Practical space-time radio method for cdma communication capacity enhancement

The First Practical Space-Time Radio Method for Improving the Capacity of Code Division Multiple Access Communication Background Field of the Invention The present invention relates to a time wireless communication system. More specifically, the present invention relates to methods for improving wireless communication performance by using the spatial domain and systems for implementing these methods.

As the demand for wireless communication increases, it is necessary to develop a technology that effectively uses the allocated frequency band, that is, increases the capacity to perform information communication within the limited available bandwidth. In a conventional low-capacity wireless communication system, information is sent from a base station to a user by broadcasting an omnidirectional signal on one of a plurality of predetermined frequency channels. Similarly, by broadcasting a similar signal on one of the frequency channels, the user sends information back to the base station. In this kind of system, multiple users access the system individually by dividing the frequency band into different sub-band frequency channels. This technique is known as Frequency Division Multiple Access (FDMA). The standard technique used in commercial wireless telephone systems to increase capacity is to divide the service area into space cells. Instead of using only one base station to provide services to all users in the area, a group of base stations are used to provide services to separate space cells individually. In this kind of cellular system, if multiple users access the system from different space cells, these users can reuse the same frequency channel without interfering with each other. Therefore, the cellular concept is a simple type of space division multiple access (SDMA)» In the case of digital communications, other technologies can also be used to increase capacity. Well-known examples are Time Division Multiple Access (TDMA) and Code Division Multiple Access (CDMA). TDMA enables several users to share a single frequency channel by assigning their data to different time slots. CDMA is generally a spread spectrum technology, This technology does not limit a single signal to a narrow frequency channel, but expands the signal across the entire frequency band. By assigning a different orthogonal digital code sequence or spread signal to each signal, the signals sharing the frequency band can be distinguished. Theoretical analysis shows that in this industry, CDMA has been recognized as the most promising method in various air interfaces (for example, see Andrew J Viterbi, CDMA Principles of Spread Spectrum Communication)^u, And Vijay K.

98811171.3 p.

Garg waits, "Applications of CDMA in Wireless/Personal Communication").

Although the prospects of CDMA are better, practical factors such as power grip speed and inter-base station interference greatly limit the effectiveness of the system in the initial stage of CDMA implementation. The capacity of a CDMA-based system mainly depends on the ability to control power extremely accurately, but in a mobile environment, the signal fluctuates so fast that the system cannot be controlled. Unfortunately, the usual characteristics of the mobile wireless environment are unstable signal propagation, severe signal attenuation between communicating entities, and co-channel interference caused by other radio resources. In addition, many urban environments contain a large number of reflectors (such as buildings), which make the signal follow multiple paths when it is sent from the transmitter to the receiver. Since the various parts of the multipath signal arrive at different phases that produce destructive interference, the multipath can cause unpredictable signal fading. In addition, the rapid fading caused by the combination of the multipath components of the signals reflected in random phases from the various units (scattere) around the moving transmitter (scattering area) is regarded as a major problem in wireless communication. The destructive combination at the antenna produces a time-varying signal level that is a function of the power density of the Rayleigh distribution. Therefore, the received power appears to be "deep" or zero at different times, which is reflected in the transmitted information In addition to fading, when the radiated power is increased in order to provide services to the blind area, the interference between base stations will also cause the system to cause obvious errors. System performance deteriorated significantly.

Modern communication systems reduce the fading effect by interleaving the transmitted data, de-interleaving the received data, and adding appropriate error correction techniques. In addition, the use of space diversity is an extremely common method for mitigating fading, for example, the correlation of received power as a function of time (power/time) in a signal received on two sufficiently separated antennas (10 wavelengths or more) Less sex. Therefore, most point-to-multipoint communication systems use space diversity combination to reduce the fading effect. In most cases, the receiver either selects an antenna with a stronger signal power ("switching diversity"), or combines the output of the two antennas after compensating for the phase and amplitude difference ("maximum ratio combination").

Spread-spectrum direct sequence systems (such as IS-95) use time diversity to provide additional fading mitigation, that is, multipath components can be separated in time due to the signal bandwidth and the autocorrelation function associated with it. If the multipath components arrive at sufficient time intervals, their power/time functions will be irrelevant. In IS-95, the RAKE receiver provides multiple demodulators ("finger^^,), each of which is assigned a different signal arrival time. Generally, the number of demodulation channels on the base station is 4. If the multipath component of the arriving signal has a significant delay spread (such as a few microseconds), the system can successfully assign different "fingers" to the incoming multipath component and achieve excellent fading mitigation. However, in most cases

98811171.3 Second, the delay spread is not enough to provide time diversity (especially in the suburbs), and the fading mitigation is still mainly provided by space diversity and coding. Because each sector of the current base station only uses two antennas, generally only two "fingers" are activated.

Recently, a high degree of attention has been paid to how to further use the spatial domain to improve the performance of wireless systems: it is well known that, in principle, SDMA technology can significantly improve the performance of CDMA-based networks. The precision and complexity of these technologies vary. The methods currently proposed are either simple but not very effective, or they are too complicated for actual implementation.

A well-known SDMA technology is to provide a set of individually controlled directional antennas for the base station so as to divide the cell into separate sectors, and each sector is controlled by a separate antenna. As a result, frequency reuse within the system can be increased and/or co-channel interference can be reduced. In addition to individually controlled directional antennas, this technology can also be implemented with coherently controlled antenna arrays. By using a signal processor to control the relative phase of the signal applied to the antenna unit, a predetermined beam can be formed in the direction of each separated sector. Similar signal processing can be used to selectively receive signals only from different sectors. However, these simple sectorization techniques only increase the capacity relatively little.

US Patent No. 5,563,610 discloses a method for mitigating signal fading caused by multipath in a CDMA system. By introducing a certain delay into the received signal, the Rake receiver can more clearly distinguish uncorrelated fading signal components. Although this diversity method can reduce the fading effect, it does not use the spatial domain and does not directly increase the system capacity. In addition, this method uses a fixed beam structure to combine angle and time diversity. Because the beam output levels are either too different or similar but highly correlated, this method is not very effective. If two signal parts arrive from similar directions, they pass through one beam and therefore cannot be distinguished. On the other hand, if these signal parts arrive between beams, the levels are similar, but at the same time they are highly correlated.

A more sophisticated SDMA technique has been proposed to significantly increase the system capacity. For example, U.S. Patent No. 5,471,647 and U.S. Patent No. 5,634,199 (both to Gerlach, etc.) and U.S. Patent No. 5,592,490 (to Barratt, etc.) disclose wireless communication systems that utilize the spatial domain to improve performance. In the downlink, the base station determines the spatial channel of each user and uses the channel information to adaptively control its antenna array to form a customized narrow beam. These beams send an information signal via multiple paths so that the signal reaches the user with maximum strength. These beams can also be selected not to point to other users, thereby reducing co-channel interference. In the uplink, the base station uses the channel information to spatially filter the received signal, thereby receiving the uplink signal with maximum sensitivity, and distinguishing the uplink signal from signals sent by other users. Selective power transmission through smart directional beams

98811171.3 The first transmission can reduce the interference between base stations and the carrier-to-interference (C/Ι) ratio at the base station receiver.

The biggest problem in adaptive beamforming is how to quickly estimate the wireless air channel for effective beam allocation. In the uplink, several signal processing techniques are known for estimating the spatial channel based on the signal received on the antenna array of the base station. These techniques usually involve inversion or singular value decomposition of the signal covariance matrix. However, the computational complexity of this calculation is so high that it cannot be put into practical use at present. These highly complex methods use array signal processing theory. This method estimates the uplink channel (such as the angle of arrival and time of the multipath signal part) to create a space-time matched filter to achieve maximum signal transmission. The proposed method involves the calculation of the signal covariance matrix and the extraction of its eigenvalue vector to determine the coefficients of the array. The main problem of array signal processing is explained by the following formula:

X=AS+N where X is a matrix of antenna array signal snapshots (each column has a snapshot of all antenna elements), S is the transmitted signal matrix (each column has a snapshot of information signals), and A is Antenna array and channel response or array cluster (manifold) matrix, N is the noise matrix. The main problem of array signal processing is to estimate S based on the statistics of A and S, that is, to reliably and correctly estimate all incoming signals with interference and thermal noise N. Over the past few years, extensive research has been conducted on this issue. There are two well-known estimation algorithms, maximum likelihood sequence estimation (MLSE) and minimum mean square error (MMSE). Using these techniques, if S represents a signal with known characteristics such as a constant module (CM) or a finite alphabet (FA), the processing can be performed by using the time-domain structure statistics of the known signal. If the array clusters are known, convergence can occur faster. However, the amount of calculation for this process is very large. In base stations that need to support more than 100 mobile units at the same time, the current computing capabilities cannot make them practical.

Most of the adaptive beamforming methods described in the art (for example, U.S. Patent No. 5,434,578) deal with uplink estimation in a large amount and also require a large amount of computing resources. However, there are few methods to deal with downlink estimation, and downlink estimation is a more difficult problem. Since the spatial channel is related to frequency, and the uplink and downlink frequencies are usually different, the uplink beamforming technology does not provide enough information for the base station to extract the downlink spatial channel information and improve the system capacity. One method for obtaining downlink channel information uses user feedback. However, the required feedback rate makes this method impractical.

Therefore, there is a need to use beamforming methods to increase the capacity of the wireless system and overcome the above-mentioned limitations in the known methods.

98811171.3 The present invention provides a wireless communication method that uses the spatial domain in both the uplink and the downlink without complicated calculation processing. This method significantly increases the capacity in both the uplink and the downlink while maintaining the simplicity of implementation. Its purpose is to eliminate the necessity of covariance matrix processing, adopt low-bit counting algorithm, and adopt signal multipath structure to achieve.

The wireless communication method of the present invention includes transmitting a code modulated signal, such as a CDMA signal, from a mobile unit, which is obtained by modulating an original symbol using a predetermined pseudo-noise sequence. The original symbol represents an original information signal. The base station antenna array then receives N complex-valued signals in parallel from the N corresponding antenna elements. Then, each of the N signal sequences is correlated with the pseudo-noise sequence to despread and select N received signals containing N received symbols corresponding to the same original symbol. Then, the N received symbols are transformed in parallel to obtain N complex-valued transformer outputs, which are then collectively correlated with a set of complex array calibration vectors to obtain spatial information about the signal. Each array calibration vector represents the response of the antenna array to the calibration signal initiated in a predetermined direction relative to the base station. Repeat the above steps to obtain spatial information (Angle of Arrival (AOA), Time of Arrival (TOA), and distance from the mobile unit) about multiple signal components corresponding to the same mobile unit. Then, the spatial information is used to spatially filter the subsequent complex-valued signal sequence. Then, the filtered signal is demodulated to obtain symbols from the original information signal.

