HK1056796B

HK1056796B - Method and apparatus for processing a modulated signal using an equalizer and a rake receiver

Info

Publication number: HK1056796B
Application number: HK03109045.3A
Authority: HK
Inventors: J‧E‧斯密; A‧贾拉利
Original assignee: 高通股份有限公司
Priority date: 2000-07-24
Filing date: 2001-07-17
Publication date: 2007-04-20

Description

Method and apparatus for processing modulated signal using equalizer and rake receiver

(1) Field of the invention

The present invention relates to data communication. More particularly, the present invention relates to a novel and improved method and apparatus for processing a received modulated signal with an equalizer and a rake receiver for improved performance.

(2) Background of the invention

Modern communication systems need to support a variety of applications. Code Division Multiple Access (CDMA) is one such system that supports voice and data communications between users over a landline. The use of CDMA technology IN MULTIPLE ACCESS COMMUNICATION SYSTEMs is disclosed IN U.S. patent application Ser. No. 4901307 entitled "CDMA COMMUNICATION System Using satellite TERRESTRIAL REPEATERS" and U.S. patent application Ser. No. 5103459 entitled "System and method FOR GENERAT ING WAVEFORMS IN A CDMA CELLULAR TELEPHONE System". Another specific CDMA system is disclosed in U.S. patent application serial No. 08/963386 entitled "METHOD AND APPARATUS FOR HIGH RATE PACKET DATA TRANSMISSION mismatch" (hereinafter HDR system), U.S. patent 6,574,211 issued 11/3 1997 AND granted 6/3/2003. These patents are assigned to the assignee of the present invention and are incorporated herein by reference.

CDMA systems are typically designed to comply with one or more standards. Such standards include "TIA/EIA/IS-95 Remote Station-Base Station Compatibility Standard for Dual-Mode Wireless spread Spectrum Cellular System" (IS-95 Standard), which IS designated by the name "3^rdThe Forum of GenerationPartnership Project "(3 GPP) provides and embodies standards in a set of documents including NOS.3G TS 25.211, 3G TS25.212, 3G TS 25.213, and 3G TS 25.214(W-CDMA standards) and" TR-45.5Physical Layer Standard for CDMA2000 Spread Spectrum Systems "(CDMA-2000 standards). New CDMA standards continue to be proposed and adopted. These CDMA systems are incorporated herein by reference.

CDMA systems typically use a rake receiver to process a modulated signal transmitted on the forward or reverse link. A rake receiver typically includes a searcher unit and finger processors. The searcher unit searches for strong received signal instances (or multipaths). The assignment finger processor processes the strongest multipaths to generate demodulated symbols for those multipaths. The rake receiver then combines the demodulated symbols in all of the newly assigned finger processors together to generate estimated recovered symbols for the transmitted data. The rake receiver efficiently combines the energy received over the multipath channel.

Rake receivers provide satisfactory performance for CDMA systems operating at low signal-to-noise ratios (S/N). For CDMA systems that transmit data at high data rates, such as HDR systems, a higher S/N is required. To achieve a higher S/N, it is necessary to reduce the components of the noise term N. The noise term includes thermal noise (No), interference (Io) from other transmission sources or transmissions to other users, and intersymbol interference (ISI) from multipath and distortion in the transmission channel. For CDMA systems operating at low S/N, the ISI component is generally negligible compared to other components. However, for CDMA systems operating at higher S/N, other noise components (e.g., interference from other transmission sources) are typically reduced, and ISI becomes a non-negligible noise component that has a large impact on the overall S/N.

As described above, the rake receiver can provide satisfactory performance when the S/N of the received signal is low. A rake receiver may be used to combine energy from multipath, but it is generally not able to cancel the effects of ISI (e.g., from multipath and channel distortion). Thus, rake receivers may not be able to achieve the higher S/N required by CDMA systems operating at higher data rates.

It can be seen that there is a need for techniques that can be used to process received demodulated signals to achieve higher S/N to support higher data rates.

(3) Summary of the invention

The present invention provides techniques that enable higher S/N to be achieved to support higher data rates. In accordance with the present invention, a number of signal processing paths can be provided to process one or more signals (e.g., signals received from one or more antennas). A signal processing path includes an equalizer that attempts to reduce ISI caused by multipath and channel distortion. The other signal processing path may be implemented by one or more conventional rake receivers. Although signal processing paths including equalizers generally provide better performance under certain operating conditions, signal processing paths that provide better signal quality estimates may be selected for processing the received signal.

One embodiment of the present invention sets forth a method of processing one or more signals in a (spread spectrum) communication system. In accordance with the method, one or more signals are received (e.g., via one or more antennas) and processed to generate one or more sample streams, which are further processed with a first processing scheme to provide a first recovered symbol stream. In a first processing scheme, the sample streams are equalized and combined by an equalizer to generate symbol estimates, which can then be further processed (e.g., despread and decovered) to produce a first stream of recovered symbols. The sample stream can also be processed with a second processing scheme having one or more rake receivers to produce a second recovered symbol stream. The signal quality associated with each processing scheme can be estimated and the first or second processing scheme can be selected accordingly.

In a first processing scheme, the sample streams are equalized and then combined. In this case, each sample stream is filtered by a respective filter by coefficients and scaled by a respective scaling factor. The scaled samples for all streams are then combined to produce symbol estimates. Conversely, the sample streams may be combined prior to equalization. In this case, each sample stream is scaled by a corresponding (complex) scaling factor. All of the stream scaled samples are then combined to produce accumulated samples, which are further filtered by coefficients to produce symbol estimates.

Each filter in the equalizer may be implemented as a Finite Impulse Response (FIR) filter, an Infinite Impulse Response (IIR) filter, or some other filter structure. The filter coefficients and scaling factors are typically adapted (e.g., trimmed) before use and further adapted (e.g., with a decision directed adaptation scheme) during use.

Various adaptation schemes may be used depending on the particular design of the equalizer. In one adaptation scheme, the filter coefficients are adapted separately and sequentially to the scaling factors. In this adaptation scheme, the coefficient adaptation may be performed with a fixed scaling factor, and the scaling factor adaptation may be performed with a fixed coefficient. The coefficient adaptation and scale factor adaptation may be repeated several times (e.g., over some expected symbol sequence whose value is known). For coefficient adaptation, the coefficients of each filter may be adapted based on (1) filtered samples derived from the filter and the expected symbols or (2) symbol estimates and the expected symbols. Alternatively, another adaptation scheme adapts the coefficients of all filters simultaneously based on the symbol estimates and the expected symbols.

In the above adaptation scheme, adaptation may be performed using a Time Division Multiplexed (TDM) pilot reference and according to a Least Mean Square (LMS) algorithm, a Recursive Least Square (RLS) algorithm, a Direct Matrix Inversion (DMI) algorithm, or some other adaptation algorithm. Before adaptation, the coefficients of each filter are initialized to a specific set of values (e.g., 0,. 0, α, 0,. 0, 0.) and the scaling factors are also initialized. The large paths for each signal being received and processed can be identified and the magnitude and phase of the multiple paths (α) can be used to initialize the coefficients and scaling factors associated with the signal. A large path to a signal being received and processed may also be identified and processed, and coefficients and scaling factors may be adapted with a time offset corresponding to the multi-path (e.g., the time offset may be used to properly generate expected values).

The signal quality associated with the first processing scheme may be estimated based on a Mean Square Error (MSE) between the symbol estimates and expected symbols. The MSE may be minimized by adapting the coefficients and scaling factors. MSE can be converted to signal-to-noise ratio (S/N). MSE or S/N may be used to select the data rate for the received signal.

Another embodiment of the present invention provides an active receiver unit capable of processing one or more signals in a spread spectrum communication system. The receiver unit includes one or more pre-processors and a first signal processing path comprised of an equalizer and a post-processor. Each pre-processor receives and processes a respective signal to produce a respective sample stream. An equalizer receives, combines, and equalizes one or more sample streams to generate symbol estimates. A post processor receives and further processes (e.g., despreads and decovers) the symbol estimates to produce a first stream of recovered symbols. The receiver unit may also include a second signal processing path and a controller. The second signal processing path is comprised of one or more rake receivers for processing the sample stream to generate a second recovered symbol stream. The controller receives signal quality estimates associated with the first and second signal processing paths and selects either the first or second signal processing path based on the received signal quality estimates.

The post-processor may include a PN despreader and a decover unit. A PN despreader receives and despreads the symbol estimates with the PN sequence at a particular time offset to produce despread samples. A decover unit decovers the despread samples with one or more channelization (e.g., Walsh) codes to produce a first stream of recovered symbols.

In one design, an equalizer may include one or more filters, one or more multipliers, and an adder. Each filter receives a respective sample stream and filters it with a set of coefficients to produce respective filtered samples. Each multiplier receives the filtered samples from a respective filter and appends them with a respective scaling factor to produce scaled samples. An adder receives the scaled samples from all of the multipliers and adds them to produce a symbol estimate.

