are generated by either the RLS coefficient determinator 307 or the LMS coefficient determinator 309. The determination of which coefficient determinator to use is made based on the state of the RLS/LMS select signal at line 231 from the controller 217. If the state of the RLS/LMS select signal at line 231 is high, the RLS coefficient determinator is enabled and LMS coefficient determinator is disabled. An invertor 311 causes the state of the RLS/LMS select signal to be inverted for the LMS coefficient detorminator 309. Likewi~e, if the state of the RLS/LMS select signal at line 231 is low the RLS coefficient determinator 307 is disabled and the LMS coefficient determinator 309 is enabled. Both the RLS coefficient determinator- 307 and the LMS
coefficient determinator 309 generate the coefficients Cn responsive to the sampled symbol xk and the error signal ek. The RLS coefficient determinator 307 also uses the desired signal rk (rs or the reference vector rv 5 depending on whether the equalizer is in a recovery or training mode, respectively); The coefficients Cn for the RLS and the LMS filtering algorithms may be determined using equations described in John Proakis, Digital Communications, Chapter 6 (1989) FIG. 4A and 4B describe the decision flow process carried out in the DSP 255. At the block 401 in FIG. 4A, symbols are retrieved from the data packet 101 that are stored in the data buffer 213. At block 403, the sampling point is detected in the sampling point ~5 detector 215 for each symbol in the data packet. The rest of the decision flow process in FIGS. 4A and 4B may be grouped into four decision portions corresponding to the four regions A-D. Blocks 409, 411, 413, 415, and 417 are used to process region A. Blocks 419, 421, 423, 425, and 427 are used to process region B. Blocks 431, 433, 435, 437, and 439 are used to process region C.
Blocks 441, 443, 445, 447, and 449 are used to process region D.
The symbols in region A of the data packet 101 are 2~ adaptively filtered forward in time. The RLS filtering algorithm is used to train the DFE 203 on the DSSW in the data packet 101 at block 409 and then continues recovering the six SACCH symbols and the eleven data symbols forward in time at block 411. The controller 30 217 determines when to switch between the training and the recovery mode via the train/recovery signal at line 221. At block 413 a determination is made if it is time to switch from the RLS filtering algorithm to the LMS filtering algorithm. If the determination is negative the flow returns to block 411 whereby the RLS
filtering algorithm is used to continue recovsring the symbols. If the determination is positive, the eighteen data symbols 115 in the data packet 101 of FIG. 1 are 6 recovered forward in time using the LMS filtering algorithm. The controller 217 determines when to switch to the LMS filtering algorithm at block 413 responsive to the symbol clock signal at line 227 of FIG.
2. The decision to switch is made by the controller 217 lD by counting the number of symbol clock signals or by comparing the error signal to the threshold signal. The controller 217 enables the LMS filtering algorithm and disables the RLS filtering algorithm via the RLS/LMS
select signal at line 231. A determination is made at ~5 block 417 if it is time to train the DFE 203 on the CDVCC backward in time. If the determination is negative, the flow returns to block 415 to continue recovering the data symbols 115 using the LMS filtering algorithm. If the determination is positive, the flow ao continues to block 419.
The symbols in region B are adaptively filtered backward in time. The decision process continues with the CVDCC in the data packet responsive to the symbol locator signal at line 233 generated by the controller 25 217 in FIG. 2. Using the RLS filtering algorithm, the DFE
is trained on the CVDCC at block 419 and continues recovering data symbols 117 backward in time at block 421. A determination is then made at block 423 if it is time to switch to the LMS filtering algorithm. If the 30 determination is negative, the flow returns to block 421 whereby the data symbols 117 are recovered backward in time. If the determination is positive at block 423, the DFE 203 switches from using the RLS filtering algorithm to the LMS filtering algorithm responsive to 207fi997 the RLS/LMS select signal at line 231 to recover the data symbols 119 backward in time at block 425. At block 427 a determination is made if it is time to train the DFE 203 on the CDVCC forward in time. If the determination at block 427 is negative, the flow continues to block 425 wherein the data symbols 119 are recovered using the LMS filtering algorithm backward in time. If the determination at block 427 is positive, the flow continues to block 431 in FIG. 4B.