The original symbols are selected from a finite alphabet of symbols. In a preferred embodiment, the finite alphabet contains no more than 64 symbols, and the calibration vector includes a complex value component with 1 or 2 bits plus sign real part and 1 or 2 bits plus sign imaginary part. If necessary, the number of bits can be increased. This simple representation allows the correlation value to be calculated via addition without the need for complex multiplication calculations. In one embodiment, the correlation step obtains spatial information about multiple signal components from a mobile unit with a small time interval (ie, a time range less than one chip). Another embodiment of the present invention includes the step of tracking time and angle information of a plurality of signal components. In one embodiment, an analog signal with a known amplitude and zero phase (ie, a "test tone" signal) is inserted into the base station's transmit and receive channels. Then, the signal on the output of each channel is decoded to determine its phase and amplitude. Then, the measured phase and amplitude data are used to correct the antenna calibration data (array cluster matrix), thereby eliminating the phase and amplitude mismatch in the multi-channel receiving and transmitting system. The mismatch is partly caused by temperature changes, component deterioration, Receive and transmit power, etc.

The present invention also provides a spatially filtered downlink information signal based on spatial information about a plurality of signal components that have been determined by the uplink. The spatial filtering includes the spatial information according to the relevant mobile unit.

98811171.3 The first message is used to assign the mobile station to a beam. The spatial information includes directionality and distance information about the mobile unit. The downlink beam is a dynamically adaptive overlapping group of wide and narrow beams, thereby assigning a wide beam to a mobile unit that is close, and a narrow beam to a mobile unit that is far away. The downlink beamwidth is determined based on the AOA distribution of the uplink signal (collecting many symbols) and the distance when possible. Since the normal AOA extension is related to the distance, the farther the mobile unit is, the smaller the AOA extension will be. Therefore, the AOA extension can be adopted as described above. The beam group is changed according to the statistics of the spatial information of all mobile units served by the base station in order to optimize the system performance. In one embodiment, multiple (2 to 4) narrow beams (2 to 3 degrees) are formed in the wide aperture antenna array to cover the scattering area to minimize the effect of fast (Rayleigh) fading. The wide aperture array allows multiple narrow beams in the same general direction to be constructed with low-correlation weight vectors, providing low correlation (about 0.7 or less) between beam outputs in the wide antenna array. By estimating the angle range of the incoming signal, especially by determining the angle range or peak value or variation of the angle of arrival histogram, the width and orientation of the beam are determined. In a preferred embodiment, the transmission of the downlink beam is performed based on beamforming information, which includes complex valued elements having 3 bits plus a sign real part and 3 bits plus a sign imaginary part. In CDMA In the preferred embodiment of IS-95, downlink traffic beams are assigned to specific mobile units, while overhead beams are maintained for 3 or 6 sector base stations. Keep a small phase difference between the service and the pilot beam to prevent the deterioration of the demodulation performance of the mobile station.

In some embodiments, for a wideband CDMA communication system, the pilot signal is code multiplexed into a signal sent from the mobile unit to the antenna array of the base station. The base station correlates the pilot signal of the incoming signal with a series of delayed pilot signals generated from the base station. These correlation values are spatially correlated with the antenna array cluster matrix to generate signal angle of arrival (AOA) and time of arrival (TOA) histograms. The obtained histogram is used to determine the "best" AOA and TOA for forming the uplink and downlink beams directed to the desired scattering area. In the spatial correlation calculation, a pilot signal is used instead of the actual The signal itself leads to the simplification of the communication system. In many cases, the AOA histogram can be assembled based on the spatial correlation between the array response vector (including the electrical amplitude and phase of all antenna array elements) and the array cluster matrix. Since the angular range between the various signals is relatively small at long distances, knowledge of array clusters is beneficial.

The present invention also provides a CDMA base station adopting the above method. The base station includes an antenna array with N antenna elements and a set of N receivers connected to the N antenna elements to generate N incoming signals. The base station also includes a set of N despreaders connected to the N receivers for generating N despread signals corresponding to a single mobile unit based on the N incoming signals. A set of N symbol converters

98811171.3 is connected to N despreaders, and generates a complex value output according to these despread signals. A spatial correlator connected to the N symbol converters correlates the complex value output with the stored array calibration data to generate beamforming information for a plurality of signal parts associated with the mobile unit. In this preferred embodiment, the array calibration data is composed of complex-valued array response elements represented as bit plus sign imaginary part and bit plus sign real part. The receiving beamformer connected to the spatial correlator and N receivers then spatially filters the N incoming signals according to the beamforming information. The Rake receiver (or other equivalent receiver) connected to the receiving beamformer generates an information signal based on the spatially filtered signal. In one embodiment, the base station also includes a tracker, which is connected to the spatial correlator and the receive beamformer. The tracker tracks multiple signal parts and optimizes the performance of the receive beamformer.

In this preferred embodiment, the base station further includes a transmit beamformer connected to the spatial correlator. The transmitting beamformer generates a spatial beam according to the beamforming information, so as to improve the system capacity. The spatial beam is a dynamically calculated downlink beam group, which includes a narrow beam and an overlapping wide beam, so that the narrow beam and the overlapping wide beam are phase-matched. These spatial beams are selected so that a narrow beam is assigned to a far mobile station and a wide beam is assigned to a close mobile station.

In one embodiment, the base station includes a compensation signal source and a compensation detector, both of which are connected between the transmitting and receiving beamformer group and the N transmitter and receiver groups. The compensation signal source inserts an analog "test tone" signal of known amplitude and zero phase into the transmission channel, and the compensation detector decodes the test tone signal and accumulates the measured phase and amplitude data, which are used for correction Phase and amplitude mismatch data.

Detailed Description Although the following detailed description contains many specific examples for explanatory purposes, those of ordinary skill in the art should understand that various changes and modifications can be made to the following description within the scope of the present invention. Therefore, the following preferred embodiments of the present invention do not impair the generality of the proposed invention, and do not limit the proposed invention in any way.

Fig. 1 shows a general schematic diagram of the system structure of the base station of the present invention. The base station includes a receiving antenna array with N antenna elements 10 = In this embodiment, the system also includes a separate antenna array 15 for transmission. However, as is well known in the art, an antenna duplexer can be used To combine these arrays. This embodiment allows low-cost duplexers and antenna filters, because each unit requires less power to provide the required effective radiated power (ERP) due to beam forming. Preferably, the number N of antenna elements is about 16.

98811171.3 p.

Each of the N antenna elements is connected to a corresponding one of a set of N conventional receivers 101. Each receiver down-converts the incoming signal and digitizes the signal to produce a received signal with I and Q (in-phase and quadrature) signal components. In this embodiment, a common local oscillator 104 is used to coherently tune each receiver to allow measurement of both phase and amplitude data, thereby generating an N-dimensional received signal vector with complex value components at any given instant. Alternatively, a fixed-frequency calibration signal can be inserted into all receiver channels together with the received signal, thereby continuously estimating the phase and amplitude difference between the receivers. The calibration signal may be different from the received signal because it is not expanded, and may have an extremely low level because of its extremely long integration period. The specific related receiver design is disclosed in US Patent No. 5,309,474.

The received signal vectors from the N receivers 101 are fed to a set of L channel estimators 11, and are also provided to a corresponding set of L receiver banks 14. Each channel estimator 11 and corresponding receiver bank 14 is used to estimate the channel and receive signals from a single mobile unit. Therefore, the maximum number of mobile units that can be served by the base station at the same time is L. In a preferred embodiment, L is at least 100. The respective estimators 11 are the same as each other in structure and operation principle. Similarly, the receiving row 14 is the same. Therefore, the following description is limited to a single estimator 11 and its corresponding receiver bank 14 for estimating the channel of a single mobile unit and receiving its signal. In this preferred embodiment, the channel estimator 11 includes a set of N A despreader 102, a corresponding set of N fast Hadamard transformers (FHT), and a spatial correlator 105. The despreader 102 is disclosed in detail in U.S. Patent No. 5,309,474 traditional code correlator. Each of the N despreaders correlates a component of the received signal vector with a pseudo-noise (PN) code sequence assigned to the associated mobile unit according to the IS-95 standard. Each code correlator or despreader 102 uses a varying time offset (synchronized with other code correlators in the same row) to separate the multipath portion that arrives with at least one PN chip period difference. The time offset is determined using repeatability assumptions, for example, the code time offset is set, and the symbol length of the sample is collected, and then the processing described here is performed. The result obtained is a CIR buffer (described later), where the peak value represents the TOA of the difference signal path. The following description discusses the processing of a multipath part. Do the same for all multipath parts that are strong enough to be isolated.

Each despreader 102 outputs a despread signal corresponding to a mobile unit received on a Yao line. The despread signal is fed to a fast Hadar code converter (FHT) 103 = The FHT used in the present invention is basically the same as that disclosed in the conventional FHT (for example, in U.S. Patent No. 5,309,474), but the difference lies in the FHT of the present invention Keep the input complex phase information. In other words, although the standard FHT output is transferred

98811171.3 is replaced by amplitude, but the FHT used in the present invention outputs complex numbers, thus retaining the two data of phase and amplitude. Each FHT in this embodiment has 64 complex outputs, and its amplitude represents the degree of correlation between the despread signal and each of the 64 symbols in the predetermined symbol alphabet. In the preferred embodiment, the symbol alphabet is a set of 64 non-crossing Walsh symbols.

For a given symbol received on the antenna array 10 (in IS-95, the symbol period is about 208 microseconds), the signals received on the N antenna elements are separated and passed through N corresponding receptions in parallel The device 101, the despreader 102 and the FHT 103 also maintain the relative phase information between the signals. The collection of N FHTs 103 together generates a complex element signal matrix B of N×64. Each column of B is an N-dimensional vector, called a spatial response vector, and its N components represent the correlation between a Walsh symbol and the signal received on N antenna elements. The matrix B is fed to the spatial correlator 105 column by column at the timing synchronized with the Walsh symbol.

As will be described in more detail later with reference to FIG. 4, the spatial correlator 105 correlates the signal matrix B with the array calibration matrix A. Matrix A is additionally obtained by calibrating the phase, amplitude and angle of the antenna array. The correlation operation generates a correlation matrix C, which represents the correlation between the signal received on the antenna array and a set of predetermined directions and a set of predetermined symbols. According to the analysis of the matrix C, the correlator 105 generates a signal angle of arrival (AOA) and a scalar value (AOA quality), which is proportional to the "purity" of the wavefront and the signal level. The data is transmitted to the controller 106, and the controller 106 uses the data to determine the best uplink beam factor for that particular signal part. Generally, as known in the art, this entire process is performed on the 4 strongest multipath parts by setting the code and the "sampling period start" time for the expected TOA. In addition, time-of-arrival (TOA) and AOA deterministic data are generated, so that a spatially matched filter containing beamforming information for each signal part can be generated. Collect the AOA results by repeating the above processing for multiple incoming information symbols. This data is used to generate an AOA histogram through which the most desired AOA and AOA distribution are calculated for each individual signal part. The AOA provides beam direction information, and the AOA distribution provides beam width information. The function of the above-mentioned channel estimator 11 is to control all other information of other mobile units operated by the base station. The channel estimator is executed in parallel.

The controller 106 receives beamforming information from each channel estimator 11. Therefore, the controller 106 obtains the spatial information of all signal parts from all mobile units. The controller 106 then downloads this information in the form of coefficients to the receiving bank 14, which uses the spatial information from the channel estimator 11 to improve the signal reception from the mobile unit. Each receiving bank 14 includes a beamformer 112 to form a narrow beam toward the portion of the signal associated with a single mobile unit. Due to selective inspection

98811171.3 The signal part of the first measured intensity. Therefore, the beamformer creates a well-matched spatial filter for incoming signals that include multipath components. The beamformer 112 feeds the spatially filtered signal to the 4 fingers of the conventional IS-95 RAKE receiver 113 (described in US Patent No. 5,309,474). It should be noted, however, that the beamformer output can be fed to other types of receivers, which are known to those of ordinary skill in the art. As a result of the above-mentioned spatial filtering processing, compared with the traditional CDMA system, its carrier-to-interference ratio (C/Ι) is significantly improved. The improvement of C/I is related to the ratio of the created effective beam (approximately 10 to 30 degrees) to the existing antenna beam (approximately 100 to U 120 degrees). It should be noted that the AOA and TOA data are also transmitted to the central controller 120, where the system determines the best downlink beam structure. A discussion of downlink processing will be given as part of the description of FIG. 4.

In another embodiment, the controller 106 may specify a narrow beam (typically 2 to 3 degrees in width) in the wide-aperture antenna array to cover different parts of the scattering area in order to slow down fading. The typical scattering area around a mobile transmitter is defined by a circle with a radius of about 30 to 100 wavelengths. Surrounding large reflectors (such as extremely large buildings or mountains) can create secondary scattering areas that produce (by spread-spectrum receivers) time-distinguishable multipath propagation, thus providing multiple scattering areas. In traditional spatial diversity, signals are collected at different spatial points, where multipath signals arriving at different phase combinations are combined. Therefore, when one antenna exhibits a destructive combination, the other antennas have a high probability of exhibiting the required constructive combination.

By forming beams that are narrow enough to distinguish the energy sets emitted from different sources, while usually pointing these beams in the same direction, the fading caused by multipath propagation can be significantly reduced. With a wide-aperture array, the weighting coefficients of the array can be changed for different beamformers to change the beams in the array. Although narrow beams are to be formed in the antenna array, high split lobes are not considered a big problem in the case of CDMA, because interference is the sum of all other active user energies, and because these split lobes are essentially narrow , Therefore, they are basically suppressed.

If the beam width is narrow enough relative to the size of the scattering area, there will be a different distribution of multipath sources in each beam, and therefore, the power/time function on each beam will not be related to other beams. The angle of the scattering area is usually between 5 and 10 degrees, but its size can vary, depending on factors such as the distance from the base station and the characteristics of the area. For a scattering area of 5 to 10 degrees, a beam width of 3 to 6 degrees is sufficient to distinguish it from other beams. Because a typical Rake receiver can receive 4 antenna beams, this embodiment provides simultaneous uncorrelated power/time function processing. The simulation results shown in Figures 2 and 3 show that the effectiveness of this method is similar to the current space diversity method.

For small angle range conditions (such as 2 to 3 degrees) associated with small scattering areas or distant users, 2 to

98811171.3 p.

Four narrow beams (for example, with a width of 3 to 6 degrees) are arranged adjacently with some overlap to cover the scattering area (for example, with an angle of 5 to 10 degrees). Figure 2 shows the cumulative probability density function (PDF) or cumulative distribution function (CDF) of this arrangement, which represents the power distribution of the symbol amplitude received by a given system. The four curves represent the CDF of various systems: the solid curve 50 on the left is used for systems that use standard space diversity, the dashed line 51 represents a system that uses a single beam pointing to the center of the scattering region, and the dashed line 52 represents that a single beam is used for fast tracking A system of varying AOA associated with varying multipath signals. The solid curve 53 on the right represents the system using the multi-beam arrangement of this embodiment. It can be seen from the shape of the CDF curve that the effect of multi-beam arrangement is similar to that of a single tracking beam, which requires extremely high processing power and standard spatial diversity. It should be noted that the horizontal axis represents the relative gain, which can be changed using different types and/or numbers of antenna elements, etc.

When the angular range increases, the angular interval between the beams also increases, and the beam width also increases. However, the number of beams remains the same. The beam is angularly expanded to sample different parts of the scattering area. In addition, the increase in beam width is limited by the size of the antenna array. As shown in Figure 3, this figure shows the CDF using the system shown in Figure 2, but with a wider angular range (10 degrees). When the angular range increases (most likely due to the gap between the mobile unit and the base station) The distance is smaller), and the fading mitigation effect of the multi-beam arrangement is improved. When the angle range increases corresponding to the increase in the viewing angle of the scattering area, the correlation between the beam outputs decreases, because each beam can be directed to a coverage area with a wider separation interval. As a result, the correlation between beam outputs is lower, which improves diversity or fading mitigation, as discussed in "Mobile Cellular Telecommunications" by William CY Lee.

The angle of arrival/time estimation described above and other descriptions below can be used for single scattering area and multiple scattering area processing. By processing the histogram of the arrival angle samples, the angle range can be determined in real time. When fading is caused by a large scattering area, the angle of arrival results (AOA samples) are distributed with a large amount of variation (and can be estimated by the variation of AOA results). However, the main AOA can be estimated with the histogram center of gravity. The center of gravity of the histogram is determined by using a low-pass filter (for example, Hamming, raised cosine, etc.) to "smooth" the histogram and find the maximum point of the "smooth" histogram.

Therefore, in the above-mentioned embodiment, the arrangement of multiple narrow beams in the wide aperture array uses multiple beams of the highly directional antenna array to achieve fading mitigation. This is because the beams "seal" the entire scattering area and therefore cannot provide Uncorrelated multipath combinations, so a moderate beamwidth (eg, 10 degrees or greater) will not provide diversity.

Although the preferred embodiment adopts the IS-95 architecture, the above processing can be used to utilize a limited alphabet or

98811171.3 The first training sequence is implemented by the wireless protocol. For example, in the GSM system, the training sequence can be obtained in each wireless burst. Because the training sequence is known, the correlation between the incoming signal and the training sequence stored on the receiver will produce the same result as above (assuming that the frequency error is not too large relative to the sequence length). In this case, a training sequence correlator (convolver) is used instead of the despreader 102 and the FHT 103. Because there is only one possibility for a training sequence, there is no need to try multiple possibilities like the Hada code converter in the preferred embodiment does. The system with training sequence used in the present invention will be described in more detail later in this article.

Figure 4 describes the spatial correlator 105 in detail. In this embodiment, the spatial correlator is an independent unit. However, due to the redundant function between the unit and the current implementation of the IS-95 RAKE receiver, the spatial correlator can be integrated with the RAKE receiver. The preferred embodiment is optimized for IS-95 (uplink with M column modulation), but any signal with a limited alphabet (a limited number of symbols) or training sequence can use the same concept. The use of the known signal structure realizes the simple determination of the array response vector, and the calculation and analysis of the complex covariance matrix are unnecessary. Therefore, this method can also be used for GSM and TDMA wireless air interfaces.

The columns of the signal matrix B (ie, the spatial response vector from the FHT) pass through the multiplexer (MUX) 206 and then correlate with the columns of the array calibration matrix A stored in the random access memory (RAM) 203. Theoretically, this correlation process is performed by multiplying the calibration or the common transposition (Hermitian) of the array cluster matrix A by the signal matrix B. The result is the correlation matrix C = A^hB. It is important to note that this theoretical calculation can be implemented in many different ways, all of which are mathematically equivalent to each other. The calibration matrix A is also called the array cluster matrix, and is generated by measuring the antenna array response within the antenna test range. Each column of A represents the response of the antenna array in one of a predetermined set of directions. For example, if the angular range is divided into 360 directions, each of the 360 columns of A represents the response of N antenna array elements from one of the 360 directions of the array. In the calculation of matrix C, these 360 vectors are correlated with the 64 column spaces of signal matrix B to generate a 360 X 64 element matrix, where element i·j represents the jth in the direction of the received signal and the i-th angle. Correlation of code elements.

In this preferred embodiment, the correlation is performed extremely efficiently by using a unique and simple calibration table representation, so that matrix multiplication can be achieved without any multiplication. Each complex value item of the calibration table matrix A is quantized so that both the real part and the imaginary part are represented by only 2 bits. More specifically, each part is represented by 2 bits, one is a value bit and the other is a sign bit, so: (0,0)=-0, (0,1)=+0, (1,0) =-1> (1,1)=+1. Therefore, only 4 bits are used to represent each complex value item. through

98811171.3 The number of array elements is increased to twice that of the current base station array to compensate for the reduced resolution in the simple quantization scheme. This simple bit-plus-sign data structure enables the use of the complex adder 204 to calculate the vector inner product between matrix columns. In a traditional implementation, the vector inner product would require a set of N multipliers. Therefore, the technology of the present invention significantly simplifies the realization of spatial correlation operations = calibration or complex-valued items of the array cluster matrix A may be caused by unforeseen factor changes, such as temperature changes and system component degradation, caused by the analog part of the transmit and receive channels. , Transmit and receive power changes, etc., causing errors. By measuring the phase and amplitude response of the channel, the behavior of the receiving and transmitting channels can be known, so that the entries of matrix A can be corrected. These entries represent the response of N antenna array elements in a given direction of the array. The measurement of the phase and amplitude response of a signal channel requires a "test tone" operation, that is, inserting an analog signal into the channel (its characteristics match the channel frequency and amplitude response), and determining the signal amplitude and phase on the channel output.

In the case of analog or TDMA base stations, the insertion of test tone signals can interfere with ongoing data transmission. If the test tone signal is made low, the accuracy of the test tone will be reduced. CDMA communication allows the test tone signal to be "embedded" in the general data stream without loss of test tone accuracy or interference with the main data signal. Because the data signal is coded and extended (IS-95 or similar), the test tone signal is either unmodulated or coded and extended so as to be statistically orthogonal to the data signal. The "matched accumulator" on the channel output (using a matched despreading code) allows the test tone signal to be coherently decoded (to determine its phase and amplitude), while the data signal's contribution to the detector output (randomly distributed in In phase and amplitude) is zero. The measured phase and amplitude data can be used to correct the analog channel response, thereby eliminating phase and amplitude mismatches in multi-channel receiving and transmitting systems.

In an embodiment of the method, as shown in FIG. 5, the compensation circuits 501 and 502 are respectively connected between the N transmitters 109 and the transmitting bank 12 (FIG. 1) and between the N receivers 101 and the channel estimator 11. (Picture 1) between. The compensation signal source circuit 501 provides the test tone signal to the transmit (TX) and receive (RX) channels. The constant generator A 503 in the compensation signal source circuit 501 provides the test transmitter 504 with a constant value A, that is, provides the test transmitter 504 with a signal of known amplitude and zero phase. The constant generator B 505 in the compensation signal source circuit 501 provides a constant value B to a selected one of the respective transmitters 109 for channel response evaluation.

In order to compensate in the receiving channel, the output of the test transmitter 504 is converted by a frequency conversion module (FCM) 506 to match the RX module, that is, to eliminate the phase and amplitude difference between the transmitter and the RX module. This can be done by measuring these values and then compensating them during matrix calculations. FCM

98811171.3 p.

506 inserts the test tone signal into all receiving channels via the equal phase and amplitude power divider 507. Each of the set of N couplers 508 couples the test tone signal with the corresponding antenna element in the receiving antenna array 10. The signal from each of the N couplers 508 is then fed to a related one of the N receivers 101, where it is down-converted to produce a digital signal with I and Q signal components. The set of N signals are then placed on the receiving bus for use, such as input to the channel estimator 11 and the receiving bank 14 in Figure 1 ο The received output or channel (digital output) to be evaluated is selected and generated from a constant Multiply the signal generated by the amplifier A'509. The signal from the constant generator A'509 is the same as the signal from the constant generator A 503 to decode or despread the test tone signal in the received data signal. Then, the compensation detector accumulator 510 accumulates the digital values (I and Q). The accumulation processing period is only limited by the channel response change rate (assuming extremely low) and the size of the register in the accumulator 510. Therefore, the integration time provided by the accumulation process is sufficient to extract the test tone signal from the mixed signal on the receiving channel under evaluation. Relative to the total signal energy in the channel, the test tone signal can be -30dB. By coherently decoding the test tone signal, the phase and amplitude of the test tone signal can be measured to determine the specific receiving channel Phase and amplitude response. For example, in the determined integration period, the I and Q samples of RX are directly accumulated. The I and Q data includes both the amplitude and phase of the measured channel.

Repeat the above processing for each receiving channel until the phase and amplitude responses of all receiving channels are known. Then, the channel compensation response of each channel is combined to form a "compensation vector" that can be used to correct the amplitude and phase of the measured or controlled data. By dividing each row of the calibration matrix A stored in the RAM 230 (FIG. 4) by the corresponding row vector (complex value) of the compensation vector, each item of the calibration matrix A can be corrected. This operation generates a corrected calibration matrix, eliminating all errors in the receiving channel.

A similar process is used to compensate the transmission channel. The transmission channel selector 511 in the compensation signal source circuit 501 selects a transmission (TX) channel from the transmission bus, which can be transmitted from the transmission bank 12 (FIG. 1). The constant (which can be extremely small) from the constant generator B 505 is added to the signal from the selected channel to be compensated. The constant signal is switched between positive and negative values so that the constant generator B 505 is the same as the constant generator B512. Then, the obtained signal is converted by the group of N transmitters 109, and then sent to a group of N couplers, and combined on the power combiner 513. The combined signal is converted by the FCM 506 and then down-converted by the test receiver 514 to generate digital I and Q signal components. Although this combination can reduce SNR conditions, it allows a completely passive configuration, which is very important when the antenna array is located at the top of the tower (usually inaccessible).

The output of the detection receiver 514 is multiplied by the signal generated from the constant generator B'512. Make from

98811171.3 The signal of the constant generator Pi 509 is the same as the signal of the constant generator B 505 to decode or despread the test tone signal of the transmitted signal. Similar to the correlation of the reception channel described above, the phase and amplitude response of the transmission channel can be used to correct the transmission coefficients in the transmission coefficient table (not shown). Therefore, the compensation system of FIG. 5 evaluates the channel response of both the transmit (TX) and receive (RX) parts of the base station.

Referring back to FIG. 4, the timing generator 201 synchronizes the spatial correlation process to the Walsh symbol period extracted from the base station pilot timing (ie, the end of the Hada code conversion). The N X 64 signal matrix is latched to the MUX circuit 206, which provides the complex adder 204 with one column vector at a time. For each vector, the complex adder 204 performs vector correlation on each column of the calibration matrix A separately. Because the calibration matrix data is only 0, 1 or -1, the data is used in the complex adder 204 to determine whether to perform null, addition, and subtraction operations on each element in the row vector. The RAM address generator 202 is also driven by the same timing generator 201 to synchronize the appearance of the column of calibration data with each latched vector.

It should be noted that the number N of array elements does not change the dimension of the correlation matrix, which is only determined by the predetermined number of symbols in the alphabet and the predetermined number of angular directions. The calibration matrix C is stored in the spatially correlated RAM 207 and processed by the maximum value selector 205. In the preferred embodiment, the selector 205 is a simple serial comparator. The final result of the spatial correlator processing is the best expected AOA for the selected signal part and the associated "inner product" value (used as a determination factor). Only when a predetermined threshold has been exceeded (as described below), the result is reported to the controller 106 (FIG. 1). When necessary, the threshold is updated from time to time. When the threshold has been exceeded, the controller registers the time offset associated with the signal part TOA. This information is used to estimate the range of the mobile unit from the base station. Using recursive processing, more than one maximum can be identified at a time: after identifying the maximum in the correlation matrix, ignore adjacent matrix elements (ignoring adjacent elements minimizes the probability of "non-peak" selection), and execute another "peak" "search for. This feature allows identification of multipath parts that cannot be distinguished only by time (as done in existing Rake receivers), which will enable beamforming reception of multipath signals in a small time range. This method has great advantages when the mobile unit is close to the base station for communication.

The threshold is calculated by averaging the reported results I and Q over a long averaging period window. For example, K reported results are accumulated at the controller 106, and the accumulated results are divided by K. Since most reported results are Non-time-related elements are generated, so the results are "noise-like", and they are averaged to provide a good estimate of the channel noise level. Since channel noise is a linear function of the number of active mobile units, therefore, As mentioned, this level needs to be updated from time to time.

FIG. 6 shows in detail the uplink beamformer 112 of FIG. 1. In this embodiment, the uplink beamformer 112

98811171.3 The first link beamformer is represented as an independent unit. However, the beamformer 112 can be integrated into the channel estimator 11 due to the "bit plus symbol" algorithm that can be made into a very low gate count device. The signals output from the N base station receivers 101 are fed to the complex adder 604 for beam forming. Since the data rate of IS-95 is about 100 trillion samples per second, the complex adder 604 can use current technology to perform at least 4 vector additions for each vector data sample. The beamforming coefficients are downloaded to the coefficient RAM 603 from the above-mentioned controller. The timing generator 601 and the address generator 602 "rotate" these coefficients to the complex adder 604. Using only the addition of complex numbers, these coefficients are used to form the inner product as described above with reference to the spatial correlator of FIG. 4. The vector addition result is fed to the interface unit 605 for transmitting the result to the Rake receiver modem. In other embodiments, any modem based on a limited alphabet or training sequence protocol can be used. The effect of the beamformer 112 is to spatially filter the incoming signal to preferentially select the signal arriving from the known direction of the signal portion of the specific mobile unit. Signals from other directions are attenuated and the reception of the desired signal is improved.

Figure 7 shows an example of the spatial distribution of downlink beams. Downlink management is very different from the uplink because IS-95 is not a symmetric protocol, and the difference between the uplink frequency and the downlink frequency is at least 60MHz (cellular). The frequency difference makes the uplink and downlink channels uncorrelated. The AOA and TOA of the uplink and downlink can be very different, although they are statistically similar. Therefore, as described above, the downlink can only be estimated statistically based on the data collected in the uplink. In addition, the downlink needs to broadcast the pilot signal to the relevant mobile unit. As a result, a single downlink beam is impossible, that is, only "mobile group" beams can be realized. Therefore, the downlink method is based on a combination of wide and narrow beams determined by data collected in the uplink.

Therefore, referring to FIG. 1, the transmitting beamformer 117 in the transmitting bank 112 is coupled to the spatial correlator 105 via the central controller 120 to receive AOA and TOA data for generating a spatial beam according to beamforming information. These spatial beams are selected from a set of calculated beams containing narrow beams and overlapping wide beams, and the narrow beams are phase-matched with the overlapping wide beams. The beamformer 117 employing a conventional digital beamformer takes signal samples (scalar I and Q) and multiplies them with weighting coefficients to generate a vector, where each element includes a scalar representation of the signal entering a single antenna. The routing circuit 116 and the routing and summing circuit 115 are data switches for routing signals from multiple transmissions to the beamformer 117 and the transmitter 109. The beam configuration is determined by the distribution of mobile units around the base station. A wide beam is required to ensure proper coverage very close to the base station, where most downlink signals reach the mobile unit by scattering from nearby reflectors. The system adjusts the wide beam 701 to ensure proper coverage of mobile units close to the base station. Mainly adjust the narrow beam 702 to adapt to the "long-range"

98811171.3 The mobile unit. Since most mobile units will be located in a relatively distant coverage area, it is expected that a narrow beam will serve most mobile units. The increase in the number of downlink beams increases the softer handover, and therefore affects the growth of passenger traffic. Therefore, extreme care must be taken when specifying the beam in the downlink.

The increase in downlink capacity can be estimated as follows:

Q*P+Q*P/N+X*P+X*P/N+(Q( 1 -P)/N+X(l -P)/N)*( 1 +B)=Q =1____________

QP + P/N + (1-P)(l + B)/N Assuming that the mobile units are evenly distributed, the maximum illumination is Q>the latter is the maximum number of simultaneous transmission channels including softer handover.

The Q*P term is the number of mobile units that come in a high angle range called "wide angle". Q*P/N is a part of the wide angle part, these wide angles are in the narrow beam, and all are in softer handover, so the irradiance is added twice in the overlapping sectors.

If X mobile units can be added as a result of beam combination, then X*P additional wide-angle types are added (assuming P remaining as mentioned above). In addition, X*P/N follows the same Q*P/ N the same rule.

In the narrow beam space, Q(lP)/N+X(lP)/N mobile units are obtained. However, since some handovers are caused by overlap, their irradiance values must be increased by a factor of 1 + B. B is the ratio of the number of users in handover to the total number of users, which can be determined by experiments. B can be kept small because the distant cells associated with the mobile unit will naturally require more narrow beams.

Fig. 8 is a graph of the capacity increase ratio relative to the number of narrow beams and the multipath probability/10 in a wide angle range with a softer handover probability fixed at 20%. Fig. 9 is a graph of the capacity increase ratio of 4 narrow beams, which is a function of the handover probability/10 and the wide angle range multipath probability/10. Fig. 10 is a graph showing two cases of variable handover rate and 4 narrow beams with wide angular range and multipath. Figure 8-10 shows the capacity improvement compared to non-adaptive array base stations.

According to the above analysis, the capacity of 4 narrow beams in a wide beam is increased to about 2. If the mobile units are not evenly distributed, the improvement can be even higher. Fig. 11 is a graph of the expected capacity ratio for different density changes of mobile units. This improvement requires adjustment of narrow beam boundaries to avoid peak mobile unit density. For example, these boundaries can be adjusted according to the description with reference to FIG. 12. This adjustment must be done step by step to avoid excessive handoffs when changing beams.

Fig. 12 shows a flowchart of downlink beamforming determination processing. In block 1200, mobile unit spatial data is collected and stored in memory. Then in block 1205, through the data

The 98811171.3 category is a two-dimensional histogram. This data is used to evaluate the distribution of mobile units around the base station. In block 1210, the histogram "peak" is identified as follows: a two-dimensional "smoothing" filter is used to eliminate noisy histogram "burrs", and the usual two-dimensional "peak search" processing is used. For a system capable of forming M downlink beams, in block 1210, M "peaks" are found. After the associated pilot signals are used to classify mobile units in block 1215, in block 1220, the number of mobile units around the M highest histogram peaks is counted. In block 1225, for each of the M peaks, the mobile unit count value is compared with the pilot count value of the closest beam. Then at block 1230, the mobile unit peak count value is compared with the pilot mobile unit count value. If the mobile unit peak count value is close to the pilot mobile unit count value, then at block 1200, the next set of spatial information is stored. However, if the mobile unit peak count value is not close to the pilot mobile unit count value, then at block 1235, the closest beam is moved to the peak. Then, at block 1225, the pilot count value of the shifted beam is compared with other mobile unit count values. Therefore, a closed loop process adjusts the boundaries of the downlink beams and uses them to equalize the number of mobile units. Narrowing the beam will cause some mobile units to handover to different pilot frequencies, Instead, only the mobile units close to the selected "peak" are placed on the relevant pilot. This process is performed at a very low speed to avoid excessive handoffs.

Figure 13 shows a device for generating an antenna array cluster (or calibration) matrix A. The antenna array 1301 provided with an assembly of antenna elements is installed on an overhead antenna mast connected to the turntable 1304. The controller 1306 instructs the turntable to rotate or rotate the number of angular directions of the array cluster A in a predetermined angular step. The network analyzer 1305 transmits an RF signal at a specific angle via the transmitting antenna 1302, and the signal is received by the antenna array 1301. The signals received by the units in the antenna array 1301 are routed through the RF switch 1303 to the network analyzer 1305 for measurement, which is well known in the art. In this preferred embodiment, the antenna array is circular, but the present invention can also be implemented using any array shape. In this case, the RF signal collected by each antenna unit can be expressed as follows: .2 Redundant Rcos (27tk/M-8)^Ak,e ^{= F}k,e*^{eJ λ} , Where α represents the array cluster function, k is the number of elements, e is the relative angle of arrival (established by rotating the array relative to the RF signal source), M is the total number of antenna elements in the circular array, and input is the RF signal wavelength. The data is collected and stored in the controller 1306, which also includes a data storage unit.

By performing spatial correlation with high angular range and unpredictable multipath in a fast fading environment, the array cluster information can be used to more accurately determine the multipath angle of arrival (AOA) value and coefficient. As mentioned before, empty

98811171.3 The second processing involves estimating the array response vector of the IS-95-based CDMA signal (including the electrical amplitude and phase of all antenna array elements) to determine the multipath angle of arrival (AOA) value and coefficient through spatial correlation. Then, these The coefficients are used (through down-conversion to baseband) to optimally combine multiple input lines. Therefore, the ability to accurately estimate the response vector of the array is an important goal in a CDMA system. However, the estimation accuracy is limited by the fading rate (Doppler shift caused by moving mobile units), because when the fading rate is Or when the Doppler rate increases, the time to collect coherent data decreases. This problem becomes more serious when the cellular system moves from the 800MHz range to the 1900MHz range or higher, which can increase the fading or Doppler rate by a factor of 2 or more.

In addition, for frequency division duplex (FDD) systems, the forward link (transmitted from the base station to the mobile unit) and the reverse link (transmitted from the mobile unit to the base station) occupy different carrier frequencies or frequency bands, but overlap in time. This difference in the frequency of the forward and reverse links reduces the correlation between the fading of the two links, so that space diversity can only be used in the reverse link, but not in the forward link , That is, it is impossible to accurately determine the array response vector estimation of the forward link array coefficients.

Various array response vector estimation methods have been proposed, some of which are characterized by using some knowledge of the time and space structure of the signal incident on the antenna array. The knowledge of the time structure of the signal (requiring known training sequence, pilot signal, constant envelope, etc.) leads to some algorithms, such as MMSE (minimum mean square error), CM (fixed modulus), etc. They are sometimes called "blind" Or "semi-blind" estimation technique. The blind technology does not use any prior information about the signal time structure and antenna array clusters, while the semi-blind technology can use the time structure. The main disadvantage of these blind technologies is the long integration time required for convergence, especially when the number of interference sources is large (CDMA is more typical). Because specific interference is not considered, this will reduce the efficiency of the solution. In addition, when a dedicated pilot signal is used on the reverse link, the pilot signal power needs to be low in order to minimize the capacity loss in the reverse link. However, the lower power pilot in coherent demodulation requires a longer integration time to ensure sufficient reference signal quality. In addition, when it is not known or the time of arrival (TOA) of the signal changes, a continuous "time search" is required. Therefore, slow convergence processing will be prohibited for each hypothesis. In a CDMA type system, signal timing must be restored before demodulation. Therefore, the search process is performed through a series of assumptions, through which the system changes the time of the reference correlation sequence and cross-correlates with the incoming signal (for example, IS-95C or cdma2000). Ru Gao Zhi Line matched filter assumption (W-CDMA), you need to adjust the sampling point. The time required for each hypothesis must be short enough to be able to perform a fast search (the hypothesis cannot be determined before spatial processing, because there may not be sufficient signal-to-noise ratio at this point).

The above-mentioned various algorithms use the statistical characteristics of the signal, but they do not use the array (ie, the array

98811171.3 cluster) any knowledge of the spatial characteristics. Although the array response vector deviates significantly from the array cluster in the environment of high angular range and unpredictable multipath structure, even partial knowledge of the array cluster can significantly reduce the required data integration time and speed up the calculation process. By using the knowledge of the relationship between the outputs of different antennas, the array cluster information enables real data processing in the spatial domain to perform two-dimensional averaging simultaneously in time and space.

As mentioned above, in a frequency division duplex (FDD) system, the coefficients in the reverse link array response vector are only statistically related to the ideal array response coefficients in the forward link. This is because in the two links There is a frequency difference between them. Therefore, the forward link transmission coefficient must also be determined by combining the estimation processing of the received power indicator fed back at the mobile unit. Due to the variable shadow or fast fading conditions unique to the mobile environment, the power feedback method that does not perform forward link estimation first may be extremely slow or may not converge at all.

However, the knowledge of the array cluster can be effectively used to speed up the determination of the weighted vector coefficients in order to improve the signal-to-interference ratio (SIR). Therefore, the CDMA network capacity can be significantly improved. In most environments, such as rural, suburban, and urban areas, emission sources close to the base station produce a large number of wide-angle range multipaths (continuously distributed in time and space). This will "push" the array response vector away from the array cluster, that is, increase the Euclidean distance between the array response vector and the array cluster. However, the distant mobile unit provides a more discrete distribution in time and space, that is, signal paths that can be distinguished in time have a smaller angular range. Since most users are located at the edge of the cell (which is the biggest problem in terms of capacity), the array cluster assisted estimation (MAE) becomes very practical, in which, by finding the array cluster closest to the measured array response vector, the array cluster Auxiliary Estimation (MAE) estimates the vector of array weighting coefficients. Although the availability of array cluster knowledge decreases when the mobile unit is close to the base station, the SIR for the entire cell is significantly improved.

The CDMA demodulator includes a time-of-arrival (TOA) search mechanism and multiple demodulation channels based on the fast Hada code converter used in the IS-95 system (in Andrew J. Viterbi, CDMA Principles of Spread Spectrum Communications) or multiple PSK demodulation channels used in other systems (eg, cdma2000, W-CDMA, and UTRAN). Each demodulation channel is usually connected to the selected antenna and tuned to the TOA determined by the search mechanism. The demodulation results of all demodulation channels are added together (coherently or incoherently combined), and a trade-off is made between performance and complexity. Coherent combination needs to determine the relationship between all elements to be combined to ensure the largest constructive combination (weighted vector). Incoherent combination is performed by squaring all combined elements, thereby avoiding potential destructive combination by eliminating the phase between the combined elements. Incoherent combination is simple and easy to implement, but not very effective, while coherent combination may be more effective, but requires complex search. When there is enough

98811171.3 For the first (ie, identifiable) TOA extension (for CDMA, it is the reciprocal of the chip rate, and for IS95, it is the chip duration, that is, 800 milliseconds), multiple demodulation channels provide time diversity , Which can enhance the standard spatial diversity used by most cellular base stations.

As described above, by calculating the weighting vector as part of the spatial processing, the demodulation results of multiple demodulation channels can be linearly combined to enhance system performance. The signal-to-noise improvement can reach 10*logM, where M is the number of antenna elements.

The effective signal combination in the antenna array depends in part on the ability to estimate the weighting vector (coefficient) used for the combination:

PW^T-V Among them, P is a scalar value representing the result of the combined processing, W is a weighted vector, and V is an array response vector.

In order to achieve the goal of weight vector estimation during the channel coherence period (less than 1/4 of the reciprocal of the Doppler rate), the combination coefficients must be estimated quickly. By using an antenna array to estimate the multipath profile in the fading channel and processing the data for beamforming, a simple method can be used to enhance the CDMA demodulation process under fast fading conditions. The CDMA demodulation process can be enhanced by using array clusters in the estimation of array coefficients in both the reverse and forward links, and by using antenna arrays to estimate the multipath profile in the fading channel and process the data to perform beamforming.

The multipath profile can be defined as the two-dimensional distribution function of multipath power to AOA and TOA. When the signal reaches the antenna array, the antenna output is collected into a single vector called the array response vector. The array cluster is generated by segmenting the signal arrival angle (in two or three-dimensional space) to form an array response vector set. The characteristics of each antenna array can be determined by the array cluster. The array cluster is a trajectory in an M-dimensional vector space, where, as described above, M is the number of antenna elements.

In the case of non-multipath (ie, ideal wavefront), the array response vector "touches" a point on the array cluster, that is, the Euclidean distance is zero. When multipath occurs, the array response vector is the linear combination of all arriving multipath wavefronts. In this case, the array response vector is "far away" from the array cluster, that is, the Euclidean distance increases. The distance between the array cluster and the array response vector increases statistically as a function of multipath signal level, multipath angular range, and interference power. Interference includes a combination of thermal noise and other incoming transmitted signals. As for the large number of contributing random units, it is assumed that the distance from the array cluster to the array response vector has a Gaussian distribution, where the mean value depends on the array cluster itself, and its change is related to the above-mentioned units. As the Euclidean distance increases, the angular range increases. Assuming that thermal noise and other transmission interference can be basically eliminated by integration and despreading (CDMA), the array response vector and the array

98811171.3 The main factors for the large separation between the clusters are the multipath level and angle range.

The operation of the spatial correlator can be described by the following operation:

Ω = V^h-A where V is the array response vector (H stands for Hermitian symbol), and A is the array cluster matrix (with columns marked with 0). Each row in A represents an element of the array cluster, and each column in A represents an angle in the array cluster. The result of spatial correlator operation is a vector. , Its amplitude value corresponds to the correlation value between the array response vector and the array cluster at all possible angles (array cluster index, 0). Sorting the magnitude of the largest zero element means choosing the best available fit (the theoretical maximum when the weighting vector W depends on the array cluster), that is, having a point that touches the array cluster. In the case where the angular range exceeds the array beam width (array beam width is well known in the art), the above processing can be described as an angular fan that includes Rayleigh fading signal sources (their combined power is constant) Move the beam in the area and search for the maximum value at a given distance. The larger the sector, the larger the sampling group, which increases the probability of obtaining a high power value.

Since all the above operations are linear, both the relative amplitude and phase of the incoming signal are preserved (Q on the selected 0). Therefore, this processing can be used in two schemes (eg, M-dimensional) incoherent demodulation and coherent phase demodulation (ie PSK). In pilot assisted or coherent demodulation, the relative phase estimation of each signal path (rake finger) can be faster, and therefore more accurate in fast fading environments, resulting in a better solution on the Rake receiver Adjusting efficiency and more precise combination of coherent fingers.

In order to construct the multipath profile, spatial processing is used to estimate the AO value. Figure 14 shows a possible embodiment of a 2D CDMA demodulator, which is used to non-coherently demodulate the transmitted (IS-95 reverse link) signal to estimate the signal AOA value. In Figure 14, a single "finger" (demodulation channel) of the cluster-assisted spatial demodulator used in the IS-95 system is shown. This type of demodulator uses the knowledge of array clusters (created in a clean environment, that is, of non-scattering sources) to enhance the demodulation process. In this case, the incoming array response vector is cross-correlated with the array cluster matrix to provide a "magnifying glass" effect. The system only "looks at" the signal after the spatial correlation (magnifying glass) has occurred, because only then is the signal-to-noise ratio sufficient for any determination. All MAD implementations include multiple MAD "fingers" (at least two for minimum time diversity). The MAD finger can perform two functions of time search and demodulation.

The I and Q components of the signal are fed from an antenna array with M elements. The output of the M antenna elements is down-converted to baseband frequency and digitized. Then, when each of the M signals is multiplied by the appropriate long and short codes from the code generator 1405, the M signals are followed by M parallel related signals.

98811171.3 The first channel is despread (as described above, or see Andrew J. Viterbi, "CDMA Principles of Spread Spectrum Communications (CDMA Principles of Spread Spectrum Communications)"). After despreading, these signals are input to a row of fast Hada code converters (FHT) 1410. The complex value outputs of M FHTs are then fed to M multiplexers 1415 to multiplex these outputs 64 (for IS-95) possible array response vectors (for each possible symbol), and fed to the spatial correlator 1420 one by one. The spatial correlator performs the aforementioned spatial correlator operation on each of the 64 candidate array response vectors. Each array response vector is cross-correlated with 256 vectors in the array cluster in the spatial correlator 1420 according to the aforementioned spatial operation, but other numbers of vectors may also be used. The number of potential complex multiplication and accumulation (MAC) operations required for each symbol (assuming 256 possible angles and M = 16 antenna elements) is:

NN=M*64*ANGLE_RANGE=16*64*256=262,100.

This corresponds to 262,100*5000=1.31P10 per second¹. A MAC operation, where the rate is 5000 Hz, because the IS-95 symbol duration is 200 milliseconds. Then, the estimated AOA output from the spatial correlator can be used for further processing, as described below.

As mentioned above, if a sufficient number of antenna elements (ie, 6 or more) are used, the array cluster can be represented by extremely low resolution or a small number of bits, and the loss of the amplitude of Ο is not too much. . Reducing the number of bits can simplify ASIC implementation and reduce the required processing speed, because real multipliers are not required and the memory size requirements are also small. Therefore, the above-mentioned processing becomes achievable in a medium-sized ASIC.

In another embodiment, as shown in FIG. 14A, the maximum absolute value sorter 1425 selects the maximum value among the matrix Ω (AO and Walsh symbol index) obtained by the spatial correlation operation; the matrix is 64 X 256-size matrix (M is 64, and the number of angle steps in the cluster calibration table is 256). Repeat the spatial correlation operation several times (the number of times depends on the available coherence time, that is, the Doppler period divided by a number ranging from 5 to 10 ). The obtained ΑΟΑ value groups are averaged to determine the columns in the cluster matrix. The columns represent the vectors in the array clusters. This column is used as the weighting vector for the weighting vector of the next coming Walsh symbol. This The "next symbol" generates a matrix containing 64 possible array response vectors, and these vectors are respectively multiplied by the weight vectors selected above to generate M vector values again. The remaining processing is well known, as described in Andrew J. Viterbi, CDMA Principles of Spread Spectrum Communications (CDMA Principles of Spread Spectrum Communications)^M > Page 100, Figure 4.7.

If a pilot signal (or a known continuous training sequence) is embedded in the transmitted signal for coherent demodulation, the FHT in Figures 14 and 14A can be replaced by a standard complex accumulator 1505 (the demodulator in Figure 14 becomes As shown in Figure 15), for example, the serial number discussed above is No.

98811171.3 p.

60/077,979, entitled "Capacity Enhancement for W-CDMA Systems" US patent provisional application. In the cdma200 system described in "The cdma2000 ITU-R RTT Candidate Submission produced by TR45.5 (TIA)" or any other embedded continuous pilot or training sequence assisted demodulation scheme, a coherent demodulator or AOA estimator.

The results of the despreading channels are combined to form an array response vector of Ινί values, and in the embedded pilot signal or training sequence, only a single possible symbol is used instead of 64 possible symbols. Search for irrelevant estimates. In this case, since the data symbols are known, the amount of calculation is significantly reduced. Therefore, for 16 antenna elements or a 16-value array response vector, the potential number of MAC operations is:

NN=M*ANGLE_RANGE=16*256=4096 This corresponds to 4096*calculation rate=4096*10000=4.096*107 MAC operations per second.

Figure 16 shows an embodiment of a general CDMA AOA/MAG (magnitude of G) estimator. In Figure 16, a single "finger" (demodulation channel) of the cluster-assisted spatial demodulator suitable for IS-95 (A, B or C), cdma2000 and W-CDMA/UTRAN proposals is shown. In this embodiment, the despreading mechanism can follow the W-CDMA (NTT/DOCOMO) and UTRAN (ETSI/SMG) proposals submitted to the ITU, which correspond to the initiated ITU third-generation cellular IMT-2000. The main difference from the current IS-95 (A and B) standard is the presence of pilot signals in the tile-direction link. IS-95C and cdma2000 proposals use continuous pilot signals, while W-CDMA uses equally spaced short bursts of pilot signals. The details of the reverse link structure are described in the CDG cdma2000 and ETSI/SMG&NTTDOCOMO W-CDMAUTRAN/ARIB proposals submitted to the ITU in June 1998. The proposal is hereby quoted in its entirety for reference.

The I and Q components of the received signal are fed into an antenna array with M elements. The output of the M antenna elements is down-converted to baseband frequency and digitized. Then, as described above, the M signals are despread along M parallel channels. For W-CDMA, in the initial mobile station access phase, the despreading block 1605 can be bypassed. When the timing of the W-CDMA mobile station has been established, the despreading block can be used for descrambling. For cdma2000, as suggested in the CDG proposal (using long and short codes), despreading block 1605 can be used. The above proposal submitted to the ITU provides other details about despreading.

The next stage on the output of the despreading block 1605 includes a row of M matched filters 1610 for W-CDMA or IS-95 or a row of M accumulators 1610 for cdma2000. These matched filters are gated to be used for discontinuous pilot signals as proposed in W-CDMA and UTRAN,

98811171.3 First, as recommended by UTRAN and W-CDMA, use a matched filter for the combination of a 256-bit sequence and scrambling code. The matched filter correlates the incoming signal sequence with the pre-stored sequence. The output of the matched filter is fed to the space passer 1615 to use the above formula

Ω = V^H-A to determine the best fit in the array cluster matrix.

It should be noted that the signal arrival time can vary, and the signal arrival time needs to be tracked. Because of the low signal-to-noise ratio conditions, the ability to determine when the matched filter produces a response to the training sequence may be limited, so it is necessary to repeat the assumption (ie, change the sampling time). This can only be done after spatial correlation, so it requires extremely fast spatial correlation calculations.

For W-CDMA, a new array response vector group is generated every 0.625 milliseconds. In the case that multipath can be distinguished in time, several array response vectors can be sequentially generated in the time frame. This time separation depends on the multipath TOA range. In a multi-path environment, there can be up to L distinguishable multi-path elements (for example, L = 3). Since the timing and phase of the training sequence cannot be accurately determined before the spatial correlation block ("magnifying glass), therefore, Time-varying sampling (time search) that requires matched filters adopts time increment assumptions. Each assumption requires spatial correlation processing. Therefore, spatial correlation processing determines the search time required to capture the mobile station. The current spatial correlator design allows high Up to 200,000 spatial correlation operations per second. For fast fading conditions, the estimated update rate of the time hypothesis can reach 500,000 times per second (1000 times faster than the maximum Doppler rate). In this case, for 16-element antennas or 16 Value array response vector and 256 possible angles, the number of MAC operations per second is 4096*500000=20.48*10⁸*L. If L=3, the number of MAC operations per second is 6.144*10⁹. Using the aforementioned low-bit count algorithm, this rate is extremely feasible with current ASIC technology.

For cdma2000, the output of the despreader is combined in an accumulator to form an array response vector of M values, and a search is performed for a single possible symbol embedded in a continuous pilot signal. This time search is similar to the search in the IS-95 system. It is also possible to support different TOA multipaths by using a single searcher (using the same spatial correlator or repeating the "finger" described in Figure 15). Figures 14, 14A, 15 and 16 describe the mechanism for processing a single path of arrival time. In the case of multiple arrival path times, multiple modules are recommended. In addition to only adding additional modules, different embodiments can share some common circuits to reduce the size and cost of the entire circuit.

Once the estimated AOA data is obtained from the spatial correlators in Figs. 14, 14A, 15 and 16, for example, the data is processed to enhance the performance of the receiver. This includes three aspects: 1) The orientation must be formed.

98811171.3 A beam with sufficient gain for the incoming signal; 2) Spatial diversity must be set; and 3) A downlink beam must be formed.

The data from the demodulator (Figures 14, 14A, 15 and 16 above) is collected to form an AOA histogram. Since the mobile unit provides a varying wavefront (the wavefront is a linear combination of many incoming wavefronts from multiple scatterers), the continuous accumulation of AOA samples can create an AOA histogram. The histogram will have a "peak" in the direction of the main scatterer and in the distribution following the angular range of the transmission source. The significant advantage of the AOA histogram is that even in the case of discontinuous transmission (for IS-95 CDMA systems) , Can also distinguish these peaks. After determining the AOA histogram friction value and variation, the beam can be formed in the direction related to the peak, and its width follows the histogram variation. In the case of a single AOA peak, the system can be in the direction Multiple beam offsets are formed in the direction of the main direction. If the array is large enough, the power of the signal extracted from each beam has a low correlation. The correlation is extracted from the inner product of different columns in the array cluster matrix Yes. With most CDMA systems using some combination of rake, each rake channel can be connected to a different beam. This configuration realizes the first two aspects mentioned above: gain and diversity.

Another feature of AOA histogram processing is the ability to estimate downlink beams, which is the third aspect mentioned above. Although there is a difference between the reverse and forward link frequencies in the FDD system, there is still a good statistical relationship. Therefore, the forward link beam is formed using the direction and distribution following the histogram distribution. In the new generation system application, the pilot signal can be obtained on the forward link, and therefore it is not necessary to adapt the phase coherence between the main frequency of the system and the forward traffic channel too laboriously. For the IS-95 system, as described above, matched phase beam synthesis is used on the forward link.

In another embodiment, as shown in FIG. 16A, the phase rotator 1620 and the inner product multiplier 1625 that can be integrated on the demodulator of FIG. 16 further process the result of the spatial correlator 1615. In both IS95C/cdma2000 and W-CDMA cases, the array response vector (or array response vector group) can be generated by integrating the data within the time limited by the Doppler rate (or a small part of it). Minimize the hysteresis error. In order to perform demodulation and beamforming, it is necessary to estimate the weight vector and carrier phase (PSK). The time for this estimation is limited by the coherence period, which is the same as a small part of the Doppler period. A spatial correlator that enhances the signal-to-noise ratio can quickly determine these values. The result of the spatial correlation is a pointer to the best fit column in the array cluster calibration table. The resulting correlation matrix. The index of the maximum value is the pointer. The phase of the maximum value (selected as part of the spatial correlator processing result) is the carrier rotation phase. The selected column (W) in the array cluster matrix containing the maximum element of the matrix from the spatial correlator 1615 is fed into the phase rotator 1620 to shift the column W. The shift is by adding

98811171.3 p.

W is multiplied by e) 0, where phase 0 is the argument of the maximum value selected from the vector G obtained by the spatial correlator.

Ω = V^h-A Then, the shifted column W'is fed to the multiplier row module 1630 of the inner product multiplier 1625, which also includes an adder circuit 1635, which performs a typical beamforming operation. The difference is that The phase of the weighting vector W'is adjusted to the response vector of the incoming signal array to maximize the PSK (Phase Shift Keying) demodulation result. The multiplier row module performs the following operations to perform beamforming: The column offset selected from A (W*^)"*array response vector.

The efficiency of the demodulation process depends on various factors, such as the accuracy of the array cluster calibration matrix selection (ie, AOA estimation), the accuracy of the rotation phase estimation, the amount of angular expansion, and the SIR. Figures 17 and 18 show the performance comparison between the standard two-unit diversity array and the above-mentioned MAD system under the same signal fading condition. Figure 17 shows the result of using QPSK (Quadrature Phase Shift Keying) MAD during random fading. Figure 18 shows the result of using a standard QPSK demodulator under the same fading conditions as in Figure 17. Simulation results show that for MAD-based systems, the average improvement is about 6 to 8 dB.

For a typical base station, the number of synchronized mobile station sessions can reach 100 or more, which will require multiplying the number of MAC operations per second by 100 or more. The ability to reduce the number of bits in this process enables actual ASIC implementation. Each voice or data channel is equipped with a spatial correlator that performs the above operations. The result is a spatially enhanced demodulator, that is, for each received symbol, the system searches for the best way to coherently combine the outputs of all antenna ports.

The effectiveness of this spatially enhanced demodulation is enhanced as the mobile unit moves away from the base station. This is because the greater the distance, the smaller the multipath angle range, and therefore, the array response vector is closer to the array cluster. When the array response vector is closer to the array cluster, since the multipath angle range is smaller, the accuracy of signal AOA and amplitude estimation increases. Assuming that the mobile units are evenly distributed on the network, most of the mobile units are in the outer area of the cell. In addition, the farther the mobile unit is from the base station, the more difficult it is to maintain communication. Since it is difficult for remote users to maintain communication, solutions for remote users are preferred. Therefore, the accuracy deterioration near the base station and the lowering of the demodulator efficiency are tolerable.

The embodiment of FIG. 19 shows a training sequence convolver. In some wireless standards, it can be used to replace the despreader 102 and FHT 103. The data register 1902 is a first-in first-out (FIFO) unit, and its word bandwidth is The receiver I and Q output widths match. The I and Q samples are shifted by the data register 1902 in twos complement format. The XOR gate is used to compare the most significant bit of I and the most significant bit of Q with the training sequence bits stored in the training sequence register 1903. The resulting XOR output

98811171.3 is fed to the adder 1901 and used to determine whether to add or subtract each I and Q sample in the data register. The output of the adder is updated every sampling period and compared with a threshold in the amplitude threshold detector 1904. When the threshold is exceeded, the I and Q values are registered as components of the signal response vector, which are sent to the aforementioned spatial correlator.

FIG. 20 shows an embodiment of the present invention including two functions (in terms of angle and time) of search and tracking. The addition of angle tracking improves the system's ability to effectively guide the receive beam at all times. The searcher 2000 still captures the new multipath parts, and the tracker 2003 tracks them. The operating principle of this embodiment is very similar to the embodiment described in FIG. 1. The main difference from Figure 1 is the addition of Tracker 2003. The N receiver outputs are fed to the beamformer 2012 in parallel. What the controller 2001 downloads to the beamformer 2012 is not one, but two beamforming information groups for each signal part to be tracked. These two groups correspond to two adjacent columns in the calibration matrix. This allows the beamformer to continuously "trigger" between two angled adjacent beams.

The beamformer output is fed to a "lead/lag gate" module 2013 known in the art. The result of the combination of the "triggered" beamformer and the "lead/lag gate" is in the form of four-level values, corresponding to: left beam/lead time, right beam/lead time, left beam/lag time, and right beam/lag time . Since the tracker is designed to track 4 multipath parts at the same time, the result is reported to the controller via the multiplexer 2015. By exchanging the beamformer coefficients and advancing/delaying the clock of the gate, the controller 2001 guides the beamformer and: advance/lag gate" to balance all the above 4 values at the same level. Angle tracking is through equalization and right sum The left-related results are achieved, while time tracking is achieved by equalizing the values related to lead and lag. This embodiment guarantees sufficient points for reliable tracking. When the searcher finds the output is much higher than the tracked output When a multipath part of the level output of the, the coefficient group is replaced as a whole. In this embodiment, each channel is assigned its own downlink beamformer 2030. It should also be noted that this embodiment supports the corresponding A single beam for each active channel.

FIG. 21 shows a schematic diagram of a base station using the channel estimator/tracker/beamformer shown in FIG. 20. The antenna array 2100 is coupled to a set of receivers 2101, which are all driven by a common local oscillator 2104, as shown in FIG. The receiver output is placed on the data bus 2110 to feed multiple channel estimators/trackers/beamformers 2105, each of which provides multiple signal parts to the BTS channel unit 2106. The unit 2106 may be a rake receiver/data transmitter of IS-95. The channel unit feeds the downlink data to the channel estimator/tracker/beamformer, which feeds the beamformed data to the summation unit 2107. The summation unit outputs the summed beamformed data to the BTS transmitter 2109 These transmitters 2109 are driven by a common local oscillator 2108. Transmitter input

98811171.3 The first output is sent through the transmitting antenna array 2111.

When applied to a CDMA IS-95 base station, the aforementioned embodiment of the downlink requires an additional "pilot". This may require some changes in the design of network control and network pilot allocation. By distributing overhead channels (pilot, paging, and synchronization) through a wide beam, while simultaneously transmitting traffic channels through narrow beams directed to the relevant mobile units, the following embodiments reduce this requirement. This method does not change the traditional ETS softer handover scheme, and therefore, does not require any changes in the network architecture.

The proposed method is realized by the fine array beam synthesis technology well known in the art. Specifically, the beam is configured to be phase-matched in the scattering area of the mobile unit. The coefficients of the beam are calculated to achieve the same wavefront between the pilot and service signals. Therefore, the current IS-95 coherent demodulation can be performed on the mobile unit. This "beam matching" is achieved by beam synthesis based on the minimum root mean square method. This method allows +/-10 degree phase matching down to the -10dB point, which is sufficient not to degrade the performance of the coherent demodulator on the mobile unit.

The coefficients of each downlink beam are set as follows: overhead data (pilot, synchronization, and paging) is transmitted through a fixed, relatively wide beam. The beam of the downlink traffic data is set to match the azimuth line measured by the uplink channel estimator, which has sufficient width margin to compensate for the azimuth error (due to the lack of correlation between the uplink and the downlink ). It should be noted that even relatively wide downlink traffic beams will provide a significant capacity increase.

Since the angular range becomes larger when the distance from the base station decreases, the narrow beam width is estimated based on the estimated distance from the BTS. This distance is derived from the delay measured by the beam steering.

Since the above method is based on the statistical distribution of the scattering area (considering various scattering models), the system must have a few exceptions: First, the allocated service has a narrow beam wider than required, and in forward power control, It gradually narrows with the frame erasure rate (similar to the bit error rate) reported on the uplink. When the frame deletion rate increases, the service beam becomes wider. This mechanism will also compensate for the large difference between the uplink angle of arrival (AOA) and the downlink AOA.

Although the above embodiments describe the use of the present invention in the current CDMA communication system, the concept of the present invention can also be used in a wideband CDMA (W-CDMA) communication system to increase the system capacity. More specifically, the W-CDMA system uses multiple antennas arranged in a wide aperture array and digital signal processing to estimate the multipath angle of arrival (AOA) and time of arrival (TOA), so that the multi-antenna beam can be assigned to the incoming Signal part, and specify an adjustable downlink beam to improve system capacity. Even though the W-CDMA specification has not yet been clearly defined, there are already some specific principles, such as the presence of pilot signals in the uplink, and W-CDMA adopts these principles (for example, IS-665 and J-STD-015

98811171.3 Part 1) in order to be able to provide effective adaptive array antenna technology implementation.

In order to increase the capacity of the W-CDMA communication system, the following features have been proposed: uplink channel estimation, uplink wave formation to provide enhanced array antenna gain, space diversity and fading mitigation (via diversity), and use To provide enhanced array antenna gain and spatial directivity downlink beamforming, which will be described in order below.

As mentioned earlier, the signal despreading and Fast Haada Transformer (FHT) can be used to estimate the array response vector (electrical amplitude and phase of all array elements) of IS-95 CDMA signals to determine multipath AOA through spatial correlation value. However, the pilot frequencies existing in the uplink of the W-CDMA system (this is the basic difference between W-CDMA and IS-95 CDMA systems) can be used to determine the array response vector and channel impulse response (CIR) to determine the uplink Channel estimated AOA and TOA values. Figure 22 shows the possible realization of the W-CDMA uplink traffic channel, which is defined and described in the CDG cdma2000 proposal submitted to the ITU in June 1998, and has been cited above.

The presence of pilot data in the uplink makes it possible to estimate the array response vector of the training sequence. By making assumptions on the period of the pilot data stream, a conventional W-CDMA receiver synchronizes its demodulator with the incoming W-CDMA signal. Each hypothesis includes: accumulating k incoming signal samples, and multiplying these samples with k copies of the internally generated pilot samples (ie, the inner product of the incoming signal and the generated copy signal). The pilot replica sequence of each subsequent hypothesis is delayed and the correlation process is repeated. When the copy of the pilot is synchronized with the incoming signal, the resulting I and Q amplitudes are maximized to indicate the "locked" condition. Continuing the above accumulation process, the precise pilot can be determined, and therefore the carrier phase can be determined.

Since the pilot part of the incoming signal is always present, the integration period is only limited by the movement of the mobile unit (Doppler shift) and the imprecision of the carrier frequency used on the receiver demodulator. Since the Doppler frequency shift is less than 100 Hz at a typical mobile unit speed, and the magnitude of the frequency error is generally several hundred hertz, the integration period range is several milliseconds, which is usually much longer than the symbol duration. This mechanism is similar to the demodulation in the IS-95 downlink described above.

Using the above-mentioned pilot correlation processing, the phase estimator 2300 as shown in FIG. 23 can be used to estimate the relative electrical phase of the carrier. The incoming signal from the mobile unit is divided into two branches by the power splitter 2301 and multiplied with the signal generated from the quadrature RF signal generator 2302 to generate I and Q signals. After each signal passes through a baseband filter 2303 and an analog/digital (A/D) converter 2304 for baseband filtering and digitization, the I and Q sample streams are fed to the multiplication and accumulation (MAC) and scaling (scale ) Circuit 2305. The above-mentioned delayed pilot code from the pilot sequence generator 2306 with variable delay circuit

98811171.3 The first sequence is multiplied by the I and Q sample stream from the A/D converter 2304, summed and scaled to produce the values SUM and SUM(Q), which represent a single element in the array response vector. If the difference between the delayed pilot code sequence and the incoming signal sequence exceeds one chip duration, the values of SUM and SUM(Q) are small (using the autocorrelation function of the pilot sequence). Therefore, by changing the delay value of the pilot sequence generator 2306 within the expected range of the incoming signal TOA, the phase estimator can be used as a channel impulse response (CIR) estimator.

CIR is necessary to accurately determine signal multipath AOA and TOA values. The pilot signal present in the uplink can be used to determine the CIR. As mentioned above, the output amplitude of SUM and SUM(Q) depends on the time difference between the incoming signal and the internally generated pilot replica sequence. The traditional searcher (in the Rake receiver) changes the delay of the internally generated pilot sequence and simultaneously evaluates the square value of the sum of SUM(I) and SUM(Q) (ie, [SUM(I)+SUM(Q)) ) To measure CIR ο Figure 24 shows a system (WCDMA BeamDirectorTM) that uses the aforementioned spatial correlation to enhance the usual search processing. The signals from the antenna elements in the receiving antenna array are processed in two rows of N phase estimators (phase estimator 2300 in FIG. 23) to generate SUM(I) and SUM(Q) signal components. The spatial correlator 2400 correlates each group of N SUM(I) and SUM(Q) components with the antenna array calibration matrix to generate a correlation matrix, which represents the signals received on the antenna array and a set of predetermined directions and a set of The correlation of predetermined symbols is as described with reference to FIG. 4. The controller 2401 reads the result of the spatial correlator to generate CIR data (both amplitude and AOA data). Figure 25 shows an example of CIR data as a function of arrival time.

The controller 2401 analyzes the CIR data to determine which TOA values are used by the "House Call" section 2402 including the first-row phase estimator and the spatial correlator. The "indoor call" part 2402 is very similar to the search part 2403, and it contains phase estimators and spatial correlators of other rows. However, this "indoor call" part deals with the TOA value determined as the multipath TOA value based on the CIR data. This mechanism can obtain a high success-to-attempt rate when measuring the AOA data of the incoming multipath part.

The above-mentioned angle of arrival/time estimation can be used in both single scattering area and multiple scattering areas. By processing the histogram of the arrival angle samples, the angle range can be determined in real time. When fading is caused by a large scattering area, the angle of arrival result (AOA sample value) is distributed with a large amount of change (which can be estimated by the amount of change in the AOA result). However, the main AOA can be estimated from the histogram center of gravity. The histogram center of gravity is determined by "smoothing" the histogram through a low-pass filter (such as Hamming, raised cosine, etc.) and finding the maximum point of the "smooth" histogram. Then, by combining the "smooth" histogram peaks with

98811171.3 Comparison of the data distribution of the first histogram to estimate the size of the multipath scattering area. When there is more than one scatter area resulting in multiple "peaks" in the CIR data, a separate histogram process is performed for each significant "peak" associated with the TOA value in the CIR.

Then, the estimated AOA value together with the scattering area size (sectorization angle) is used to determine the coefficient of the uplink beamformer row 2404, which is fed to the uplink rake receiver. As the number of "fingers" of the Rake receiver is limited, the uplink beam assignment is optimized to maximize the efficiency of the Rake combination. For example, if only a single scattering area is recognized, configure all beams to uniformly cover the recognized scattering area. If multiple scattering areas are distinguished, these beams are allocated to first ensure that all the different scattering areas are covered, and then the remaining available beams are added to provide diversity in the stronger scattering area. The CIR data from the controller can also be used to determine the coefficients of the downlink beamformer 2405 sent to the transmitting antenna array based on the same downlink principle as described above. The beam width is determined according to the uplink multipath distribution, and the beam coefficient is set to ensure that the illuminance of the scattering area is the illuminance determined according to the multipath distribution.

As can be seen from the various embodiments described above, the present invention includes various changes within its scope. Those of ordinary skill in the art will understand that other modifications may be made to the above-mentioned embodiments without departing from the scope of the present invention. Therefore, the true scope of the present invention is not limited by the above-mentioned details provided for explanatory purposes. On the contrary, the true scope of the present invention is determined by the appended claims.

98811171.3