In another design, the equalizer may include one or more multipliers, an adder, and a filter. Each multiplier receives a respective sample and scales it with a respective scaling factor to produce scaled samples. The adder receives the scaled samples from all of the multipliers and adds them to produce added samples. A filter receives the summed samples and filters them with a set of coefficients to produce symbol estimates.

In the above design, the equalizer further comprises a coefficient adjustment unit that is capable of adapting the coefficients of each filter and the scaling factor of each multiplier. As described above, various adaptation schemes may be used. The filter coefficients may be adapted according to the selected adaptation scheme, either from the filtered samples received from the filter or from the symbol estimates. Likewise, the adaptation may also be implemented with a pilot reference and LMS, RLS, DMI, or some other algorithm.

The receiver unit may operate as a base station or a remote terminal in a spread spectrum (e.g., CDMA) communication system.

In accordance with one aspect of the present invention, a method of processing one or more signals in a spread spectrum communication system is provided. The method comprises the following steps:

receiving and processing the one or more signals to provide one or more sample streams; and

processing the one or more sample streams for a first time to provide a first recovered symbol stream, wherein the processing for the first time comprises:

equalizing the one or more sample streams with an equalizer to generate symbol estimates, the equalizing step comprising quantizing the symbol estimates and processing the symbol estimates to provide the stream of recovered symbols, and comprising the steps of:

despreading the symbol estimates with a PN sequence to generate despread symbols; and

decovering the despread symbols to generate the first stream of recovered symbols;

wherein the despreading step and the decovering step are selectively performed based on a data rate of the one or more received signals;

second processing multipaths of the one or more sample streams with one or more rake receivers to provide a second recovered symbol stream;

estimating a signal quality associated with each of the first and second processes; and

and selecting the relevant first or second processing according to the estimated signal quality.

In accordance with another aspect of the present invention, a method of processing one or more signals in a communication system is provided. The method comprises the following steps:

receiving and processing the one or more signals to provide one or more sample streams;

processing the one or more sample streams a first time to provide a first recovered symbol stream, the step of first processing comprising the steps of:

equalizing the one or more sample streams with an equalizer to generate symbol estimates, the equalizing step comprising quantizing the symbol estimates and processing the symbol estimates to provide the stream of recovered symbols,

estimating a signal quality associated with each of the first and second processes;

selecting a first or second associated treatment according to said estimated signal quality; and

adapting coefficients of each of one or more filters within the equalizer;

wherein the coefficients of each filter within the equalizer are adapted with a Time Division Multiplexed (TDM) pilot reference or a Code Division Multiplexed (CDM) pilot reference.

In accordance with yet another aspect of the present invention, a method of processing one or more signals in a communication system is provided. The method comprises the following steps:

adapting coefficients of each of one or more filters within the equalizer;

wherein the coefficients of each filter within the equalizer are adapted using a Least Mean Square (LMS) algorithm, a Recursive Least Squares (RLS) algorithm, a Direct Matrix Inversion (DMI) algorithm, or a combination thereof.

In accordance with yet another aspect of the present invention, a receiver unit for processing one or more signals in a communication system is provided. The receiver unit includes:

one or more preprocessors to receive and process the one or more signals to provide one or more sample streams;

an equalizer coupled to the one or more preprocessors for receiving, combining, and equalizing the one or more sample streams to generate symbol estimates; and

a post-processor coupled to said concatenated equalizer for receiving and processing said symbol estimates;

wherein the equalizer comprises:

one or more multipliers coupled to the one or more preprocessors, respectively, each multiplier for receiving a corresponding sample stream and multiplying the sample stream by a corresponding scaling factor to provide scaled samples;

an adder coupled to said one or more multipliers for receiving said scaled samples from said one or more multipliers and adding them to provide added samples;

a slicer coupled to the summer for receiving and quantizing the symbol estimates to generate sliced symbol estimates;

a filter coupled to said adder and configured to receive the summed samples and filter them with a set of coefficients to provide said symbol estimates;

a coefficient adjusting unit, connected to the filter, for adapting the set of coefficients of the filter based on the symbol estimates; and

a delayer coupled to said filter for receiving and layering said symbol estimates to provide layered symbol estimates;

wherein the coefficient adjustment unit adapts the set of coefficients of the filter according to the layered symbol estimates.

In accordance with another aspect of the present invention, an apparatus is provided. The apparatus comprises:

an equalizer for processing signal samples to provide symbol estimates, the equalizer comprising a filter and a coefficient adjustment unit for adjusting coefficients by which the signal samples are filtered by the filter;

a despreader for despreading the symbol estimates from the equalizer to generate despread symbols;

a decover unit to decover the despread symbols to generate a first stream of recovered symbols;

a rake receiver for processing one or more multipaths of the signal samples to provide a second stream of recovered symbols;

a controller for estimating a first signal quality of said first recovered symbol stream and a second signal quality of a second recovered symbol stream, said controller comparing said first and second signal qualities; and

a switch for selecting either the equalizer or the rake receiver to generate recovered data symbols.

In accordance with yet another aspect of the present invention, a system is provided. The system comprises:

a base station for transmitting a pilot signal and a data signal;

a remote terminal for receiving the pilot signal and the data signal, the remote terminal comprising:

an equalizer for processing samples of a pilot signal and a data signal to provide symbol estimates, the equalizer comprising a filter and a coefficient adjustment unit configured to adjust coefficients by which the filter filters the samples;

a rake receiver for processing one or more multipaths of samples of said pilot and data signals to provide a second stream of recovered symbols;

The present invention also provides other methods and receiver units that implement various aspects and features of the present invention, as described below.

(4) Description of the drawings

The features, objects, and advantages of the present invention will become more apparent from the detailed description set forth below when taken in conjunction with the drawings. Like elements in the drawings are identified with like reference numerals, wherein:

FIG. 1 is a simplified block diagram illustrating one embodiment of signal processing for data transmission in a communication system;

fig. 2A is a block diagram of one embodiment of a receiver in a communication system;

FIG. 2B is a block diagram of one embodiment of a data processor in a receiver;

FIG. 3 is a block diagram of a receive data processor in a receiver unit in accordance with one embodiment of the present invention;

FIGS. 4A and 4B are block diagrams of two embodiments that can be used to implement the equalizer shown in FIG. 3;

FIG. 5A is a block diagram of one embodiment of an FIR filter that may be used to implement each of the filters shown in FIGS. 4A and 4B;

FIG. 5B is a block diagram of one embodiment of a post-processor that may be used to process symbol estimates from an equalizer to generate recovered symbols;

FIG. 6 is a diagram of a data frame format for forward link transmission in an HDR CDMA system;

fig. 7 is a block diagram of one embodiment of a rake receiver.

(5) Detailed description of the preferred embodiments

Fig. 1 is a simplified block diagram illustrating one embodiment of signal processing for data transmission in a communication system 100. At the transmitter unit 110, data is sent, typically in the form of data packets, from a data source 112 to a Transmit (TX) data processor 114, which formats, codes, and processes the data to generate one or more analog signals. The analog signals are then provided to a transmitter (TMTR)116 that amplifies, filters, quadrature modulates, and upconverts the analog signals to generate a modulated signal suitable for transmission to one or more receiver units via one or more antennas 118 (only one shown in fig. 1).

At receiver unit 130, the transmitted signal is received by one or more antennas 132a-132k and provided to a receiver (RCVR) 134. In receiver 134, each received signal is amplified, filtered, frequency down-converted, quadrature demodulated, and digitized to produce in-phase (I) and quadrature (Q) samples. The samples are digitally processed and provided to a Receive (RX) data processor 136 that further processes and decodes the samples to recover the transmitted data. The method of processing and decoding in the receive data processor 136 is the inverse of the method of processing and encoding in the transmit data processor 114. The decoded data is then provided to a data sink 138.

The signal processing procedures described above support unidirectional transmission of traffic data, messaging, voice, image, and other communication types. A bi-directional communication system supports bi-directional data transmission. The processing illustrated in fig. 1 may represent forward link processing in a CDMA system where transmitter unit 110 represents a base station and receiver unit 130 represents a remote terminal. For simplicity, the signal processing of the reverse link is not shown in fig. 1.

Fig. 2A is a block diagram of one embodiment of a receiver 134 in a communication system. In this embodiment, receiver unit 130 includes several antennas 132a through 132 k. In the receiver 134, each antenna 132 is connected to a respective received signal processor (or pre-processor) 210. In each pre-processor 210, the received signal from antenna 132 is amplified (low noise) by amplifier 222, filtered by Receive (RX) filter 224, down-converted and quadrature demodulated by frequency converter/demodulator 226, and digitized by one or more analog-to-digital converters (ADCs) 228 to provide ADC samples. Digital processor 230 further processes the ADC samples to generate complex I_INAnd Q_INSampled and passed to the receive data processor 136. The digital processor 230 is described in detail below.

As shown in fig. 2A, the receiver unit 130 includes a number of antennas 132A through 132k coupled to a number of pre-processors 210a through 210k for processing signals received by the antennas. Each set of antennas 132 and pre-processor 210 forms part of a signal path for processing a particular received signal. Multiple antennas 132 in receiver unit 130 provide spatial diversity and may further suppress interference from other transmission sources, which can improve performance. It is also within the scope of the present invention that receiver unit 130 could be designed to have a single signal path.