The symbols in region C are adaptively filtered forward in time. Using the RLS filtering algorithm, the DFE is trained on the CDVCC at block 431 and the data symbols 121 are recovered forward in time at block 433.
At block 435, a determination is made if it is time to switch to the LMS filtering algorithm. If the determination is negative at block 435, the flow returns to block 433 whereby the data symbols 121 are recovered forward in time. If the determination at block 435 is positive, the flow continues to block 437 wherein aD the data symbols 123 are recovered forward in time using the LMS filtering algorithm. At block 439 a determination is made if it is time to train the DFE 203 on the ASSW backward in time. If the determination is negative at block 439, the flow returns to block 437 - 26 wherein the data symbols continue to be recovered using the LMS filtering algorithm forward in time. If the determination at block 439 is positive, the flow continues to block 441.
The symbols in region D are adaptively filtered backward in time. Using the RLS filtering algorithm, the DFE 203 is trained on the ASSW at block 441 and continues recovering the six RSV symbols and the data symbols 125 backward in time. At block 445, a determination is made if it is time to switch from the RLS filtering algorithm to the LMS filtering algorithm.
If the determination at block 445 is negative, the flow returns to block 443 wherein the symbols continue to be recovered using the RLS filtering algorithm. If the 5 determination at block 445;is positive, the flow continues to block 447 whereby the data symbols 127 are recovered backward in time using the LMS filtering algorithm. While each of the 18 data symbols 127 are recovered, a determination is made at block 449 if the lD last symbol in the data packet has been recovered. If the determination is negative at block 449, the flow returns to block 447 whereby the data symbols 127 continue to be recovered using the LMS filtering algorithm backward in time. If the determination at 15 block 449 is positive, the flow continues to block 451 wherein other receiver functions are processed such as mobile assisted handoff (MAHO) and voice or data determination.
Thus, a receiver 201 having an adaptive filter uses aD multiple filtering algorithms; to train and recover information in a time-varying channel. The switch from one filtering algorithm to the other may be responsive to a predetermined location of a symbol in the data packet or the quality of the error signal in the adaptive 25 filter. In the preferred embodiment of the present invention, multiple filtering algorithms were used to minimize processing time in the DSP while providing substantially equivalent bit error rate performance. In other applications where processing time is not as 30 critical, other advantages may be gained such as improving the bit error rate performance and improving the stability of the adaptive filter.
What is claimed is:
Claims (5)
1. A method for adaptively equalizing a signal received by a receiver, wherein the signal includes at least one predetermined codeword and data, the method comprising the steps of:
storing the signal;
adaptively filtering the signal in a forward and a reverse direction in time responsive to a predetermined location of the at least one predetermined codeword in th signal and adaptive filter coefficients to produce an equalized signal;
updating the adaptive filter coefficients using a first algorithm while adaptively filtering at least the at least one predetermined codeword; and updating the adaptive filter coefficients using a second algorithm while adaptively filtering the data.
storing the signal;
adaptively filtering the signal in a forward and a reverse direction in time responsive to a predetermined location of the at least one predetermined codeword in th signal and adaptive filter coefficients to produce an equalized signal;
updating the adaptive filter coefficients using a first algorithm while adaptively filtering at least the at least one predetermined codeword; and updating the adaptive filter coefficients using a second algorithm while adaptively filtering the data.
2. A method according to claim 1 further comprising the steps of:
determining whether a predetermined number of symbols in the at least one predetermined codeword and a first data portion of the signal have been adaptively filtered;
selecting the first algorithm responsive to a determination that the predetermined number of symbols and the first data portion have not yet been adaptively filtered; and selecting the second algorithm responsive to a determination that the predetermined number of symbols and the first data portion have been adaptively filtered.
determining whether a predetermined number of symbols in the at least one predetermined codeword and a first data portion of the signal have been adaptively filtered;
selecting the first algorithm responsive to a determination that the predetermined number of symbols and the first data portion have not yet been adaptively filtered; and selecting the second algorithm responsive to a determination that the predetermined number of symbols and the first data portion have been adaptively filtered.
