CN101331688A - Receiver with Chip Level Equalization - Google Patents

Abstract

The present invention relates to receiver apparatuses and methods of controlling weight adaptation in a receiver of a code multiplex telecommunications system with orthogonal spreading codes, wherein received discrete time signal samples are chip-level filtered by using a first equalising step. Additionally, the received discrete time signal samples are delayed by a time period corresponding to a data symbol and used in a second equalising step. Symbol estimates obtained from the first equalising step are non-linearly filtered and used as a desired response for the second equalising step in the following symbol period, wherein equaliser weights adapted in the second equalising step are used for the first equalising step. Alternatively, the second equalising step may be dispensed with and weight adaptation may be incorporated in a single equalising step. As an additional or alternative option, a hybrid equaliser architecture may be provided, where the above two-step equalisation is used during an active phase where a channel is allocated, while another weight updating scheme is used during an inactive phase where no channel is assigned. Thereby, detrimental effects of interference power can be reduced at low increase in complexity.

Description

Receiver with Chip Level Equalization

technical field

The present invention relates to a receiver apparatus and method for controlling weight adaptation in a receiver of a code division multiplexing communication system using orthogonal spreading codes. As an example, the present invention relates to a receiver device and weight adaptation method for a High Speed Downlink Packet Data Access (HSDPA) system introduced by the Universal Mobile Telecommunications System (UMTS) Release 5 standard.

Background technique

A Code Division Multiple Access (CDMA) system is based on a digital wideband spread spectrum technique in which multiple independent user signals are transmitted over an allocated available radio frequency band. In CDMA, each user signal includes a different orthogonal code and a pseudo-random binary sequence that modulates the carrier, thereby spreading the spectrum of the waveform, thereby allowing a large number of user signals to share the same spectrum. In the receiver, the user signals are separated using correlators, which are only capable of despreading signals with selected orthogonal codes. Other user signals whose orthogonal codes do not match cannot be despread, and such signals contribute to system noise. The signal-to-noise ratio (SNR) of a system is determined by the ratio of the desired signal power to the sum of the powers of all interfering signals, and is enhanced by the system processing gain and the ratio of the spreading bandwidth to the baseband data rate. In the third generation of Wideband CDMA (WCDMA), different spreading factors and variable user data rates can be supported simultaneously.

By using spreading codes, the frequency band of the transmitted signal is widened to a chip rate, which is greater than the actual data or information symbol rate. For example, if the spreading code used has a length of 8 data symbols (called "chips"), then 8 chips will be transmitted for each data symbol. The orthogonal nature of the spreading codes provides the property of a unique code, which mathematically means that the respective inner product or correlation of the spreading codes used or intended for communication is zero. The orthogonality of the spreading codes ensures that the transmission of the signal or sequence of data symbols respectively encoded by the spreading codes neither generates nor propagates other signals corresponding to other users in the communication system encoded by other orthogonal spreading codes side effect. A receiver looking for a certain spreading code for a certain transmitter sees a signal encoded by an orthogonal spreading code as noise on the radio frequency (RF) channel. Since spreading codes can have different lengths, there must also be orthogonality between spreading codes of different lengths.

The construction of spreading codes can be realized by an Orthogonal Variable Spreading Factor (OVSF) tree as shown in Figure 2, where the abbreviation "SF" indicates the spreading factor that characterizes the length of the spreading code and the level of the OVSF tree. Within each tree level, the available spreading codes have the same length and are orthogonal. The spreading factor can also be expressed by the ratio between the chip rate and the data symbol rate or the ratio between the chip period and the data symbol period. Spreading codes for different users can fall into different levels of the OVSF tree, thereby providing various levels of quality of service (QoS). User symbols may be spread by a spreading factor ranging from 4 to 512.

However, in CDMA systems, generally speaking, due to multipath propagation and frequency-selective fading, the orthogonality between individual user waveforms degrades, and multiple access interference impairs receiver performance. Although the user signals transmitted at the base station (BS) side are orthogonal, due to the multipath effect of the propagation channel between the transmitter and receiver, this orthogonality may no longer exist at the mobile station (MS) front end, causing The reason for multipath effects is that a channel may consist of more than one distinct propagation path for each signal of a user. Thus, multipath is a propagation phenomenon that causes a radio signal to take two or more paths to reach a receiver antenna, so that the radio signal arrives at the receiver with different time delays. Causes of multipath propagation include atmospheric ducting, ionospheric reflection and refraction, and reflections from terrestrial objects such as mountains and buildings.

和接收机

基站BS的发射机

在下行链路或前向链路分别向每个用户站MS₁，...，MS_K发送数据，基站BS₁的接收机在上行链路或反向链路分别向从每个移动用户站MS₁，...，MS_K接收数据。基站BS₁与移动用户站MS₁，...，MS_K之间的大气空间通常为上行链路和下行链路通信提供了多径环境，如图3中箭头所示。Figure 3 shows a typical CDMA communication system comprising a plurality of mobile or subscriber stations MS ₁ , ..., MS _K and enabling a plurality of users (1, ..., K) to communicate with a base station BS ₁ . Each base station BS ₁ and mobile station MS ₁ , ... MS _K includes a transmitter

and receiver

Transmitter of base station BS

Sending data to each subscriber station MS ₁ ,..., MS _K in the downlink or forward link respectively, the receiver of the base station BS ₁ Data is received from each mobile subscriber station MS ₁ , . . . , MS _K on the uplink or reverse link respectively. The atmospheric space between _the base station BS ₁ and the mobile subscriber stations MS ₁ , .

The following three common methods are used to overcome the problem of orthogonality loss or interference, these three methods are:

The first and most straightforward approach is to treat the interference caused by multipath propagation as Additive White Gaussian Noise (AWGN) and implement a conventional Rake receiver to detect Energy in multiple delayed forms of the received signal is harvested to detect user symbols independently of other user symbols.

The second approach is interference suppression, which partially restores orthogonality by using a chip-rate channel equalizer, and again by correlating with a spreading code, estimating a particular user's symbols independently of other users' symbols.

Finally, the third method is interference cancellation (IC). First, the symbols of known effective interfering spreading codes are estimated by one of the first two methods. Thereafter, the estimated symbols are respread, rechannelized, and deleted from the original received signal.

As mentioned above, due to the multipath effects of the propagation channel between the transmitter and the receiver, the orthogonality may not exist at the front end of the mobile station MS. This loss of orthogonality can cause intersymbol interference (also known as multiuser interference or multiple access interference), interchip interference, and intersymbol interference in symbol estimation. Receivers of optimal or sub-optimal type, i.e. multiuser detectors (MUDs) and interference cancellers (ICs), in most cases need to know the signal and channel parameters of all effective users in order to mitigate multipath effects and to Most reliable way to detect desired data flow. However, the possibility to implement a MUD or IC in a mobile station is limited due to its high complexity and the fact that the transmission parameters of all users are usually not known. A very practical and widely exploited sub-optimal solution is a conventional Rake receiver according to the first method above, which performs a matched filtering operation on the codes of the desired users so that the multi-user interference is seen as additive white noise.

However, when using smaller spreading factors to achieve high data rates, such as in HSDPA systems, the performance of the Rake receiver degrades due to the fact that multipath interference becomes significant and the correlation properties of the spreading sequences are destroyed. Therefore, systems with smaller spreading factors consider an equalizer according to the second method described above to restore the orthogonality between users and limit interference so that higher data rates can be achieved. This is especially important for systems like HSDPA that aim to provide very high data rates.

In the UMTS standard, four QoS classes with different delay and different sequence requirements are defined. The four classes are: Conversational classes with low latency and strict order (e.g. voice), Streaming classes with moderate delay and strict order (e.g. video), Interactive classes with moderate delay and order (e.g. web browsing), and Background classes that have no latency guarantees and no order (such as bulk data transfers). Among these business classes, the background class and the interactive class have burst characteristics. The burstiness triggers the idea of users sharing some resources in time, the most important of which are the orthogonal codes for the downlink, and other supporting techniques, extensions, changes and deletions applied on these channels. Thus, HSDPA emerged as a system that can improve downlink data throughput by using a combination of fast physical layer retransmissions and transmissions and link adaptation controlled by the base station (or Node B in UMTS technology). In HSDPA, two main features of WCDMA, variable spreading factor and fast power control, are disabled. They are replaced by adaptive coding rate and adaptive modulation and extensive multicode operation. The spreading factor is fixed at SF=16. Users can use up to 15 codes at the same time, which enables HSDPA link adaptation to have a large dynamic range and maintain good spectral efficiency. The scheduling process is done in the Node B so that capacity can be allocated to a user when necessary and channel conditions make this policy effective.

In order to support new HSDPA functions, two additional types of channels have been introduced. In the downlink direction from the BS or Node B to the MS, one or more shared control channels (HS-SCCH) broadcast the HSDPA channel assignment identification, transport format and hybrid automatic repeat request (HARQ) process identifier. In the uplink direction, the High Speed Dedicated Physical Control Channel (HS-DPCCH) carries HARQ status reports and Channel Quality Indicators (CQI).

The concept of equalization based on the above two methods has been applied to different systems for many years. Therefore, there are multiple equalizer schemes.

As an example, US 6 658 047 discloses an adaptive channel equalizer for a receiver of a CDMA communication system. An estimator for estimating the impulse response of the channel provides a reference for the adaptive equalizer, and the adaptive equalizer operates to estimate the chip sequence transmitted by the channel and restore the orthogonality between the received signals. An adaptive equalizer includes circuitry for using a blind adaptive algorithm, known as the Griffith algorithm, to estimate the sequence of chips transmitted by the channel.

In addition, "Adaptive Chip-Rate Equalization of Downlink Multirate Wideband CDMA" by Schniter P. et al., IEEE Transactions on Signal Processing, Volume 53, Issue 6, June 2005, pp.2205-2215 discloses a method using Multiple Access Interference (MAI) Decision-directed (DD) chip-rate adaptive equalization scheme for filtering and/or cancellation. In acquisition mode, the equalizer is adapted from the start of code or loss of lock using code-multiplexed pilots. The use of MAI filtering yields a 3rd order least mean square (LMS) algorithm which has significant advantages over standard (ie first order) LMS algorithms in non-stationary environments. In tracking mode, decision direction facilitates MAI cancellation in equalizer updates, thereby enhancing performance.

将这个重新加扰的信号从第二均衡器f^H(i)的输出中减去，将延迟N_max个码片的输入信号提供给第二均衡器，其中N_max表示最低速率用户的扩频增益。因此，提供了两个均衡器功能，即自适应更新的N_max延迟均衡器功能f(i-N_max)、以及用于产生符号估计的临时均衡器功能

可以在预测单元2中使用延迟均衡功能f(i-N_max)的N_max步前向预测计算临时均衡器功能

延迟均衡器的适配是基于从延迟均衡器的输出x(i-N_max)中减去重新加扰的输出

来执行的。Fig. 4 shows a schematic block diagram of DD chip rate adaptive equalization described in the above prior art. In DD mode, the receiver makes a temporary hard decision on the symbols of all active users and uses this decision to construct a delayed approximate copy of the transmitted sequence. Next, the transmitted sequence is used to update the adaptive filter at the chip rate. Since it is impossible to make reliable signal estimates without properly tuning the equalizer, the DD mode only works after the previous pilot training mode has converged. According to Fig. 4, the symbol estimation process consists of multiplying the descrambled signal s ^* (iv) to the temporal equalizer descrambling of the output. Next, the output of the matched filter is calculated for each effective user in the descrambling unit 4 , and the output of the matched filter is quantized in the detection unit 6 . The hard symbol estimates and the spreading code are then used to regenerate a delayed approximate copy of the multi-user sequence by respreading in the respreading unit 8 and multiplying by the rescrambling signal s(iN _max -v). The rescrambling operation produces a signal

This re-scrambled signal is subtracted from the output of the second equalizer f ^H (i), and the input signal delayed by N _max chips is provided to the second equalizer, where N _max represents the spread spectrum of the lowest rate user gain. Therefore, two equalizer functions are provided, an adaptively updated N _max delay equalizer function f(iN _max ), and an interim equalizer function for generating symbol estimates

The temporal equalizer function can be computed in prediction unit 2 using N _max steps forward prediction of the delayed equalization function f(iN _max )

The adaptation of the delay equalizer is based on subtracting the rescrambled output from the output x(iN _max ) of the delay equalizer

to execute.

In HSDPA systems, there are two possible phases during which a MS or User Equipment (UE) in UMTS technology can track (estimate) and/or equalize the channel, these two phases are inactive phases and valid phase. The inactive phase (or state) is when the user listens to the channel, but is not assigned a High Speed Downlink Shared Channel (HSDSCH); and the active phase is when at least one HSDSCH code is assigned to the user. The adaptive equalizer described above by Schniter et al. does not provide an optimal solution for the high speed channels provided in HSDPA systems. Since the adaptation branch of the delayed equalizer introduces a large delay, the adapted filter weights or taps cannot be directly used in the filtering operation of the upstream branch of the temporal equalizer. In a very rapidly changing channel, the prediction mechanism of the prediction unit 2 essentially guesses the filter weights of the upstream branch of the temporary equalizer based on the filter weights of the downstream branch of the adaptive equalizer. Furthermore, significant delays are introduced corresponding to the maximum effective spreading factor in the system, which may even reach 512 chips in some cases. Furthermore, in the adaptive equalizer scheme proposed by Schniter et al., it is assumed that all valid codes in the system are known. Therefore, there is a need to detect where in the OVSF hierarchy the valid code is located and estimate its magnitude. However, this is a very complicated process and thus not easy to implement. Even if achieved, problems with false detections, missed detections, and wrongly estimated magnitudes may arise. Furthermore, despreading is performed independently for each active code in each level of the OVSF tree, which results in high computational complexity.

Contents of the invention

Therefore, the object of the present invention is to provide an improved weight adaptation control method at the receiver, by which the adverse effect of interference power can be reduced with little increase in complexity. According to the first aspect, the object of the present invention is achieved by the receiver device as claimed in claim 1 and the weight adaptation control method as claimed in claim 16, respectively.

Accordingly, the delay introduced in the adaptation branch is fixed and reduced to one symbol period of only 16 chips. Therefore, the filter weights can be copied from the equalizer of the adaptation branch to the equalizer of the filtering branch without any prediction. Furthermore, non-linear filtering of symbol estimates can be based on knowledge of specific channel codes, enabling more robust methods and systems.

Furthermore, in the solutions according to claims 3 and 16, only one equalization function or unit is required, which significantly reduces complexity, overhead and power consumption.

According to an additional or alternative second aspect, the above objects can be achieved by the receiver device as claimed in claim 19 and the weight adaptation control method as claimed in claim 21 .

Correspondingly, a hybrid equalizer architecture utilizing a selective weight adaptation scheme is proposed, where the updated algorithm is selected based on the phase of the communication system (i.e. active and inactive). Thus, it is desirable that the filter weights or taps be more reliable at the beginning of the active phase to reduce computational load and complexity. The update rate of the second update scheme used during the non-active phase may be chosen to be a lower value than the update rate used during the active phase. The selection of the second weight update scheme must be done in such a way that the channel tracking capability is not relaxed.

In the specific example of an HSDPA system, the time period may correspond to a fixed symbol length of 16 chips for an orthogonal spreading code. Therefore, only a short delay of 16 chips is introduced so that the updated weights in the adaptation branch can be immediately used in the first equalizer means of the filtering branch.

The filtering branch may comprise means for despreading the descrambled and equalized signal samples by employing a Fast Walsh Hadamard Transform (FWHT) at the level of a single code tree, wherein the despread signal samples are provided to the feedback means. Utilizing a fixed level of FWHT offers the advantage of reducing despreading complexity. In the example of a fixed symbol length of 16 chips, the despreading complexity can be reduced to 1/4. The estimate of the filtered signal fed back by the feedback means may be respread again using the FWHT in the spreading means.

Furthermore, subtraction means may be provided for obtaining the difference between the equalized signal samples output from the second equalizer means and the feedback signal samples obtained from the feedback means, and supplying this difference to updating means, which update means uses for adapting the equalizer weights of the second equalizer means. Thereby, the estimated chip-level signal can be used as a kind of training sequence or desired signal for the second equalizer means located in the adaptation branch. In an alternative implementation of a single equalizer, the subtraction means may be arranged to obtain the output of the equalizer means and delayed via another delay means arranged to delay the filtered signal samples by a time period corresponding to a data symbol The difference between the filtered signal samples and the feedback signal samples obtained from the feedback means is then provided to update means for adapting the equalizer weights of the equalizer means.

Furthermore, selection means are provided for selecting the second equalizer means during an active phase in which at least one channel code is assigned to the receiver device and for selecting the other weight updating means during an inactive phase in which no channel code is assigned to the receiver device . Such a hybrid equalizer architecture offers the advantage of reducing computational load by using different update mechanisms and supporting desired signal or statistical properties during the inactive and active phases of the channel. In particular, the other weight updating means may be arranged to operate based on a direct comparison of the input and output of the first equalizer means. As a specific but not limited example, other weight updating means can be configured to operate based on the Griffith algorithm.

Alternatively, the feedback means may be arranged to classify the symbol estimates obtained in the filtering branch into: a first group of branches which should carry the most reliable estimate of the downlink shared channel to be hard detected, the first output The second group of branches excluded from feedback and replaced by a known constant sequence or to be fed back scaled by linear minimum mean square error (LMMSE) weighting, and the remaining third branch fed back again scaled by LMMSE weighting Group. If the estimated LMMSE weights are negative at any particular branch, they are replaced by zero. This last case is equivalent to blocking the feedback of these branches, and this occurs when the power of a particular branch is below a predetermined threshold σ _th ² . Such a hybrid mechanism provides the advantage of only explicitly using the knowledge of a specific downlink shared channel, thereby improving the robustness of the system. In particular, the predetermined threshold applied to the third branch group may correspond to the average energy level between the hard detection and soft detection values of all downlink shared channels. Averaging across downlink shared channel codes reduces the variance of the estimation error.

LMMSE weighting is a measure of the signal-to-interference-plus-noise ratio (SINR). With this measure, a clear and reliable measure is provided for a mixture of hard linear decisions and weighted linear decisions.

而预定阈值如之前定义的为σ_th ²。因此，该分支上的LMMSE权值为 $c_k=\frac{{\left|{\hat{s}}_k\right|}^2-{\mathrm{\σ}}_{\mathrm{th}}^2}{{\left|{\hat{s}}_k\right|}^2}.$ 权值中的分子项对应于有用信号功率，分母项对应于有用功率和噪声加干扰功率之和。The LMMSE weights for the remaining code branches can be calculated as follows. Suppose the instantaneous power of any branch k is

And the predetermined threshold is σ _th ² as defined before. Therefore, the LMMSE weight on this branch is $c_k=\frac{{\left|{\widehat{\mathrm{the\; s}}}_k\right|}^2-{\mathrm{\&sigma;}}_{\mathrm{the\; <figure-callout\; id="2"\; label="th"\; filenames="A20068004756500251.png,A20068004756500261.png"\; state="\{\{state\}\}">th</figure-callout>}}^2}{{\left|{\widehat{\mathrm{the\; s}}}_k\right|}^2}.$ The numerator item in the weight value corresponds to the useful signal power, and the denominator item corresponds to the sum of useful power and noise plus interference power.

The multiplexing mechanism between hard pilot addition and scaled linear feedback of the first branch requires a threshold. If the pilot tone power is P _CPICH , the calculated optimal threshold is $P_{\mathrm{thr}}=P_{\mathrm{CPICH}}+2*{\mathrm{\&sigma;}}_{\mathrm{the\; <figure-callout\; id="2"\; label="th"\; filenames="A20068004756500251.png,A20068004756500261.png"\; state="\{\{state\}\}">th</figure-callout>}}^2.$ If the power of the first branch is less than P _thr , the pilot signal is added to the feedback path in a hard way. Otherwise, the output of the first branch is fed back by LMMSE scaling like the other remaining branches.

Further advantageous developments are defined in the dependent claims.

Description of drawings

Now, based on a preferred embodiment, the invention is described with reference to the accompanying drawings, in which:

Figure 1 shows a schematic block diagram of a receiver architecture in which the present invention can be implemented;

Figure 2 shows a graphical representation of an Orthogonal Variable Spreading Factor (OVSF) tree;

Figure 3 shows a typical structure of a multi-user communication system;

Fig. 4 shows a schematic block diagram of a decision-oriented chip rate adaptive equalizer according to the prior art;

Figure 5 shows a time diagram of the non-valid phase and the valid phase with corresponding update algorithms according to a preferred embodiment;

Figure 6 shows a block diagram of an equalizer architecture according to a first preferred embodiment;

Figure 7 shows a diagram related to the selective non-linear filtering operation in the preferred embodiment; and

Fig. 8 shows a block diagram of an equalizer architecture according to a second preferred embodiment.

Detailed ways

In the following, preferred embodiments will be described based on the HSDPA data access system according to the UMTS standard Release 5 specification. HSDPA has been developed to provide high speed data rates in the downlink direction. Due to this feature and due to the dispersion properties of the HSPDA channel, the traditional Rake receiver according to the first method above is no longer considered, while the equalizer scheme according to the second method above is considered as a key solution.

Fig. 1 shows a schematic block diagram of a receiver 10 in which the preferred embodiment of the invention may be implemented. The receiver 10 has an adaptive interference suppression algorithm based on channel equalization and adapted to synchronous CDMA systems using orthogonal spreading codes in a scrambling manner. In particular, the receiver 10 does not necessarily have any training sequence or training information for adaptive equalization. The receiver 10 only needs initial weight values, which may or may not require a channel estimation scheme requiring training sequences.

According to Fig. 1, at least one of the antennas ₁₀₀₁ to _100N receives a signal from a communication channel. This signal is coupled to a conventional RF transceiver 110 that includes analog-to-digital (A/D) conversion. The conventional RF transceiver 110 optionally performs chip waveform filtering. The converted signals r ₁ to r _N are forwarded to the channel impulse response estimator 120 and the adaptive chip estimator 130 . The channel impulse response estimator 120 is operative to estimate the impulse response of the channel and provides reference input factors or weights h ₁ to h _N to the adaptive chip estimator 130 to provide initial weights. The output of adaptive chip estimator 130 is coupled to symbol sync code correlator 140 . The correlator 140 despreads the output d of the adaptive chip estimator 130 by multiplying the output d of the adaptive chip estimator 130 by the output of the code generator 150, and integrates over a symbol period. The code generator 150 can generate the required spreading codes according to the above-mentioned OVSF tree in FIG. 2 . The output of the correlator 140 is coupled to a conventional de-interleaver 160, the interior of which is coupled to a conventional decoder 170, which outputs a data decision.

The proposed equalizer function, implemented by the channel impulse response estimator 120 and the adaptive chip estimator 130, strives to restore the integrity of the user waveform by estimating the transmitted multi-user chips at the receiver and thereby equalizing the channel. Orthogonality to suppress multiple access interference. From the perfectly estimated chip sequence, the desired user signal can be recovered without any residual interference from other users by correlating the multi-user chip sequence with the scrambling code and the user spreading code.

One problem with adaptive methods for CDMA in UMTS based systems is that no reliable training multi-user chip sequences are available. However, by using the knowledge of the correlation between the desired signal (multi-user chip sequence) and the received signal, such training sequences may not be needed. To this end, the receiver 10 employs a channel impulse response estimator 120 to estimate the channel impulse response.

According to a preferred embodiment, a hybrid equalizer architecture based on a variant of the chip-level LMS algorithm is proposed. The update rule for the equalizer weights is either Griffith's algorithm in the inactive period of the HSDPA channel, or a new decision-oriented scheme in the active period, which utilizes all available power by exploiting partial code knowledge of the HSDPA code. Thus, the update of filter taps or weights can be performed at a rate lower than the chip rate, resulting in important savings in complexity. The preferred embodiment provides significant gains compared to conventional Rake receiver based solutions.

In existing examples of HSDPA systems, two possible phases are provided during which a terminal device such as a mobile terminal or user equipment (UE) in third generation conditions can track the channel or Estimation and/or equalization. These two phases are referred to as the non-valid phase and the valid phase. The inactive stage or state is when the user monitors the channel, but no high-speed downlink shared channel HSDSCH code is allocated to him; on the other hand, the active stage or state is to allocate at least one HSDSCH code to the user. During the inactive and active phases, a similar adaptive equalizer architecture can be used, but with different update mechanisms and supported desired signals or statistical properties.

Fig. 5 shows a schematic timing diagram indicating the non-active phase I and the active phase A and the corresponding update mechanism for updating equalizer weights. The update mechanism is chosen to optimize the equalizer operation in terms of computational load, estimation speed, interference suppression, and the like. In the inactive phase I, the update mechanism should simply not relax the ability to track the time-varying channel. For this reason, the known Griffith algorithm G is generally used when a training sequence (desired signal) is not available or reliable. The Griffith algorithm G used during the inactive phase I approximates the input/output cross-correlation part of the Wiener filter by exploiting the variance of the base station total signal and the channel estimate. This provides the advantage that the filter taps at the beginning of the active phase A will be more reliable than simply considering the Rake receiver as a temporary initial solution at the transition instant. In order to reduce the computational load, it can be considered that the update rate of the Griffith algorithm G is lower than the update rate selected during the active phase A. However, it should be noted here that the symbol-level LMS mechanism using pilot tones can also be well used in the non-active phase I, where filter taps or weights are updated at most every 256 chips.

On the other hand, a new update mechanism is proposed based on a new variant of the Decision-Directed Least Mean Square Error (DD-LMS) equalizer during active phase A in which a user is allocated at least one HSDSCH.

Fig. 6 shows the proposed hybrid equalizer architecture according to the first preferred embodiment with a new variant of the DD-LMS equalizer.

The dashed lines and boxes in Figure 6 represent the parts of the equalizer architecture that operate during the non-active phase (ie, when Griffith's algorithm is used for weight updates). The first update function or unit 280 controls the weight determination function or unit 285 to apply filter weights to the first equalizer 215 in the filter branch of the equalizer architecture. The update rule of the first update unit 280 is given by the following equation (1), and implements the recursive filter update process according to the Griffith algorithm, where w _Rake represents the finite impact of a traditional Rake receiver with as many taps as the channel length Response (FIR) form. The FIR version of the Rake receiver is called a Channel Matched Filter (CMF). It corresponds to the conjugate symmetry of the channel, ie w _Rake [n] = h ^* [-n], where h[n] denotes a channel tap with delay n. σ _d ² represents the variance of the total base station signal, w _l represents the filter column vector of the current state, w _l+1 represents the filter vector of the next state, u _l represents the row vector of the input regression, and μ represents the step of the algorithm long. The update vector is normalized to the Griffith calculator part of the normalized minimum mean square error (NLMS) by entering the energy of the regression vector.

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; Rake</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Therefore, in the inactive stage, the input and output values of the first equalizer 215 are provided to the first update unit 280 for processing based on the above-mentioned Griffith algorithm to apply the adaptive weight update mechanism to the first equalizer 215 .

Continuous non-dashed lines and boxes are used during the active phase when the user is assigned at least one channel. The new update algorithm (considering a software implementation of a digital signal processor or a vector processor) or architecture (considering a hardware context) is arranged to route the received discrete-time samples y[l] to two branches, the filtering branch (upstream branch) and the adaptation branch (downstream branch). The signal samples are routed to the adaptation branch via a delay function or element 290 in which the signal samples are delayed by eg 16 chips (corresponding to one HSDSCH symbol period). The short symbol duration of 16 chips is small compared to relatively large typical correlation times in wireless channels. Thereby, the fact that a channel and its associated optimal equalizer weights do not change much within such a short period of 16 chips can be exploited.

The suggested switching or selection according to phase may be achieved by a switching or selection function or unit (not shown) responding to control information, such as a flag, indicating the dominant phase (active phase or inactive phase). In a hardware implementation, the conversion or selection unit may be an analog or digital electronic switch or selector. In software-based implementations, the switching or selection functions may be implemented by conventional branch or jump operations within software routines.

As shown in Figure 1, the upstream branch corresponds to the typical data flow path in any receiver with a chip-level filtering followed by descrambling and spreading structure. Initially, the filter weight h[l] is provided by the external filter setting mechanism 210 . A Fast Walsh Hadamard Transform (FWHT) is provided for efficiently implementing multiple despreading operations. If M codes are despread at spreading level N, using FWHT instead of M independent correlators reduces the complexity from M·N units to Nlog ₂ (N) units. Therefore, FWHT remains advantageous as long as M > log ₂ (N). The cross-M value for SF=16 is log ₂ (16)=4. Therefore, to jointly despread multiple HSDPA channels corresponding to the user of interest and other users, a FWHT of length 16 can be used, which inherently has 16 outputs. Hereinafter, such a FWHT is referred to as FWHT-16.

As can be seen from Figure 6, the signal samples output from the first equalizer 215 in the filtering branch are descrambled by a descrambling function or unit 220 and then provided to a despreading function or unit 230 applying FWHT-16. The FWHT-16 output associated with the respective HSDSCH of the user of interest passes through a decision block (slicer) and is fed forward to a decoding unit, such as decoder 170 in Fig. 1, or other bit-level processing unit.

However, all hard-detected HSDSCH symbols (e.g., from Quadrature Phase-Shift Keying (QPSK) or 16-Quadrature Amplitude Modulation (16-QAM) constellations) are also correlated with other hard-detected HSDSCH symbols provided their channelization codes are also known. The detected or hard-decided HSDSCH symbols are fed back to the downstream adaptation branch together. Since the code search space is limited to at most 14 codes (since at most one code is assigned to interested users, at most 15 codes can be assigned to HSDPA traffic), it is easy to detect even if it is unknown. Also, HSDSCH codes are placed consecutively (which makes detection an easier task), and the constellation is also limited to QPSK and 16-QAM, which also facilitates detection. Therefore, with little effort, other possible HSPDA codes can be exploited.

However, the symbol estimation fed back to the downstream branch is not limited to HSDPA codes. More suitably, all other FWHT-16 outputs are provided to a non-linear filtering function or unit 240, where all other FWHT-16 outputs are provided for non-linear filtering, e.g. by LMMSE scaling in the non-linear filtering module Block or enable all other FWHT-16 outputs, non-linear filtering controls the feedback to the adapted branch. Next, in the spreading unit 260, the estimated or detected symbols fed back to the adaptation branch are re-spreaded according to the FWHT-16 algorithm, and then re-scrambled in the scrambling unit 250 to obtain a symbol that can be used as a A training sequence or a desired signal for a desired signal.

The difference between the desired signal d[l] and the output of the second equalizer 255 supplied with delayed input signal samples u[l] is provided as error signal e[l] to a second updating unit 270, which Weight updating is performed for the second equalizer 255 . Then, the updated weights of the second equalizer 255 can be directly used as the weights of the first equalizer 215 of the upstream filtering branch.

Feedback of hard-detected HSDPA symbols and linearly estimated other valid symbols provides the advantage that the total chip-level signal synchronously transmitted by the base station can be estimated as precisely as possible and used as a second equalizer located in the downstream adaptation branch 255 training sequences. In particular, the second equalizer 255 operates on a delayed version of the received signal samples (delayed by one HSDPA symbol period). Thus, the detected or estimated signal (in the case of the linear feedback part) serves as the expected chip-level response for perfectly synchronized downstream adaptation branches in the following symbol period. Although the data flowing through the upstream filtering branch is one symbol period ahead of the data flowing through the adaptation branch, the equalizer weights adapted at the second equalizer 255 of the adaptation branch can be safely used for upstream filtering The actual data of the branch is filtered. This is possible because the symbol period of one HSDPA (ie, 16 chips) is almost negligible compared to the correlation time of a typical wireless channel.

FIG. 7 shows the non-linear filter operation indicative of a particular type of output samples applied to the FWHT-16 processing at the despreading unit 230 . In particular, as shown in the processing flow in the middle of FIG. 7 , the nonlinear filtering unit 240 classifies the sample outputs into HSDSCH branches for hard detection. The HSDSCH branch should carry the most reliable estimate of the desired signal. This is achievable as long as the decision is correct most of the time and the amplitude of the HSDSCH symbols is estimated accurately. The amplitude estimate may be based on the power offset value between the HSDSCH and the Control Pilot Channel CPICH signaled by the base station (ie NodeB).

Furthermore, the first output branch corresponds to the first output, i.e. all 1 codes for the partially despread CPICH pilot tones, PCCPCH codes and all under the OVSF subtree rooted at code c _16,0 of the OVSF tree of Fig. 2 Valid output of despreading of valid codes. There are two possible options here. According to a first option, the first output branch is excluded from the feedback operation and instead the sequence of CPICH chips is added, since this sequence is a known constant sequence. However, this approach has the following disadvantages: it also needs to estimate the magnitude of the CPICH, and once the CPICH is fed back explicitly, other codes under the OVSF subtree rooted at code c _16,0 cannot be included. According to a second option, as shown in the upstream flow of Fig. 7, the first output branch is fed back without any processing, or fed back with LMMSE weighted scaling.

For the rest of the other branch OBs, the treatment is chosen to be completely covert. It cannot be known in advance whether the OVSF subtrees rooted in each of them have apparent validity. The trick is that you don't need to know the effective generalization (spreading code) code, and don't need to estimate its actual symbols, as long as you don't take into account hard decisions or other nonlinear operations that explicitly require constellation and symbol magnitude information. It is also sufficient to obtain a pseudo-symbol estimate reflected from an actual symbol at a particular position in the OVSF hierarchy to its parent or child. Therefore, as shown in the lower processing flow chart of FIG. 7 , other branch OBs can be processed by the following operations: first estimate the energy threshold σ _th ² , and then compare the energy of other branch OBs with the threshold. Branches above this energy threshold σ _th ² pass through the nonlinear filtering unit 240 and are fed back, while other branches are blocked.

It should be emphasized here that it is not very useful to correctly detect all validity under either branch. Rather it is sufficient to determine whether it is beneficial to include or exclude any particular branch. For example, there may be weak effectiveness under some branches, but the interference and noise captured by this FWHT-16 branch are more dominant. In this case it is better to block that branch.

As an optional improvement mechanism, the LMMSE weighting mechanism can also be introduced into the nonlinear filtering unit 240 or as a separate unit, which is used for hard detection and based on reliability measures such as SINR values of branches. The branch works after nonlinear processing.

The feedback strategy improves the energy of the desired signal, allowing better channel tracking. In addition, recursive processing can be understood as also a learning process for the desired signal. At each recursive step, the quality of the filter weights is improved, thereby improving the quality of the detected or estimated feedback signal (ie the desired signal).

Updating the filter taps or weights at the second update unit 270 can be done at the chip rate or even lower than the chip rate, thus reducing the complexity. The update rule implemented in the second update unit 270 may be a recursive equation for updating filter taps under the DD-LMS algorithm proposed by Schniter et al. in the prior art as mentioned at the beginning, where μ is the algorithm the step size.

Note that despreading unit 230 performs despreading jointly at a single level, ie SF=16. Therefore, in the example of FWHT-16 operation, the despreading complexity can be reduced to 1/4. Computational complexity is even lower compared to other methods. The number of codes is significantly reduced due to multiple effective codes at higher spreading factors of OVSF and a single dummy code at SF=16 spreading factor. Furthermore, a mix of hard and weighted linear decisions and an optional explicit reliability measure make the proposed scheme more efficient.

In this solution, the initialization through the external weight setting mechanism 210 can be based on the traditional Rake principle. Thereafter, in the non-active phase, the Griffith algorithm or another suitable algorithm may be used.

Any DD scheme tends to have the problem of non-convergence. This behavior occurs when the equalizer is locked into a rotating constellation (state) and cannot recover from it. In order to avoid non-convergence, the PCPICH signal can be utilized, which is a 45-degree vector at both the chip level and the symbol level after descrambling. First, Super-PCPICH-Symbol (the sum of one PCPICH symbol) is obtained every 5 or 10 PCPICH symbol periods. The PCPICH symbol period is a design parameter that depends on Doppler spread and noise, and can be less or more. The equalizer filter The weights are rotated by an angle equal to the difference between the estimated phase of the Super-PCPICH-Symbol and 45 degrees, which is the correct phase of the pilot signal.

The complexity of adaptive LMS filtering is based on the components of adaptation and filtering. Both have almost the same complexity, about 4 times the number of taps or weights for real multiplication and addition. The DD algorithm for the second update unit 270 requires two filtering and one adaptation mechanism. Therefore, when adapting at chip rate, the DD update scheme is 50% more complex than adaptive LMS filtering. However, when the adaptation rate is reduced to 1/χ, the complexity of the adaptation part of the LMS algorithm and the complexity of one filtering part and the adaptation part of the DD-LMS algorithm are both reduced proportionally. This can be expressed by the following equation (2):

$\frac{\mathrm{<span\; class="notranslate">\; DD</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; LMS</span>}}_{\mathrm{<span\; class="notranslate">\; complexity</span>}}}{{\mathrm{<span\; class="notranslate">\; LMS</span>}}_{\mathrm{<span\; class="notranslate">\; complexity</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Thus, for example, the increase in complexity to the LMS algorithm can be reduced to only 6% when the filter is adapted at the HSDSCH symbol rate. The additional re-spreading unit 260 and re-scrambling unit 250 add a negligible amount of complexity to the filtering operation.

The proposed equalization scheme reduces the impact of interfering power, and since the nonlinear filtering unit 240 is used to utilize all available power, the power of the desired feedback signal d[l] is much better than conventional solutions. Significant gains are achieved compared to conventional NLMS algorithms where adaptation of filter taps or weights using only pilot tones is not possible due to the supplied power being too small compared to the interference level conduct. Since the complexity depends on the adaptation rate, the required increase in complexity can be reduced or tuned.

Fig. 8 shows a schematic block diagram of an alternative hybrid equalizer architecture according to a second preferred embodiment. Its principle is similar to that of the first preferred embodiment. Therefore, only those modules that are new or whose functions and operations are changed in the second preferred embodiment are described below. The functions and operations of the remaining modules are similar to those of the first preferred embodiment and will not be described here. Furthermore, the features of the first preferred embodiment described in conjunction with FIG. 7 are also applicable to the following second preferred embodiment.

The following architectural changes are introduced in the DD-LMS of the second preferred embodiment, reducing the adaptation complexity by 50% and the overall complexity by 33%.

In the first preferred embodiment shown in FIG. 6, the input signal y[l] is delayed by 16 chips, ie one HSDPA symbol, in the delay element 290 and fed forward to the downstream adaptation branch as signal u[l]. The upstream data filtering branch estimates user symbols. The upstream data filtering branch consists of filtering in the first equalizer 215, descrambling in the descrambling unit 220, despreading all dummy codes in the despreading unit 230 via FWHT with spreading factor level 16, and Finally, in the nonlinear filtering unit 240, a hard decision is made on the HSDPA code and LMMSE weighting is performed on other codes. The estimated symbols are fed back to the downstream branch through the FWHT of the re-spreading unit 260 and the re-scrambling of the re-scrambling unit 250 . The resulting BS chip estimate d[l] is used to adapt the expected response of the second equalizer 255. The resulting delay is 16 chips plus some processing delay by units 260 and 250 (which can be done at high processor speed but not necessarily at chip rate), which is much less than typical correlation times for mobile channels. Thus, the estimated filter weights of the second equalizer 255 can be directly used in the first equalizer 215 of the data filtering branch.

The main complexity of the overall architecture of the first preferred embodiment of FIG. 6 comes from the two filter functions of the first and second equalizer 215 , 255 . In the second equalizer 255, the chip rate filtering and adaptation process is done. In the first equalizer 215, only filtering is done since the second equalizer 255 has provided weights. Both operations have O(N) complexity, where N is the number of FIR filter taps. Therefore, two modules 215 and 255 result in a total computational complexity of 3 units.

In the second preferred embodiment of Fig. 8, in the upstream branch, the output signal of the first equalizer 215 is also delayed by the second delay element 290 by the same amount of delay (ie, one symbol period). subtracting or comparing the delayed output signal of the first equalizer 215 from the re-scrambled feedback signal from the re-scrambling unit 250, and The resulting error signal e[l] is provided to the second update unit 270 which is now directly fed to the first equalizer 215 . Furthermore, the delayed output signal u[l] originating from the first delay element 290 is supplied to the first equalizer 215 . Thereby, the first equalizer 215 of the upstream branch can be adapted without the need for the second equalizer 255 of the downstream branch. Therefore, the adaptation process is moved from the second equalizer 255 to the first equalizer 215, and the second equalizer 255 can be deleted. Figure 8 shows the resulting architectural changes. The rest of the architecture of the second equalizer 255 (ie, symbol estimation processing, etc.) is the same as in the first preferred embodiment.

Due to the architectural changes of the second preferred embodiment, the adaptation overhead can be reduced by 50% (from two units to one unit), and the overall complexity can be reduced by 33% (from three units to two units). Furthermore, this also achieves proportional power savings.

It should be appreciated that the functions or modules of Figures 6 and 8 may be implemented using discrete circuit elements or software routines executed by a suitable data processor. A combination of circuit elements and software routines may also be employed. Other weight update algorithms that provide adaptive complex chip estimation filters can also be used.

It should be noted that the two aspects of the invention (ie choosing different update architectures or algorithms according to the stage on the one hand and employing new variants of the DD-LMS equalizer on the other) can be implemented in separate embodiments. That is, the dashed blocks in Figures 6 and 8 can be considered optional, and the remaining new variants of the DD-LMS equalizer architecture or algorithm corresponding to the non-dashed blocks in Figures 6 and 8 can be Provided without transition or selection according to stage. Furthermore, a hybrid equalizer architecture with stage-based transition and selection between different equalizer architectures or update algorithms can be implemented without new variants of the DD-LMS equalizer. Alternatively, the switch or selection can be made between two known equalizer architectures or update algorithms, such as the Griffith algorithm and the traditional DD-LMS algorithm disclosed in the aforementioned prior art by Schniter et al.

In summary, a receiver apparatus and method for controlling weight adaptation in a receiver of a code division multiplexing communication system using orthogonal spreading codes is described, wherein the received Chip-level filtering of discrete-time signal samples. In addition, the received discrete-time signal samples are delayed by a time period corresponding to the data symbols and used in a second equalization step. The symbol estimates obtained from the first equalization step are non-linearly filtered and used in subsequent symbol periods as the expected response for the second equalization step, where the equalizer weights adapted in the second equalization step are used by in the first equalization step. Alternatively, the second equalization step can be omitted, and the weight adaptation can be incorporated into a single equalization step. As an additional or alternative option, it is possible to provide a hybrid equalizer architecture in which the above two-step equalization is used during the active phase for allocated channels and another weight update scheme is used for unassigned channels. During the inactive phase of the channel. As a result, the detrimental effect of interference power can be reduced with a small increase in complexity.

Last but not least, it should be noted that the term "comprising" when used in a specification including claims is intended to specify the presence of stated features, means, steps or components, but not to exclude one or more other features. , the presence or addition of means, steps, components or combinations thereof. Furthermore, in the claims, the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Furthermore, any reference signs do not limit the scope of the claims.

Claims (22)

1. receiver device that is used to utilize the code division multiplex communication system of orthogonal intersection, described receiver device (10) comprising:

A. filtering branch has first equalizing device (215) that is used to receive discrete time signal sample and is used for described received signal sample is carried out chip-level filtering;

B. feedback device (240) is used for the sign estimation that obtains at described filtering branch is carried out nonlinear filtering, and filtered sign estimation is fed back to first equalizing device (215);

C. adaptive device (255 or 215) is used for receiving described received signal sample via deferred mount (290), and wherein said deferred mount (290) was arranged to described received signal sample delay and corresponding time cycle of data symbol; And

D. wherein, in symbol period subsequently, described adaptive device carries out adaptive Expected Response with described filtered sign estimation with the equaliser weights of described first equalizing device (215) of opposing.

2. receiver device as claimed in claim 1, wherein, described adaptive device comprises second equalizing device (255), and will locate adaptive equaliser weights at described second equalizing device (255) and be used for first equalizing device (215).

3. receiver device as claimed in claim 1, wherein, described first equalizing device (215) comprises adaptive device.

4. receiver device as claimed in claim 2, also comprise substracting unit, be used for obtaining from the equalizing signal sample of described second equalizing device (255) output and from the difference between the feedback signal sample of described feedback device (240) acquisition, and described difference offered updating device (270), described updating device (270) is used for carrying out adaptive to the described equaliser weights of described second equalizing device (255).

5. receiver device as claimed in claim 3, also comprise substracting unit, be used to obtain described equalizing device (215) output and through another be arranged to after described filtering signal sample delay and the deferred mount of corresponding time cycle of data symbol (290) postponed the filtering signal sample and from the difference between the feedback signal sample of described feedback device (240) acquisition, and described difference offered updating device (270), described updating device (270) is used for carrying out adaptive to the described equaliser weights of described equalizing device (215).

As before the described equipment of arbitrary claim, wherein, the described time cycle is corresponding with the mark-hold length of described orthogonal intersection.

As before the described equipment of arbitrary claim, wherein, described filtering branch comprise be used for single code tree grade to descrambling and balanced after the sample of signal despreading device (230) that carries out despreading, wherein said de-spread signal samples is provided for described feedback device (240).

8. equipment as claimed in claim 7, wherein, the described filtering sign estimation of in spread spectrum device (260) described feedback device (240) being fed back is carried out spread spectrum once more.

As before the described equipment of arbitrary claim, also comprise choice device, be used for during not having channel code to distribute to non-effective stage of described receiver device selecting other right value update devices (280,285).

10. equipment as claimed in claim 9, wherein, described other right value update devices (280,285) are arranged to directly relatively operating based on the input and output of described first equalizing device (215).

11. as each described equipment in claim 9 or 10, wherein, described other right value update devices (280,285) are arranged to based on the Griffith algorithm to be operated.

12. as before the described equipment of arbitrary claim, wherein, described feedback device (240) is arranged to the described sign estimation that will obtain and is categorized as in described filtering branch: should carry first branch group of the downlink sharied signal channel of reliable estimation that will be detected firmly, first output got rid of from feedback and replaced or do not carry out any processing and second branch group that is fed by known constant sequence, and the 3rd branch group that is lower than the residue branch that will get clogged of predetermined threshold.

13. equipment as claimed in claim 12, wherein, described predetermined threshold is corresponding with the hard detection and the averaged energy levels between the soft detected value of all downlink sharied signal channels.

14. as before the described equipment of arbitrary claim, wherein, use weight mechanism after described feedback device (240) is arranged in described nonlinear filtering, described weight mechanism is based on reliability measurement.

15. equipment as claimed in claim 14, wherein, described reliability measurement is the measurement to Signal to Interference plus Noise Ratio.

16. the method for the weight adaptation of a receiver (10) that is used for controlling the code division multiplex communication system that utilizes orthogonal intersection said method comprising the steps of:

A. use first equalization step reception discrete time signal sample and described received signal sample is carried out chip-level filtering;

B. the discrete time signal sample with described reception postpones and the corresponding time cycle of data symbol, and the time signal sample of described delay is used for the adaption function of described first equalization step; And

C. apply the nonlinear filtering operation to the sign estimation that obtains from described first equalization step, and in symbol period subsequently, with filtered sign estimation as at Expected Response described first equalization step, that equaliser weights carried out adaptive described adaption function.

17. method as claimed in claim 16, wherein, described adaption function comprises second equalization step.

18. a computer program comprises the code device that produces when being suitable for moving on computers as the step of claim 16 or 17 described methods.

19. a receiver device that is used to utilize the code division multiplex communication system of orthogonal intersection, described receiver device comprises:

A. filtering branch is used to receive discrete time signal sample, and is used for described received signal sample is carried out chip-level filtering;

B. equalizing device (215) is used for recovering being included in the orthogonality between the spreading code of described received signal sample;

C. the weight adaptation device (255; 285), be used to upgrade the filter weights of described equalizing device; And

D. choice device is used to control described weight adaptation device (255; 285) during effective stage of distributing at least one channel code for described receiving equipment, first update scheme is used for weight adaptation, and during not having channel code to distribute to non-effective stage of described receiver device, second update scheme is used for weight adaptation.

20. method as claimed in claim 19, wherein, described first update scheme is based on the normalization minimum mean-square error scheme of decision-directed, and described second update scheme is based on the Griffith algorithm.

21. the method for the weight adaptation of a receiver that is used for controlling the code division multiplex communication system that utilizes orthogonal intersection, described method comprises:

A. receive discrete time signal sample and described received signal sample is carried out chip-level filtering;

B. described received signal sample is carried out equilibrium, to recover to be included in the orthogonality between the spreading code in the described received signal sample; And

C. control is applied to the renewal of the filter weights of described equalization step, so that during effective stage of distributing at least one channel code for described receiving equipment, use the first right value update scheme, and during not having channel code to distribute to non-effective stage of described receiver device, use the second right value update scheme.

22. a computer program comprises the code device that is suitable for producing when moving on computers the step of method as claimed in claim 21.

Images