Fig. 2A shows some of the functional units that may be used to implement the pre-processor 210. In general, the preprocessor 210 may include any combination of the functional elements shown in FIG. 2A, and may be arbitrarily arranged to obtain a desired output. For example, the preprocessor 210 typically includes a multi-stage amplifier and filter. It is also within the scope of the present invention that different functional elements than those shown in fig. 2A can also be included in the preprocessor 210.

Figure 2B is a block diagram of one embodiment of a digital processor 230. The ADC 228 may be implemented at a particular sampling rate f, depending on the particular design of the receiver unit 130_ADCThe received signal is sampled and another sampling rate f is available to the subsequent rx data processor 136_SAMPAnd (5) operating sampling. For example, the received signal may be sampled at a rate that is approximately twice, four times, or eight times the chip rate. The ADC sampling rate may or may not be synchronous with the chip rate, depending on the design of the particular receiver. The rx data processor 136 is designed to operate on samples at a particular sampling rate (e.g., chip rate), which may be different from the ADC sampling rate. The digital processor 230 is used to implement sample rate conversion. In some designs, the subsequent equalizer within rx data processor 136 has a higher sampling rate (e.g., f) than the ADC sampling rate_SAMP＝2f_ADC) Is advantageous. The digital processor 230 is designed and operated to provide upsampling.

In the embodiment shown in fig. 2B, digital processor 230 includes an upsampler 242, a Finite Impulse Response (FIR) filter 244, and a downsampler 246, all connected in series. The up-sampler 242 receives the ADC samples and up-samples them by a factor P. Upsampling may be performed by inserting (P-1) zero samples between each pair of successive ADC samplesAnd (5) realizing. FIR filter 244 receives and filters the samples resulting from the up-sampling to remove the image frequency generated by the up-sampling. The FIR filter 244 further (match) filters the received samples. The filtered samples are supplied to decimator 246 for decimation by a factor Q to generate (complex) samples x₁(n) to the receive data processor 136. The decimation may be accomplished by simply deleting (Q-1) samples from every Q filtered samples.

It is within the scope of the invention that the data processor 230 can also be implemented by a sample rate converter (especially when P and Q are not integers) or some other design. Also, the data processor 230 may be designed to provide additional and/or different functionality while remaining within the scope of the present invention.

In an IS-95CDMA system, receiver units are designed to operate at low signal-to-noise ratios (S/N), where S represents the desired signal and N represents the total noise. The total noise includes thermal noise (No), interference (Io) due to other transmission sources or transmissions made to other users, and intersymbol interference (ISI) from multipath and distortion in the transmission channel. ISI is caused by multipath or frequency distortion in the received signal caused by the transmitter unit, the transmission channel, and the receiver unit. For IS-95CDMA systems, multi-user data IS transmitted simultaneously over the same system bandwidth, resulting in a low S/N. At the receiver unit, the desired signal may be recovered from the highly degraded received signal by accumulating energy over a longer period of time using spread spectrum processing, thereby producing a signal with improved S/N.

Conventional processing of spread spectrum signals is accomplished with a rake receiver that searches for strong received signal instances (or multipaths), processes the strongest multipaths, and combines the results of the processed multipaths to generate recovered symbols that are more accurately estimated for the transmitted data. The design and operation of the rake receiver is described in further detail below. The rake receiver effectively combines the energy from multiple signal paths to optimize the S/N. However, because of the limited number of finger processors used to process multipath, the ability of a rake receiver to correct for distortions in the received signal caused by the channel is also limited.

For CDMA systems that transmit at higher data rates, such as HDR systems, a higher S/N is required to support the higher data rates. To achieve a higher S/N in the forward link of an HDR system, data is transmitted to one user at any given time. This eliminates interference due to transmissions by other users. Furthermore, by operating the HDR system with a reuse factor greater than one, interference from other transmitting base stations may be reduced by using directional antennas at either the transmitter unit or the receiver unit, or both. To further improve S/N, ISI needs to be reduced as well (which IS usually negligible in IS-95 systems). The present invention provides a technique for reducing ISI caused by multipath and channel distortion in order to obtain a higher S/N.

Fig. 3 is a block diagram of the receive data processor 136 according to one embodiment of the invention. In this embodiment, the rx data processor 136 includes two signal processing paths that can be operated simultaneously to achieve improved performance, especially at higher data rates. The first signal processing path includes an equalizer 310 coupled to a post-processor 320 and the second signal processing path includes a rake receiver 330.

In the rx data processor 136, the sample streams obtained from the pre-processor 210 are supplied to an equalizer 310 and a rake receiver 330, respectively. Each sample stream is generated from a respective received signal. Equalizer 310 equalizes the received sample stream and provides symbol estimates to post-processor 320. Post-processor 320 may further process the symbol estimates to provide recovered symbols, depending on the processing performed by transmitter 110. In particular, if PN spreading and covering are performed at the transmitter unit, the post-processor 320 may be configured to despread with the complex PN sequence and decover with one or more channelization codes. After the filter coefficients are selected, the equalizer 310 can clearly perform phase rotation (via pilot demodulation for the rake receiver).

Rake receiver 330 may be configured to process one or more multipaths of each received signal to provide recovered symbols for the received signal. For each sample stream, the rake receiver 330 may be configured to perform PN despreading, decovering, and coherent demodulation of several multipaths. Rake receiver 330 then combines the demodulated symbols for all multipaths of the received signal to generate recovered symbols for the received signal. Rake receiver 330 may further combine the recovered symbols for all received signals to provide a total recovered symbol provided by the receiver.

The recovered symbols produced by the post-processor 320 and rake receiver 330 are provided to a Switch (SW)340, which selects recovered symbols from either the post-processor 320 or rake receiver 330 and provides them to a deinterleaver 350. The selected recovered symbols are then reordered by a deinterleaver and then decoded by decoder 360. The controller 370 is connected to and controls the operation of the equalizer 310, the post-processor 320, the rake receiver 330, and the switch 340.

According to the present invention, the received signal may be equalized by equalizer 310, thereby reducing the amount of ISI in the received signal. The characteristics of the transmitter unit, the transmission channel, and the receiver unit distort each received signal. The equalizer operates to equalize the overall response of each received signal, thereby reducing the amount of ISI. Lower ISI improves S/N and can support higher data rates.

Fig. 4A is a block diagram of one embodiment of an equalizer 310a, which may be used to implement the equalizer 310 shown in fig. 3. As shown in fig. 4A, the received signal from each antenna 132 is processed by a corresponding pre-processor 210 to produce a stream of samples x_i(n) of (a). In the equalizer 310a, the samples generated by the preprocessors 210a to 210k are each sent to filters 410a to 410 k. Each filter 410 applies a particular set of coefficients to the received samples x based on a particular set of coefficients adapted to the signal it receives_i(n) performing an equalization process. The symbol estimates (per antenna) generated in filters 410a through 410kToAre provided to respective multipliers 412a to 412k, which also receive respective scaling factors s₁To s_k. Each multiplier 412 is scaled by a scaling factor s_iEstimating received symbolScaled and the scaled samples are provided to adder 414. The scaling factors may be complex, however, in particular, the coefficients in the filter 410 may be adapted to use real numbers. Adder 414 receives and adds the scaled samples from multipliers 412a through 412k to generate an output (combined) symbol estimate

A slicer 418a is provided in the equalizer 310a to receive the output symbol estimatesAnd layered (i.e., quantized) to produce layered symbol estimatesIn an adaptation scheme for data pointing, the layered symbol estimates may be used to adapt the variables in the equalizer 310a, as will be described later. The design of the delayer 418a may provide more layers for higher order modulation schemes (e.g., 16QAM, 64QAM, etc.) depending on the particular quadrature modulation supported. Although not shown in FIG. 4 for simplicity, the symbol estimates generated for filters 410a through 410k are usedToLayering, one or more laminators can be provided to generate respective layered symbol estimatesTo

Coefficient adjustment unit 420a receives symbol estimatesToOutput symbol estimationLayered symbol estimationActual (expected) symbols y (n) or a combination thereof. Then, the coefficient adjustment unit 420a adapts (trims or adjusts) the coefficients of the filters 410a to 410k and the scaling factors of the multipliers 412a to 412k according to the received symbol estimates and symbols. Coefficient adjustment element 420a may implement a Least Mean Square (LMS) algorithm, a Recursive Least Squares (RLS) algorithm, a Direct Matrix Inversion (DMI) algorithm, some other algorithm, or a combination thereof. The coefficient adjustment element 420a and the adaptation algorithm are described in further detail below.

Fig. 5A is a block diagram of one embodiment of a filter 410 that may be used to implement filters 410a through 410k shown in fig. 4A. In the present embodiment, the filter 410 is implemented by an FIR filter. However, an Infinite Impulse Response (IIR) filter or some other filter structure may also be used and fall within the scope of the present invention.

In filter 410, received samples x_i(n) are sent to a number of delay elements 512a to 512m connected in series. Each delay element 512 generates a particular amount of delay (e.g., one clock cycle, 1/f, of the received sample rate clock)_SAMP). Received sample x_iThe outputs of (n) and delay elements 512a through 512m are provided to respective multipliers 514a through 514 l. Each multiplier 514 also receives a coefficient c_i，jEach received sample is scaled by the coefficient and a scaled sample is generated and sent to adder 516. Adder 516 adds the scaled samples from multipliers 514a through 514l and generates a symbol estimateSymbol estimationIt can be calculated as follows:

equation (1)

Where the symbol (—) represents the complex conjugate and (2M +1) is the number of taps of the filter 410.

Returning again to FIG. 4A, the output symbol estimates of adder 414It can be calculated as follows:

equation (2)

Where K is the number of signal processing paths allocated in equalizer 310 a.

Each filter 410 in FIG. 4A produces a frequency response C_i(w) which equalizes the overall frequency response G experienced by the signal processed by the filter_i(w) is carried out. Total frequency response G of signal_i(w) comprises a transmitter unit, a transmission channel and a receiverThe frequency response of the cell (i.e. all parts up to the filter as well). The coefficients of each filter 410 are "adapted" or "trimmed" to achieve ISI reduction.

As shown in fig. 4A, a set of signals is received by several antennas 132 and processed by a corresponding number of pre-processors 210. Each pre-processor 210 produces a sample stream for a respective filter 410. In addition, the symbol estimates obtained in the K filters 410 are scaled by K scaling factors and combined to generate an output symbol estimateThus, equalizer 310a can adjust up to K · (2M +1) + K variables. According to various adaptation schemes, the K · (2M +1) coefficients and the K scaling factors of the K filters 410 may be adapted. Some adaptation schemes are described below.

In a first adaptation scheme, the coefficients of the filter 410 are adapted "time-domain" before the scaling factors are adapted "space-domain". Time domain adaptation adapts the coefficients of each filter 410 to equalize the received signal processed by the filter. The spatial domain adaptation then adjusts the scaling factors associated with all of the assigned filters 410 to provide equalization for the equalizer 310 a.

In time domain adaptation, filters 410a through 410k may be individually adapted according to their respective received samples and equalized outputs. Scaling factor s before the first adaptation process₁(n)) to s_K(n) is initialized to a specific set of values (e.g., 1/K each). The scaling factors are generally normalized so that their sum equals one, which can be expressed as follows:

equation (3)

Samples x received according to a particular filter 410i_i(n) and expected received samplesThe set of coefficients of the filter can be adapted,c _i(n)＝c_i，-M(n)，...，c_i，M(n)。

initially, for each sampling period n where coefficient adaptation is implemented, the filter 410i uses the coefficient set c_i(n) sampling a set of received samplesx _i(n)＝x_i(n-M)，...，x_i(n + M) filtering to generate a symbol estimateCan be expressed as:

equation (4)

Whereinc _i ^H(n) is a set of coefficients for a conjugate transpose (i.e., complex conjugate transpose). Equation (4) is expressed in vector form but is the same as equation (1) above. Then estimating the symbolTo the coefficient adjustment unit 420 a.

The equalizer 310a is typically adapted with known symbols before it is first used to process the received samples. For CDMA systems that transmit a pilot reference, the pilot reference may be used to adapt the equalizer. If the pilot reference is sent using Time Division Multiplexing (TDM),for example, in an HDR system, the actual symbols y (n) output by equalizer 310a are known at the pilot reference period and may be compared with the symbol estimates (for each antenna)Or symbol estimation outputAnd (4) comparing. The error between the actual symbol and the symbol estimate can be calculated and used to adapt the equalizer. The generation of the actual symbols y (n) in the HDR system is described further below.

On the other hand, if the pilot reference IS sent using Code Division Multiplexing (CDM), such as in an IS-95 system, the symbol estimates are further processedOrTo release CDM processing and thereby generate corresponding recovered pilot symbolsOrWhich are each associated with a corresponding actual pilot symbol p_i(n) or p (n) for comparison.

The error between the actual symbol and the symbol estimate can then be calculated and used to adapt the equalizer. The process of adapting the equalizer 310a with the TDM pilot reference is described in detail below. The process of adapting equalizer 310a with a CDM pilot reference is described in the following paragraphs.

In the known sampling period n of the actual symbol y (n), the coefficient adjustment unit 420a can receive or generate the actual symbol y (n). Symbol estimation in filter 410iError e between actual symbols y (n)_i(n) can be calculated by the following formula:

equation (5)

The error at a particular point in the processing path may be calculated from the difference between the symbol estimate and the expected symbol at that point.

On the other hand, in sample periods n where the actual symbols y (n) are unknown, symbol estimates may be computed for filter 410IAnd layered symbol estimationError e between_i(n) the following:

equation (6)

Error e calculated by equation (5) or equation (6)_i(n) may be used to generate a new set of coefficients for the next sampling periodc _i(n+1)。

As mentioned above, a variety of adaptation algorithms, some of which are described below, may be used to generate the new coefficients.

In the LMS algorithm, new coefficientsc _i(n +1) can be calculated by the following formula:

equation (7)

Where μ is the unitless adaptation constant.

In the RLS algorithm, new coefficientsc _i(n +1) can be calculated by the following formula:

equation (8)

And

where λ is the memory factor (typically 0.95 < λ ≦ 1.0),k _i(n) is the gain vector, I is the identity matrix (i.e., the matrix with values on the diagonal all equal to one), and P is the inverse correlation matrix. Initially P_i(0) δ · I, where δ is a small positive number (e.g., 0.001).

In DMI algorithms, N may be used_SYMCalculating a new coefficient c for one trimming period of the symbol_i(n +1) represented by the following formula:

equation (9)

And

whereinIs an estimate of the autocorrelation matrix (i.e. the filter content) of the received samples,is the cross vector of the filter content and the desired output. Estimated valueAndmay be aggregated over multiple (disjoint) trim intervals. The equation (9) is performed whenever necessaryIs inverted (e.g., when each slot in the HDR system is based on two separate pilot references, as shown in fig. 6).

The LMS, RLS, DMI algorithms all attempt (directly or indirectly) to minimize the Mean Square Error (MSE) as:

equation (10)

Where E { x } is the expected value of x. LMS, RLS, DMI and other adaptation algorithms are described in further detail by Simon Haykin in a book entitled "Adaptive Filter Theory" (third edition, prentier Hall, 1996), which is incorporated herein by reference.

When the filters 410a to 410k are individually adapted in the above-described manner, an adaptation of the spatial domain may be performed to adapt the scaling factors. The adaptation in the spatial domain may be achieved by methods similar to those used for the time-domain adaptation described above. In particular, the coefficients of the filters 410a to 410k are fixed, and the scaling factor s is adjusted₁(n) to s_K(n)。

Initially, for each sampling period n for which the scaling factor is adapted, each filter 410I uses coefficientsc _iSampling a set of received samplesx _i(n) filtering to generate symbol estimatesSymbol estimates from filters 412a through 410kToAre fed to respective multipliers 412a to 412k by respective scaling factors s₁(n) to s_K(n) scaling to produce an output symbol estimateIs represented as follows:

equation (11)

Output symbol estimationPasses through the coefficient adjustment unit 420a and generates an errorThe difference e (n), expressed as:

or equation (12)

Equation (13)

Depending on whether the actual symbol is known (equation 12) or unknown (equation 13). The calculated error e (n) is then used to generate a new scaling factor s according to the LMS, RLS, DMI or some other algorithm₁(n +1) to s_K(n +1) (i.e.s(n +1)), which is similar to the above-described filter coefficient adaptation method.

The desired performance may be achieved by several adaptation iterations in the time and space domains (e.g., using the same or different pilot references). The iterative adaptation process may be accomplished by:

1) initializing the scaling factor to 1.0 (i.e., s)₁＝s₂＝...＝s_K＝1/K)；

2) Initializing coefficients for each filter 410a through 410k (e.g., in one of the ways described below);

3) coefficients for each filter 410ic _i(n) performing time domain adaptation;

4) coefficient of filters 410a to 410kc ₁To c _kKeeping unchanged:

5) for scaling factors s when the filter coefficients are constant₁(n) to s_K(n) performing adaptation of the spatial domain;

6) maintaining the scaling factor unchanged;

7) steps 3 to 5 are repeated several times until the desired result is obtained.

The symbol estimates for equalizer 310a are adapted once the coefficients and scaling factors for filters 410a through 410k are adaptedIt represents a good estimate of the transmitted sample.

After the initial adaptation process, the equalizer may be adapted again (periodically) when the known symbol is received again (e.g., in a subsequent pilot reference period). In addition, equalizer 310a may adapt with an adaptation scheme for the decisions (e.g., in practical applications, where the actual symbols are typically unknown). Adaptation to the decision may be done based on the error between the hierarchical symbol estimate (i.e., the expected signal level of the symbol estimate) and the symbol estimate (i.e., the actual level at which the symbol was received).

The first adaptation scheme performs the adaptation in the time domain and the spatial domain separately and sequentially, which reduces the required computational complexity. For example, if each filter 410 comprises (2M +1) coefficients, the adaptation in the time domain is achieved by performing operations on vectors of length 2M +1 and (possibly) 2M +1 dimensional matrices. If K received signals are processed in parallel and the operations are performed on K signal processing paths, the adaptation of the spatial domain can be achieved by operating on vectors of length K and possibly on K dimensional matrices.

In a second adaptation scheme, the coefficients and scale factors in the equalizer 310a are adapted in both the time and spatial domains. In the present embodiment, all filters 410a to 410k are operated in parallel to apply a corresponding sample stream x₁(n) to x_K(n) filtering. Filters 410a through 410a can be calculated from the following equation410k generated symbol estimatesTo

To equation (14)

Symbol estimationToThen to respective multipliers 412a through 412k, with respective scaling factors s₁(n) to s_K(n) scaled and combined to produce an output symbol estimateThis can be calculated by equation (11).

The symbol estimates can be computed as shown in equations (12) or (13)And actual symbols y (n) or layered symbol estimationWith a corresponding error e (n). Then, the K groups of filter coefficients are simultaneously adapted by using the error e (n)c ₁(n) toc _K(n) (K scaling factors s)₁(n) to s_K(n) remain constant). The K sets of filter coefficients can be concatenated into a vector of length K (2M +1) which is adjusted by the error e (n). This embodiment adapts up to K · (2M +1) variables simultaneously.

In a second adaptation scheme, the filter coefficients are adapted simultaneously by operating on a vector of length K (2M +1) and a (possibly) K (2M +1) -dimensional matrix. The required computation is more complex than for the first adaptation scheme described above. However, because all coefficients can be considered together to equalize the received signal, a symbol estimate is outputThe value of the actual symbol y (n) is well estimated and thus improved performance can be achieved. For example, the filter coefficients for antenna 1 can be based on signals received through other antennas.

The adaptation of the filter coefficients and the scaling factors may be achieved with the aid of a rake receiver 330. In a typical design, rake receiver 330 includes a searcher unit and finger processors. The searcher unit may process samples of a particular received signal at different time offsets and search for the strongest multipath under the control of controller 370. For CDMA systems, the searcher unit typically correlates the received samples with the (complex) PN sequence used to spread the samples in the transmitter unit. The searcher unit may be designed to correlate with a particular time offset or multiple time offsets. Each time offset is equal to a specific delay of the PN sequence associated with the absolute PN sequence (zero delay). The searcher unit may also be designed to produce (complex) correlation results for each time offset measured, or only correlation results that exceed a particular threshold. The searcher unit or controller 370 may be designed to hold a list of the correlation results for each received signal and their respective time offsets.

Controller 370 may identify the strongest multipath (e.g., the multipath with the greatest amplitude or energy) from the list of correlation results for all received signals. The functions performed by the searcher unit and finger processor are described in further detail below.

In one embodiment, the filter coefficients are initialized based on the correlation results of the rake receiver 330. For each received signal, the rake receiver 330 can be used to search for the strongest multipath. The magnitude of the correlation result for the strongest multipath of each received signal may be ranked. The index symbol Ji corresponding to the strongest multipath of a particular signal may be determined by:

equation (15)

The strongest multipath for a particular received signal can then be determined by:

equation (16)

Wherein | α_i，jI is the amplitude of the jth multipath of the ith received signal, | α_JiI is the strongest of the ith received signalThe magnitude of the diameter. Similarly, the strongest multipaths of all received signals may be ranked, where the strongest multipath may be determined by:

equation (17)

Wherein | α_JAnd | is the amplitude of the strongest multipath of all received signals.

Once the strongest multipath of all signals is determined, the time offset relative to that multipath can be identified. Then, the coefficient c of each filter 410_i，0(n) may be initialized to one of: (1) the value associated with the "finger value", which represents the quality of the received signal of the strongest multipath (e.g.,γ is a constant that depends on the variance of the noise), (2) the value 1.0, or some other value. Each remaining coefficient is initialized to zero (i.e., c)_i，-M(n)＝...＝c_i，-1(n)＝c_i，1(n)＝...＝c_i，M(n) ═ 0.0). Corresponding to the strongest multipath alpha_JiTime offset τ of_JiCan be divided into a "coarse" part and a "fine" part. The coarse portion may be used as a coarse tune to properly generate the actual symbols y (n) used to adapt the coefficients and scaling factors, as described below. The fine part may be used as a fine-tune to determine the received samples x_i(n) time history. In particular, the digital processor 230 uses the precision part to adjust the timing of the re-sampling clock to generate the received samples x that are positioned by the time offset_i(n) of (a). Generating the time offset for the actual symbols y (n) also takes into account each filterThe number of taps and the initial value of the coefficients of the filter 410.

The scaling factor may also be initialized by the rake receiver 330. For example, each scaling factor can be initialized with a value related to the amplitude of the strongest multipath of the received signal. Thus, the scaling factor s can be adjusted₁(n) to s_KThe values of (n) are respectively set to be alpha_J1To alpha_JKThe value of interest. In addition, each scaling factor may be set to a particular value (e.g., 1/K) prior to the adaptation process. Other methods of initializing coefficients and scaling factors are also within the scope of the present invention.

Fig. 4B is a block diagram of another embodiment of an equalizer 310B that can also be used to implement the equalizer 310 shown in fig. 3. Unlike the equalizer 310a shown in fig. 4A, which performs equalization first and then spatial domain combining, the equalizer 310B shown in fig. 4B, which performs spatial domain combining first and then equalization. The combination of spatial domains can be viewed as a process of forming beams with an array of K antennas 132. The structure of equalizer 310b is simpler than that of equalizer 310a, and its dimensionality is reduced to approximately 1/K of the latter (i.e., equalizer 310b has (2M +1) + K variables, while equalizer 310a has K (2M +1) + K variables). Equalizer 310b may produce a good estimate under certain operating circumstances (e.g., when the scattering is not large and when the antennas have nearly identical frequency responses with some phase offset).

As shown in fig. 4B, the signal received by each antenna 132 is processed by a corresponding pre-processor 210 to produce a stream of samples x_i(n) of (a). In the equalizer 310b, the sample streams x generated by the preprocessors 210a to 210k₁(n) to x_K(n) to respective multipliers 422a to 422k, which also receive respective scaling factors s₁(n) to s_K(n) of (a). Each multiplier 422 uses a scaling factor s_IFor received sample x_i(n) are scaled and the scaled samples are provided to adder 424. The scaling factor of the multiplier 422 is typically complex to account for complex combinations of signals received through the multiple antennas 132. Adder 424 is coupled from multipliers 422a through 422kThe scaled samples are received and summed to produce spatially combined samples x (n) and provided to filter 410 x.

Filter 410x equalizes samples x (n) according to a particular set of coefficients. Symbol estimates generated by filter 410xAs the output of equalizer 310b and is provided to coefficient adjustment unit 420 b. The slicer 418b may be used in the filter 410x310b to receive symbol estimatesAnd layered to generate layered symbol estimates

Coefficient adjustment element 420b also receives actual symbols y (n), the symbol estimatesAnd hierarchical symbol estimationOr a combination thereof. Coefficient adjustment unit 420b then scales the coefficients of filter 410x and scaling factor s based on the received symbol estimates and symbols₁(n) to s_K(n) adapting. The coefficient adjustment unit 420b can also be designed to implement LMS, RLS, DMI or some other algorithm or combination thereof. The coefficient adjustment unit 420b is implemented similarly to the coefficient adjustment unit 420a shown in fig. 4A.

The adaptation of the scaling factors and coefficients in the equalizer 310b is achieved using a similar adaptation scheme to that described above for the equalizer 310 a.

In a first adaptation scheme, time-domain adaptation is performed first and then spatial-domain adaptation is performed. For time-domain adaptation, the scaling factors are initialized to a set of specific values and then the filter coefficients are adapted. The initial value of the scaling factor can be estimated using the direction of arrival (DOA), which is known in Radar theory and described in the book entitled "Adaptive Radar Detection and Estimation" by John Wileyand Sons, 6 months 1992, by s.haykin and a.steinhardt. Alternatively, each scaling factor may be initialized, such as to 1/K. Maintaining the scaling factor constant, the filter coefficients may be adapted using LMS, RLS, DMI, or some other algorithm, similar to that described above.

Once the first iteration of time-domain adaptation is complete, the coefficients can be kept constant and spatial-domain adaptation can be performed (e.g., for the same pilot reference or for different pilot references). For adaptation in the spatial domain, symbol estimates are computedAnd used to adapt the scaling factor, again using LMS, RLS, DMI, or some other algorithm, similar to that described above. The adaptation process of the time domain and the space domain is iterated several times to achieve the desired result. Similar to that described in equalizer 310a, rake receiver 330 can be used to initialize the coefficients and scaling factors for filter 410 x. For example, rake receiver 330 is used to search for the strongest multipath of each received signal being processed. The strongest multipath of all received signals can be identified according to equation (17). The coefficients of filter 410x may be initialized based on the magnitude of the strongest multipath. The actual symbol y (n) used for adaptation may be generated with a time offset corresponding to this multipath.

Likewise, the strongest multipath of each received signal may be used to initialize the corresponding scaling factor. For example, the scaling factor s may be initialized based on the magnitude of the strongest multipath of the signal received by antenna 132a₁The scaling factor s may be initialized based on the magnitude of the strongest multipath of the signal received by antenna 132b₂And so on. The initial value of the scaling factor may also be set according to an index value representing the received signal quality (S/N).

For some CDMA systems, such as the HDR system, the pilot reference is time division multiplexed with other data and transmitted from the base station to the remote terminal. For these CDMA systems, the filter coefficients and scaling factors may be adapted using the transmitted pilot reference. Thereafter, the coefficients and scaling factors are maintained constant and used to process the data transmitted during the time period between pilot references.

Fig. 6 shows the format of a data frame for forward link transmission in an HDR CDMA system. On the forward link, traffic data, pilot reference, and signaling data are time division multiplexed within one frame and transmitted from the base station to the remote terminal. Each frame covers a unit of time called a slot (e.g., 1.67msec designed for a particular HDR). Each time slot includes traffic data segments 602a, 602b, and 602c, pilot reference segments 604a and 604b, and signaling data (OH) segments 606a and 606 b. Traffic data segment 602 and pilot reference segment 604 are each used to convey traffic data and pilot reference. The signaling data section 606 is used to convey signaling information, such as a forward link active (FAC) flag, a reverse link busy flag, a reverse link energy control command, and so forth. The FAC flag indicates whether the base station is to transmit traffic data in a certain number of slots in the future. The reverse link busy flag indicates whether the reverse link capacity limit of the base station has been reached. The power control commands direct the transmitting remote terminals to increase or decrease their transmit power.

According to the HDR CDMA system, traffic data is covered by Walsh codes corresponding to channels to be used to transmit the data before transmission, and energy control data of each remote terminal is covered by Walsh codes assigned to the remote terminal. The pilot reference, overlaid traffic data and energy control data are then spread with a complex PN spreading sequence obtained by multiplying the short IPN and QPN spreading sequences assigned to a particular transmitting base station by the long PN sequence assigned to the receiving remote terminal. At the highest data rates, the bit rate matches or exceeds the chip rate of the PN spreading sequence and Walsh code, and therefore direct sequence spreading of the data cannot be achieved. The format of the data frames for forward link transmission in HDR systems and their processing is described in further detail in the aforementioned U.S. patent application serial No. 08/963,386, currently U.S. patent 6,574,211.

For some CDMA systems, e.g. IS-95CDMA systemOne pilot reference is Code Division Multiplexed (CDM) with other data and transmitted from the base station to the remote terminal. For these CDMA systems, the filter coefficients and scaling factors may be adapted with the transmitted CDM pilot reference. However, since the pilot reference is code division multiplexed with other data, to extract the transmitted pilot reference, it is necessary to estimate each antenna symbolOr output symbol estimationAdditional processing is performed.

For the IS-95CDMA system, the pilot reference (full one sequence) IS covered by a 64-chip full zero Walsh sequence and combined with other covered data by a PN sequence. By processing the output symbol estimates by complementary methodsThe transmitted CDM pilot reference may be recovered. As shown in FIG. 5B, which will be described in detail below, in post-processor 320, symbol estimationDespread by the PN sequence, decovered by the pilot Walsh sequence, and accumulated over a 64-chip pilot symbol period to produce recovered pilot symbolsThe decovering process removes data on other traffic channels that are covered by Walsh sequences that are orthogonal to the pilot Walsh sequence.

Recovered pilot symbolsThe error e (n) between the actual (expected) pilot symbols p (n) can be given by:

equation (18)

The error e (n) may then adapt the filter coefficients and the scaling factors using any combination of the adaptation schemes and adaptation algorithms described above (e.g., LMS, RLS, or DMI). For the equalizer 310a shown in fig. 4A, the symbol estimates for each antenna from each filter 410 may also be usedAn adaptation process is performed. Symbol estimationMay be processed by post-processor 320 in the manner described above to generate recovered pilot symbolsRecovered pilot symbolsAnd the expected pilot symbol p_iError e between (n)_i(n) can be derived from the following formula:

equation (19)

The recovered symbols may generally be computed at any particular point in the receiver unit (e.g.,orAnd the expected symbol (e.g., y (n) or p (n)). Is suitably processedSo as to compare "equal" or "similar" symbols.

For a CDMA system that transmits a CDM pilot reference, the error e (n) should be calculated for each pilot symbol period (e.g., every 64 chips for an IS-95CDMA system). If the CDM pilot reference IS sent continuously, such as for an IS-95CDMA system, the adaptation IS done in a continuous manner despite the lower pilot symbol rate. For a CDMA system that transmits a TDM pilot reference, the error per chip period may be calculated. However, the adaptation process using TDM pilot reference is typically only performed in the pilot reference period.

As described above, the information obtained from the rake receiver may be used to (1) initialize the filter coefficients prior to the first adaptation process, (2) appropriately generate the desired symbols y (n) or p (n), and (3) determine the received samples x as described above_i(n) time history. For the equalizer 310a in FIG. 4A, the center coefficient c of each filter 410 may be set_i，0(n) the finger values for the strongest multipath for the signal being processed by the filter are initialized (e.g.,time offset tau for the strongest multipath_JiCan be used to generate expected symbols y (n) or expected pilot symbols p (n) and to generate received samples x that are positioned in time by a time offset_i(n) of (a). The expected symbol y (n) or p (n) can also be calculated when considering the delay associated with the processing element up to the point where the error is calculated. E.g. for a certain time offset tau_JiThe delay associated with the desired symbol y (n) is used to generate the desired pilot symbol p (n) to account for the delay of the post-processor 320.

The filter coefficients and scaling factors may also be adapted with the received data. The received samples are processed, decoded, and CRC checked to determine if the packet was received without error. Correctly received data packets canIs re-encoded and re-processed in a similar manner as at the transmitter unit. The regenerated symbols and recovered symbols (e.g.,by comparison, the error between the two is used to adapt the filter coefficients and the scaling factors. The recovered symbols are appropriately buffered to account for delays incurred by the decoding, re-encoding and re-processing processes.

Fig. 7 is a block diagram of one embodiment of a rake receiver 330. Due to multipath and other phenomena, the transmitted signal may reach the receiver unit through multiple signal paths. To improve performance, rake receivers are designed to process multiple (and strongest) instances of the received signal (or multipath). For a conventional rake receiver, several finger processors 710 are employed to handle several multipaths. Each finger processor 710 consists of one finger of the rake receiver, which can be assigned to process a particular multipath.

As shown in FIG. 7, I received from a particular preprocessor 210_INAnd Q_INThe samples are sent to several finger processors 710a through 710 l. In each assigned finger processor 710, the received I_INAnd Q_INThe samples are provided to a PN despreader 720, which also receives the complex PN sequences PNI and PNQ. The complex PN sequence is generated according to the particular design of the CDMA system being implemented, and for HDR systems, is obtained by multiplying the short IPN and QPN sequences with the long PN sequence using multipliers 738a and 738 b. Short IPN and QPN sequences are used to spread data from a transmitting base station, and long PN sequences are assigned to a receiver unit at the receiver and used to scramble the data. The IPN and QPN sequences are generated with a time offset corresponding to the particular multipath being processed by the finger processor.

PN despreader 720 for complex I_INAnd Q_INComplex multiplication of the samples with a complex PN sequence and despreading the complex I_DESAnd Q_DESThe samples are sent to decover units 722 and 732. Decover unit 722 decovers the despread samples using one or more channelization codes (e.g., Walsh codes) used to cover the data and generatesAnd repeating the covering sampling. The decovered samples are then provided to a symbol accumulator 724, which accumulates the samples over the entire channelization code length, thereby producing decovered symbols. The decovered symbols are provided to pilot demodulator 726.

For an HDR system, the pilot reference is transmitted during part of the forward link transmission. The decovering unit 732 then decovers the despread samples using a particular channelization code (e.g., Walsh code 0 for HDR systems) that is used to cover the pilot reference at the base station. The decovered pilot samples are then provided to accumulator 734 and accumulated over a particular time interval to generate pilot symbols. The accumulation time interval may be the duration of the pilot channelization code, the entire pilot reference period, or some other time interval. The pilot symbols are then provided to a pilot filter 736 and used to generate pilot estimates, which are provided to a pilot demodulator 726. The pilot estimate corresponds to an estimated pilot symbol or a predicted pilot symbol for a time period in which data is present.

Pilot demodulator 726 coherently demodulates the decovered symbols from symbol accumulator 724 with pilot estimates from pilot filter 736 and provides the demodulated symbols to symbol combiner 740. Coherent demodulation can be achieved by dot and cross product of the decovered symbols with pilot estimates. Dot and cross products effectively phase demodulate the data and further scale the synthesized output with the relative strength of the decovered pilots. The scaling with the pilot effectively weights the effects of different multipaths according to multipath quality for efficient combining. Thus, the dot and cross products perform the dual roles of phase prediction and signal weighting, which are both characteristic of coherent rake receivers.

A symbol combiner 740 receives the demodulated symbols from all assigned finger processors 710 and coherently combines them to produce recovered symbols for the particular received signal being processed by the rake receiver. The recovered symbols for all received signals are then combined as described below to generate a total recovered symbol that is sent to subsequent processing units.

The searcher unit 712 should include a PN despreader, a PN generator, and a signal quality measurement unit. The PN generator may be controlled by a controller 370 for searching for the strongest multipath to generate complex PN sequences at different time offsets. For each time offset to be searched, the PN despreader receives the I with a complex PN sequence of a particular time offset_INAnd Q_INSamples are sampled and despread to produce despread samples. The signal quality of the despread samples can then be estimated. This can be done by calculating the energy (i.e., I) of each despread sample_DES ²+Q_DES ²) And accumulating energy in a particular time period (e.g., a pilot reference period). The searcher unit searches at many time offsets and selects the multipath with the highest signal quality measurement. The available finger processors 710 are then assigned to process these multipaths.

The design and operation of a rake receiver in a CDMA SYSTEM is described in further detail in U.S. patent No. 5,764,687 entitled MOBILE DEMODULATOR or receiver FOR compressed speech COMMUNICATION SYSTEM and U.S. patent No. 5490165 entitled MOBILE DEMODULATOR ELEMENT ASSIGNMENT IN A SYSTEM CAPABLE reception SIGNALS. Both of which are incorporated herein by reference and assigned to the assignee of the present invention.

In the embodiment shown in fig. 2A, several forward link signals are received by antennas 132A through 132k and processed by respective pre-processors 210a through 210k to generate a sample stream x₁(n) to x_K(n) of (a). Thus, several rake receivers may be provided to process the K sample streams. The symbols recovered from the overall processed received signal are then combined by a combiner. Alternatively, one or more rake receivers may be time division multiplexed to process the K sample streams. In this TDM rake receiver architecture, samples in the K flows may be temporarily stored in buffers and later retrieved and processed by the rake receiver.

For each received signal, rake receiver 330 processes up to L multipaths, where L represents the number of available finger processors 710. Each of the L multipaths corresponds to a particular time offset that is helped to be identified by the searcher unit 712. Controller 370 or searcher unit 712 should hold the strongest multipath (α) for each of the K received signals being processed_Ji) And its corresponding time offset (τ)_i) A list of (a). These amplitude and time offsets may be used to initialize the coefficients and scaling factors of equalizer 310, as described above. In one implementation, the magnitude of each multipath of interest may be calculated by dividing the square of the accumulated energy value by the number of samples (N) used in the accumulation.

FIG. 5B is a block diagram of one embodiment of post-processor 320 that processes symbol estimates generated from a received signalPost processor 320 is used to undo PN spreading and covering processes that may have been performed at the transmitter unit. For example, HDR systems perform PN spreading and covering of data at a lower data rate before transmission. With higher data rates (matching or exceeding the chip rate), the PN spreading and covering process may result in symbol inversion (i.e., a change in polarity) but no direct sequence spreading. Thus, the post-processor 320 can perform PN despreading and decovering processes at lower data rates and symbol inversion at higher data rates. In general, post-processor 320 is designed to perform functions that are complementary to those performed at the transmitter unit.

In post-processor 320, the symbol estimates are provided to a PN despreader 520, which also receives a complex PN sequence having the same time offset as that used to adapt the filter coefficients. PN despreader 520 despreads the symbol estimates with the complex PN sequence, producing despread samples, which are provided to decover unit 522. The decover element 522 then decovers the samples using one or more channelization codes used to cover the base station data. The decovered samples for each channel are accumulated throughout the length of the channelization code by a symbol accumulator 524 to generate recovered symbols for subsequent processing units.

The equalizer 310, post-processor 320, and rake receiver 330 may use a set of processing elements to time-division multiplex and process the samples of all received signals. Also, some elements of the post-processor 320 are similar to those of the rake receiver 330. Thus, the post-processor 320 and the rake receiver 330 may implement at least some of the same functions via a shared set of processing elements.

In the adaptation period of the filter 410 shown in fig. 4A and 4B, the actual symbols y (n) may be generated based on the time offsets of the strongest multipaths of all received signals. For an HDR system, the pilot reference is an all-one sequence, and the channelization code used to cover the pilot reference is Walsh code 0 (i.e., an all-zero sequence). The complex PN sequence is then used to despread the pilot reference. Thus, the pilot reference and the complex PN sequence being transmitted are equal in duration of the pilot reference transmission.

At the receiver unit, the actual symbol y (n) in the pilot reference period may be generated as the complex PN sequence corresponding to the time offset of the strongest multipath of all received signals (i.e., y (n) ═ pni (n) + jpnq (n)).

Referring again to fig. 3, rx data processor 136 includes two signal processing paths that may be used to process the received signal. The first signal processing path includes an equalizer 310 and a post-processor 320 and the second signal processing path includes a rake receiver 330. In one embodiment, the two signal processing paths may operate in parallel (e.g., during an adaptation period) and a signal quality estimate for each signal processing path can be calculated. The signal processing path with the better signal quality estimate is then selected to process the received signal.

For a conventional rake receiver, the received signal quality can be estimated by calculating the signal-to-noise ratio (S/N). For a CDMA system that transmits a TDM pilot reference, the S/N may be calculated at the pilot reference period when the received signal is known. A signal quality estimate is generated for each assigned finger processor. The estimates for all assigned finger processors can then be weighted and combined to generate the total S/N, which can be calculated by:

equation (20)

Where β is a weight factor used by the rake receiver to combine demodulated symbols from the assigned finger processors to produce recovered symbols, which are an improved estimate of the transmitted data. Es is the energy per symbol of the desired signal (e.g., pilot signal) and Nt is the total noise of the received signal being processed by the finger processor. Nt generally includes thermal noise, interference from other transmitting base stations, interference from different multipaths of the same base station, and other components. The energy per symbol can be calculated by:

equation (21)

Wherein P is_IAnd P_QIs an in-phase filtered and quadrature filtered pilot symbol, N_SYMIs the number of symbols that accumulate the energy and produce the value of Es. Referring to fig. 7, filtered pilot symbols may be generated by accumulating the despread samples during the entire length of the channelization code used to cover the pilot reference. The total noise can be estimated as the energy variation in the desired signal energy and can be calculated as follows:

equation (22)

Measurement of received signal quality is described in further detail in U.S. patent No. 5903554 entitled "METHOD AND APPARATUS FOR MEASURING LINK quality a stream characteristics SYSTEM" AND U.S. patent No. 5799005 entitled "SYSTEM AND METHOD FOR DETERMINING RECEIVED level POWER AND PATH LOSS in CDMA characteristics SYSTEM". Both of these patents are assigned to the assignee of the present invention and are incorporated herein by reference.

For signal processing paths including equalizer 310, different criteria including Mean Square Error (MSE) may be used to estimate signal quality. Also, for a CDMA system that transmits a TDM pilot reference, the MSE may be estimated during the pilot reference period and can be calculated as follows:

equation (23)

Wherein N is_SAMIs the number of samples that accumulate the error and produce the MSE. In general, the mean square error is averaged over several samples under one or more pilot references to obtain the desired measurement confidence. The mean square error is then converted to an equivalent signal-to-noise ratio, expressed as follows:

linear squareJourney (24)

S/N of the signal processing path where equalizer 310 is located_EQMay be associated with the S/N in the signal processing path of the rake receiver 330_RAKEAnd comparing to select a signal processing channel which can provide better S/N to process the received signal.

Alternatively, the MSE (using equation 23) determined for the signal processing path in which rake receiver 330 is located may be compared to the MSE determined for the signal processing path in which equalizer 310 is located, thereby selecting the signal processing path having the better MSE.

For an HDR system, estimating the S/N at the remote terminal is used to determine the maximum data rate that the remote terminal can receive under operating conditions. The maximum data rate is then sent back to the base station of the S/N estimation object. Thereafter, the base station transmits to the remote terminal up to the identified maximum data rate band.

The data rate in a data transmission can be estimated in a number of ways in accordance with the invention. In one approach, the S/N may be estimated for a Rake receiver or equalizer based on the calculated MSE, as shown in equation 24. The best S/N among all signal processing channels can then be used to determine the maximum data rate that can be supported. Alternatively, MSE can be used to directly determine the maximum data rate. The optimal S/N, MSE or maximum data rate may be sent to the base station.

Under certain operating conditions, the signal processing path in which the equalizer resides may exhibit better performance than the signal processing path in which the rake receiver resides. For example, the signal processing path in which the equalizer resides generally performs better when the S/N is high and for channels in which ISI is present. Rake receivers are used to handle multipath, which can also cause ISI. In practice, the rake receiver can be viewed as a filter having L taps (L corresponding to the number of finger processors), each tap corresponding to an adjustable time delay. However, rake receivers do not effectively reduce ISI caused by frequency distortion in the received signal.

The equalizer can more effectively reduce ISI caused by frequency distortion. This is achieved by producing a response that is approximately opposite to the frequency distortion when trying to minimize the total noise including ISI. The equalizer thus "inverts" the channel and attempts to eliminate multipath effects. In practice, each filter 410 is equivalent to a finger processor when the filter coefficients are initialized to { 0., 0, 1, 0., 0 }. Then, when the zero-valued coefficients are adapted, the frequency response of the filter is changed to equalize the channel distortion. Therefore, the equalizer can effectively cope with the multipath-induced ISI and the channel-induced ISI.

For simplicity, many aspects and embodiments of the present invention have been described for a spread spectrum communication system. However, many of the principles of the invention described herein can also be applied to non-spread spectrum communication systems, as well as communication systems that can selectively implement direct sequence spreading, such as HDR systems.

The filter 410 shown in fig. 4A and 4B may be designed to be any length (i.e., with any number of taps and coefficients). More taps allow the filter 410 to better correct for frequency distortion in the received signal and to handle multipaths with larger time offsets. However, more taps means more complexity, more computation of the equalization taps, and possibly longer convergence time. The number of taps is therefore a design choice and is selected based on several factors. These factors include cost, performance, complexity, etc. For example, it may be desirable to equalize over a particular time window (e.g., 20 microseconds), in which case the number of taps is determined by the sampling rate f_SAMP. Any number of taps may be used and this is within the scope of the invention.

The processing elements described herein (e.g., filter 410, equalizer 310, post-processor 320, rake receiver 330, etc.) may be implemented in a number of ways, such as by designing a Digital Signal Processor (DSP), microcontroller, microprocessor, or other electronic circuit in one or more Application Specific Integrated Circuits (ASICs) to perform the functions described herein. Likewise, a processing unit may also implement the functions described herein with a general-purpose or specially designed processor for executing instruction code. Thus, the processing units described herein may be implemented in hardware, software, or a combination of both.

The previous description of the preferred embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without the use of the inventive faculty. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for processing one or more signals in a spread spectrum communication system, the method comprising the steps of:

2. The method of claim 1, wherein the signal quality associated with the first processing is estimated based on a mean square error between the symbol estimates and expected symbols.

3. The method of claim 2, wherein the data rate of the one or more signals is selected based in part on the mean square error.

4. The method of claim 1, wherein for the first processing, the step of processing symbol estimates comprises combining the symbol estimates, and the step of equalizing one or more sample streams precedes the combining step.

5. The method of claim 1, wherein the first processing step further comprises combining the one or more sample streams, the combining step preceding the step for equalizing one or more sample streams.

6. The method of claim 1, further comprising:

the coefficients of each of one or more filters within the equalizer are first adapted, one filter for filtering each of the one or more sample streams.

7. The method of claim 6, wherein the first adapting is performed for each filter based on filtered samples produced by the filter.

8. The method of claim 6, wherein the first adapting is performed for the one or more filters based on the symbol estimates.

9. The method of claim 6, wherein coefficients for each filter are initialized to a set of particular values.

10. The method of claim 6, further comprising the steps of:

identifying a multipath of one of the one or more signals being received and processed, and

the first adaptation is performed based on a time offset corresponding to the identified multipath.

11. The method of claim 6, wherein the first adapting attempts to minimize a mean square error between the symbol estimates and expected symbols.

12. The method of claim 6, wherein the first adapting attempts to minimize a mean square error between the filter-filtered samples and expected symbols.

13. The method of claim 6, further comprising the steps of:

the symbol estimates are layered to generate layered symbol estimates, and

performing the first adaptation using the layered symbol estimates.

14. The method of claim 6, wherein each filter within the equalizer is implemented as a finite impulse response filter.

15. The method of claim 6, wherein the first adapting is performed with a time division multiplexed pilot reference.

16. The method of claim 6, wherein the first adapting is performed using a code division multiplexed pilot reference.

17. The method of claim 6, wherein the first adapting is performed using a least mean square algorithm.

18. The method of claim 6, wherein the first adapting is performed using a recursive least squares algorithm.

19. The method of claim 6 wherein said first adapting is performed using a direct matrix inversion algorithm.

20. The method of claim 6, wherein the step of first processing further comprises combining the one or more sample streams according to one or more scaling factors, wherein each sample stream of the one or more sample streams has one scaling factor.

21. The method of claim 20, further comprising the steps of:

prior to the combining step, performing a second adaptation of the one or more scaling factors.

22. The method of claim 21, further comprising the steps of:

identifying a multipath for each of the one or more signals being received and processed; and

each respective scaling factor is initialized based on the identified multipath.

23. The method of claim 21, wherein the second adapting is performed based on the symbol estimates.

24. The method of claim 4, further comprising the steps of:

first adapting coefficients of each of one or more filters within the equalizer, one filter for filtering each of the one or more sample streams; and

the second adaptation is made to the one or more scaling factors used for the combining.

25. The method of claim 24, wherein the first and second adapting are performed separately and sequentially, wherein the one or more scaling factors are held fixed during the first adapting and coefficients of the one or more filters are held fixed during the second adapting.

26. The method of claim 24, wherein the first and second adapting are performed repeatedly a plurality of times.

27. The method of claim 24, wherein said first and second adapting are repeated a plurality of times over a particular sequence of expected symbols.

28. The method of claim 24, wherein the first and second adapting are performed based on the symbol estimates.

29. A method of processing one or more signals in a communication system, the method comprising the steps of:

adapting coefficients of each of one or more filters within the equalizer;

wherein the coefficients of each filter within the equalizer are adapted using either a time division multiplexed pilot reference or a code division multiplexed pilot reference.

30. A method of processing one or more signals in a communication system, the method comprising the steps of:

adapting coefficients of each of one or more filters within the equalizer;

wherein the coefficients of each filter within the equalizer are adapted using a least mean square algorithm, a recursive least square algorithm, a direct matrix inversion algorithm, or a combination thereof.

31. A receiver unit for processing one or more signals in a communication system, the receiver unit comprising:

wherein the equalizer comprises:

32. An apparatus, comprising:

33. The apparatus of claim 32, wherein the equalizer and rake receiver both operate on pilot samples, whereupon the switch selects either the rake receiver or the equalizer to process data samples based on the first and second signal qualities estimated by the controller.

34. The apparatus of claim 32, wherein the despreader and decover unit is configured to selectively despread and decover based on a data rate of the received signal relative to the signal samples.

35. The apparatus of claim 32, wherein the controller is configured to estimate the first signal quality for the first stream of recovered symbols using a mean square error between the symbol estimates and expected symbols.

36. The apparatus of claim 32, wherein the equalizer is structured to use information from the rake receiver.

37. The apparatus of claim 36, wherein the rake is configured to identify a strongest multipath of the signal samples, and wherein the coefficient adjustment unit is configured to adjust the coefficient based on information associated with the identified strongest multipath.

38. The apparatus of claim 37, wherein the information comprises a time offset of the strongest multipath identified.

39. The apparatus of claim 32, wherein the coefficient adjustment unit is configured to minimize a mean square error between the symbol estimates and expected symbols.

40. The apparatus of claim 32, wherein the coefficient adjustment unit is configured to minimize a mean square error between the filtered sample sum and a desired symbol from the filter.

41. The apparatus of claim 32, wherein the filter comprises at least one of a finite impulse response filter and an infinite impulse response filter.

42. The apparatus of claim 32, wherein the coefficient adjustment unit is configured to adjust the coefficient using one received pilot signal and one expected pilot signal.

43. The apparatus of claim 32, wherein the coefficient adjustment unit is configured to use at least one of time-division multiplexed pilot signals and code-division multiplexed pilot signals.

44. The apparatus of claim 32, wherein the coefficient adjustment unit is configured to use at least one of a least mean square algorithm, a recursive least square algorithm, and a direct matrix inversion algorithm.

45. The apparatus of claim 32, further comprising a plurality of antennas for receiving signals, and wherein the equalizer comprises a plurality of filters, each filter configured to filter samples of the signal received by a particular antenna, and a plurality of scaling units, each scaling unit configured to scale the filtered samples from a particular filter, the coefficient adjustment unit configured to provide scaling factors to the plurality of scaling units.

46. The apparatus of claim 45, wherein the equalizer further comprises an adder to combine outputs of the plurality of scaling units.

47. The apparatus of claim 45 wherein the rake receiver is configured to identify a strongest multipath of the signal samples, the coefficient adjustment unit configured to adjust the scaling factor based on the identified strongest multipath.

48. The apparatus of claim 45, wherein the coefficient adjustment unit is configured (a) to adjust the coefficients if the one or more scaling factors remain fixed, and (b) to adjust the scaling factors if the coefficients of the one or more filters remain fixed.

49. The apparatus of claim 32, wherein the apparatus constitutes a remote terminal.

50. A system, comprising:

a base station for transmitting a pilot signal and a data signal;

51. The system of claim 50 wherein the remote terminal is configured to train the equalizer coefficients with the pilot signal and to demodulate the data signal.