3. A method according to claim 1 further comprising the steps of:
determining whether a quality measure of the equalized signal is either favourable or unfavourable;
selecting the first algorithm responsive to a determination that the quality measure is favourable;
and selecting the second algorithm responsive to a determination that the quality measure is unfavourable.
determining whether a quality measure of the equalized signal is either favourable or unfavourable;
selecting the first algorithm responsive to a determination that the quality measure is favourable;
and selecting the second algorithm responsive to a determination that the quality measure is unfavourable.
4. A method according to claim 1:
wherein the data includes a first data portion and a second data portion;
wherein the adaptive filter coefficients are updated using the first algorithm while adaptively filtering the at least one predetermined codeword and the first data portion; and wherein the adaptive filter coefficients are updated using the second algorithm while adaptively filtering the second data portion.
wherein the data includes a first data portion and a second data portion;
wherein the adaptive filter coefficients are updated using the first algorithm while adaptively filtering the at least one predetermined codeword and the first data portion; and wherein the adaptive filter coefficients are updated using the second algorithm while adaptively filtering the second data portion.
5. A method according to claim 1:
wherein the first algorithm updates the adaptive filter coefficients at a first rate; and wherein the second algorithm updates the adaptive filter coefficients at a second rate substantially less than the first rate.
wherein the first algorithm updates the adaptive filter coefficients at a first rate; and wherein the second algorithm updates the adaptive filter coefficients at a second rate substantially less than the first rate.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US76998391A | 1991-09-30 | 1991-09-30 | |
| US769,983 | 1991-09-30 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA2076997C true CA2076997C (en) | 1996-12-10 |
Family
ID=25087110
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA 2076997 Expired - Fee Related CA2076997C (en) | 1991-09-30 | 1992-08-27 | Apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithms |
Country Status (3)
| Country | Link |
|---|---|
| CA (1) | CA2076997C (en) |
| GB (1) | GB2260068B (en) |
| MX (1) | MX9205516A (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2738967B1 (en) * | 1995-09-15 | 1997-12-05 | France Telecom | ADAPTIVE EQUALIZATION DEVICE FOR DIGITAL COMMUNICATIONS SYSTEMS |
| US5742642A (en) * | 1996-10-29 | 1998-04-21 | Telefonaktiebolaget Lm Ericsson | Signal processing method and apparatus for reducing equalizer error |
| US6581080B1 (en) | 1999-04-16 | 2003-06-17 | Sony United Kingdom Limited | Digital filters |
| GB9908813D0 (en) * | 1999-04-16 | 1999-06-09 | Sony Uk Ltd | Digital filtering |
| JP3296330B2 (en) * | 1999-05-20 | 2002-06-24 | 日本電気株式会社 | Adaptive filter, control method of adaptive filter, and storage medium storing program |
| GB2357938A (en) * | 1999-12-24 | 2001-07-04 | Nokia Networks Oy | Selecting one of a plurality of equalisation algorithms |
| DE10013790A1 (en) * | 2000-03-20 | 2001-11-22 | Sel Alcatel Ag | Distorted input signal processing circuit has filter whose filter coefficients are influenced by comparison signal produced by equalized signal from filter and output signal of threshold decider |
| WO2002003635A2 (en) * | 2000-07-05 | 2002-01-10 | Nokia Networks Oy | Two-step equaliser for receiving station and associated method |
| US6853681B1 (en) * | 2000-11-28 | 2005-02-08 | Telefonaktiebolaget L M Ericsson (Publ) | Using different channel models for different training sequences |
| FR2871236B1 (en) * | 2004-06-02 | 2006-09-01 | Gervais Danone Sa | METHOD FOR CONTROLLING A MICROBIOLOGICAL PROCESS FROM SUCCESSIVE TIME DERIVATIVES OF STATE VARIABLES |
| US9100257B2 (en) * | 2012-01-25 | 2015-08-04 | Marvell World Trade Ltd. | Systems and methods for composite adaptive filtering |
| CN105866068B (en) * | 2016-03-28 | 2020-04-24 | 北京航空航天大学 | Method for extracting harmonic signal peak point of laser absorption spectrum based on recursive least square method |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2554996B1 (en) * | 1983-11-14 | 1986-02-28 | Labo Electronique Physique | METHOD AND DEVICE FOR DETERMINING THE OPTIMAL POSITION OF THE REFERENCE COEFFICIENT OF AN ADAPTIVE EQUALIZER |
| GB8511835D0 (en) * | 1985-05-10 | 1985-06-19 | British Telecomm | Adaptive digital filter |
| DE68920362D1 (en) * | 1988-06-09 | 1995-02-16 | Nat Semiconductor Corp | Hybrid stochastic gradient algorithm for the convergence of adaptive filters. |
| FI83010C (en) * | 1989-08-24 | 1991-05-10 | Nokia Data Systems | Method for fine tuning an equalizer for use in a data transmission system receiver |
| EP0426026B1 (en) * | 1989-10-31 | 1996-08-28 | Mitsubishi Denki Kabushiki Kaisha | Equalizer |
| US5230007A (en) * | 1991-06-28 | 1993-07-20 | Motorola, Inc. | Method for optimizing an adaptive filter update coefficient |
| AU655384B2 (en) * | 1991-07-26 | 1994-12-15 | General Instrument Corporation | Carrier phase recovery for an adaptive equalizer |
-
1992
- 1992-08-27 CA CA 2076997 patent/CA2076997C/en not_active Expired - Fee Related
- 1992-09-24 GB GB9220184A patent/GB2260068B/en not_active Expired - Fee Related
- 1992-09-28 MX MX9205516A patent/MX9205516A/en not_active IP Right Cessation
Also Published As
| Publication number | Publication date |
|---|---|
| MX9205516A (en) | 1993-03-01 |
| GB2260068B (en) | 1995-11-15 |
| GB2260068A (en) | 1993-03-31 |
| GB9220184D0 (en) | 1992-11-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US5241545A (en) | Apparatus and method for recovering a time-varying signal using multiple sampling points | |
| CA2076084C (en) | Adaptive mlse-va receiver for digital cellular radio | |
| CA2076123C (en) | Decision feedback equalization for digital cellular radio | |
| EP1254542B1 (en) | A method of receiving a signal and a receiver | |
| US5249205A (en) | Order recursive lattice decision feedback equalization for digital cellular radio | |
| US7580455B2 (en) | Decision feedback equalization with fractionally-spaced feedback data | |
| EP0615347B1 (en) | Adaptative equalizing receiver and maximum likelihood sequence estimation receiver | |
| EP0519891A2 (en) | PSK demodulator for multipath fading channels | |
| CA2076997C (en) | Apparatus and method for adaptively filtering a time-varying signal using multiple filtering algorithms | |
| US5257265A (en) | Method and apparatus for reducing multipath distortion | |
| WO1997010553A1 (en) | Device and method for adaptive narrowband interference suppresson in multiple-access communication | |
| KR20020064996A (en) | A method of detecting a sequence of information symbols, and a mobile station adapted to performing the method | |
| US5182749A (en) | Receiver for recovering data in a forward and reverse direction in time | |
| US5361400A (en) | Apparatus and method for removing distortion in a received signal | |
| AU682700B2 (en) | Digital radio communication equipment | |
| Masoomzadeh-Fard et al. | Nonlinear equalization of multipath fading channels with noncoherent demodulation | |
| US6782046B1 (en) | Decision-directed adaptation for coded modulation | |
| JP3224468B2 (en) | Data receiving device | |
| US20020167999A1 (en) | Equalizer, receiver, and equalization method and reception method | |
| JP2003502910A (en) | Method and system for binary information detection in the presence of slowly changing disturbances | |
| JPH0490614A (en) | Equalizer | |
| KR960014680B1 (en) | Signal equalizer & method of equalising received data signal | |
| JPH06350467A (en) | Receiver for digital communications | |
| US20060193411A1 (en) | Digital receiver method and apparatus using differential detection and equalization-based techniques | |
| Dahlman | Improved adaptive Viterbi detectors for fast-varying mobile-radio channels |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |