Multipath search processor for a spread spectrum multiple access communication system

ES 2 201 123 T3

DESCRIPTION

Multipath search processor for spread spectrum multiple access communication system.

Background of the invention

I. Field of the invention

The present invention relates generally to spread spectrum communication systems and, more particularly, to signal processing in a cellular telephone communication system.

II. Description of related techniques

In wireless telephone communication systems, such as cellular telephone systems, personal communication systems, and wireless local loop systems, many users communicate over a wireless channel to connect to wired telephone systems. Communication over the wireless channel can be carried out by one of several multiple access techniques that allow large numbers of users in a limited frequency spectrum. These multiple access techniques include time division multiple access (TDMA), frequency division multiple access (FDMA), and code division multiple access (CDMA). The CDMA technique has many advantages. An example of a CDMA system is described in US Pat. No. 4,901,307, published February 13, 1990, by K. Gilhousen et al., entitled "SPREAD SPECTRUM

MULTIPLE ACCESS COMMUNICATION SYSTEM USING SATELLITE OR TERRESTRIAL REPEATERS "and granted to the assignee of the present invention.

In the aforementioned patent, a multiple access technique is disclosed, in which a large number of users of a mobile telephone system, each of whom owns a transceiver, communicate through satellite repeaters or terrestrial base stations. by CDMA spread spectrum communication signals. When CDMA communications are used, the frequency spectrum can be reused multiple times, thus increasing the user capacity of the system.

The CDMA modulation techniques disclosed in US Patent No. 4,901,307 offer many advantages over narrowband modulation techniques used in communication systems using satellite or terrestrial channels. The terrestrial channel poses special problems to any communication system, in particular problems related to multipath signals. The use of CDMA techniques makes it possible to overcome the particular problems of the terrestrial channel, reducing the adverse effect of multiple paths, eg fading, and exploiting their advantages at the same time.

CDMA techniques disclosed in US Patent No. 4,901,307 include the use of coherent modulation and demodulation for both directions of the link in mobile-satellite communications. Consequently, this patent discloses the use of a pilot carrier signal as a coherent reference phase for the satellite-mobile unit link and the base station-mobile unit link. However, in the terrestrial cellular environment, the severity of multipath fading with the consequent phase interruption of the channel, as well as the high power required to transmit a pilot carrier signal from the mobile unit, makes it impossible to use demodulation techniques. coherent for the mobile unit-base station link. In the US patent No. 5,103,459 entitled "SYSTEM AND METHOD FOR GENERATING SIGNAL WAVEFORMS IN A CDMA CELLULAR TELEPHONE SYSTEM", published June 25, 1990 and granted to the assignee of the present invention, means are provided to overcome the adverse effects of multipath in the base mobile unit link, using non-coherent modulation and demodulation techniques.

In a CDMA cellular telephone system, the same frequency band can be used for communication by all base stations. At the base station receiver, separable multipaths, such as a line-of-sight path and another reflective path from a building, can undergo diversity blending to improve modem performance. The properties of the CDMA waveform that provide processing gain are also used to differentiate signals that occupy the same frequency band. Furthermore, high-speed pseudo-noise (PN) modulation allows many different propagation paths of the same signal to be differentiated, as long as the difference between the delays of the paths exceeds the duration of the PN segment. If a PN segment frequency of approximately 1 MHz is used in a CDMA system, the full spread spectrum processing gain, which is equal to the ratio of the spread bandwidth to the data rate of the system, can be used, relative to paths that have delays that differ by more than one microsecond. A path delay differential of one microsecond corresponds to a differential path distance of approximately 300 meters. The urban environment typically provides differential path delays of more than one microsecond.

Due to the multipath properties of the terrestrial channel, the receiver receives signals that have been transmitted on several differentiated propagation paths. One of the characteristics of multipath channels is the time spread introduced into the signal that is transmitted on the channel. For example, if an ideal pulse is transmitted through a multipath channel, the received signal appears as a train of

ES 2 201 123 T3 pulses. Another feature of multipath channels is that each channel path can cause a different attenuation factor. For example, if an ideal pulse is transmitted through a multipath channel, each pulse in the received pulse train generally has a different signal strength than the other received pulses. Finally, another feature of multipath channels is that each channel path can cause a different phase in the signal. For example, if an ideal pulse is transmitted through a multipath channel, each pulse in the received pulse train generally has a different phase than the other received pulses.

In the radio channel, multiple paths are created due to the reflection of the signal from surrounding obstacles (eg buildings, trees, vehicles and people). In general, the radio channel is a time-varying multipath channel, due to the relative movement of the structures that create the multipath. For example, if an ideal pulse is transmitted through the time-varying multipath channel, the received pulse train changes over location, attenuation, and phase time, depending on the time it was transmitted. the ideal boost.

The multipath characteristic of a channel can cause the signal to fade. The fading is determined by the phasing characteristics of the multipath channel. Fading occurs when the vectors of the multiple paths add destructively, receiving a signal that is less than any of the individual vectors. For example, if a sine wave is transmitted through a multipath channel that has two paths, one of which has an attenuation factor of X dB, a propagation time δ, and a phase shift of Θ radians, and the other, an attenuation factor of X dB, a propagation time δ and a phase shift of Θ + π radians, no signal will be received at the output of the channel.

In narrowband modulation systems, such as FM analog modulation systems used in conventional radiotelephony systems, the existence of multipath in the radio channel causes considerable multipath fading. However, as noted above in connection with a wideband CDMA system, the various paths can be differentiated in the demodulation procedure. This differentiation not only greatly reduces the severity of multipath fading, but provides an advantage to the CDMA system.

Diversity is one of the ways to mitigate the damaging effects of fading. Therefore, it is desirable to provide some form of diversity to allow the system to reduce fading. There are three main types of diversity: time diversity, frequency diversity, and space and trajectory diversity.

The best way to obtain time diversity is by using repetition, time interleaving, and coding to detect and correct errors that introduce redundancy. A system comprising the present invention may employ any of these techniques as a form of time diversity.

The CDMA technique, because of its inherent broadband character, offers a form of frequency diversity by spreading signal energy across a large bandwidth. Therefore, frequency selective fading only affects a small part of the bandwidth of the CDMA signal.

Space and path diversity is achieved by providing multiple signal paths over simultaneous links between a mobile unit and two or more base stations and employing two or more separate antenna elements in a single base station. Furthermore, path diversity can be achieved by exploiting the multipath environment, through spread spectrum processing, and allowing signals arriving with different propagation delays to be received and processed separately, as described above. Examples of path diversity are provided in US Pat. No. 5,101,501 entitled "SOFT HANDOFF IN A CDMA CELLULAR TELEPHONE SYSTEM", published on March 21, 1992, and in US Patent No. 5,109,390 entitled "DIVERSITY RECEIVER IN A CDMA CELLULAR TELEPHONE SYSTEM", published on 28 April 1992, both granted to the assignee of the present invention.

The detrimental effects of fading in a CDMA system can be further limited, up to a certain level, by controlling the transmit power. A system for controlling the power of the base station and the mobile unit is disclosed in US Patent No. 5,056,109 entitled "METHOD AND APPARATUS FOR CONTROLLING TRANSMISSION POWER IN A CDMA CELLULAR MOBILE TELEPHONE SYSTEM", published on October 8, 1991 and also granted to the assignee of the present invention.

The CDMA techniques disclosed in US Patent No. 4,901,307 envisage the use of relatively long PN sequences, a different PN sequence being assigned to each mobile unit user. The cross-correlation between the different PN sequences and the autocorrelation of a PN sequence, for all non-zero time shifts, have an almost zero mean value that allows differentiating the signals from the various users at the time of reception. (In order for the autocorrelation and cross-correlation to have a mean value of zero, it is necessary for the logical “0” to take the value “1” and for the logical “1” to take the value “-1” or a similar assignment.)

However, said PN signals are not orthogonal. Although the cross-correlation has basically a zero average throughout the sequence, for a short time interval, such as the time of a bit of information, the

ES 2 201 123 T3 cross correlation is a random variable with a binomial distribution. Strictly speaking, the signals interfere with each other almost as they would if they were high-bandwidth Gaussian noise, at the same power spectral density. Therefore, the other user signals, or mutual interference noise, ultimately limits the achievable capacity.

As is well known in the art, a set of n orthogonal binary sequences can be constructed, each with a length n, where n is any power of 2, see SW Golomb et al., Digital Communications with Space Applications, Prentice-Hall, Inc., 1964, pp. 45-64. Actually, orthogonal binary sequence sets can also be constructed for most lengths that are a multiple of four and less than two hundred. One class of such sequences that can be easily generated is the so-called Walsh function (or Hadamard matrix).

A Walsh function of order n can be defined recursively as follows:

W (n) =

W (n / 2), W (n / 2) W (n / 2), W '(n / 2)

<img file="ES2201123T3_D0001.tif" />

where W 'denotes the logical complement of W, and W (1) = | 0 |. Therefore,

<img file="ES2201123T3_D0002.tif" />

Each row in a Walsh function matrix is ​​a Walsh symbol, sequence, or code. A Walsh function matrix of order n contains n sequences of n Walsh segments of length each. Each Walsh code has a corresponding Walsh index that refers to the number (from 1 to n) corresponding to the row in which the Walsh code is found. For example, for the Walsh function matrix n = 8 provided above, the row made up of all zeros corresponds to the Walsh index 1 and the Walsh code 0, 0, 0, 0, 1, 1, 1, 1 corresponds to the Walsh index 5.

A Walsh function matrix of order n (as well as other orthogonal functions of length n) has the property that, over the entire n-bit interval, the cross-correlation between all the different sequences in the set is zero. As can be seen, the sequences differ from each other by exactly half their bits. It should also be noted that there always exists a sequence that contains only zeros and that the other sequences contain half ones and half zeroes. The Walsh symbol that consists of only logical zeros, instead of half ones and half zeros, is the so-called Walsh symbol zero.

In the reverse link channel between the mobile unit and the base station, no pilot signal is available to provide a reference phase. Therefore a procedure is needed to provide a link

ES 2 201 123 T3 high quality on a fading channel having a low Eb / No ratio (energy per bit / noise power density). Modulation of the Walsh function in the reverse link is a simple procedure to obtain 64-ary modulation with coherence in the set of six code symbols correlated with the 64 Walsh codes. The characteristics of the terrestrial channel determine that the rate of change of the phase is relatively slow. Consequently, if a Walsh code is selected that has a short duration compared to the rate of change of the channel phase, it will be possible to carry out coherent demodulation over the entire length of a Walsh code.

On the reverse link channel, the Walsh code is determined by the information that is transmitted from the mobile unit. For example, a three-bit information symbol can be mapped to the eight sequences of W (8) given above. At the receiver, a "decorrelation" of the Walsh encoded symbols can be performed to obtain an estimate of the original information symbols, using a fast Hadamard transform (FHT). A preferred "decorrelation" or selection procedure generates programmable decision data that can be provided to a decoder for maximum probability decoding.

To carry out the "decorrelation" procedure, an FHT is used. The FHT correlates the received sequence with each of the possible Walsh sequences. Selection circuits are employed to select the most probable correlation value, which is scaled and provided as programmable decision data.

A spread spectrum receiver of diversity or "rake" type design comprises various data receivers to mitigate the effects of fading. Typically, each data receiver is tasked with demodulating a signal that has traveled a different path, either due to the use of various antennas or due to the multipath properties of the channel. In the demodulation of modulated signals according to an orthogonal signaling system, each data receiver correlates the received signal with each of the possible correlation values, by means of an FHT. The FHT results from each data receiver are combined, and then the selection circuits select the most likely correlation value based on the highest combined FHT result to generate a demodulated programmable decision symbol.

In the system described in the aforementioned US Patent No. 5,103,459, the ringing signal starts as a 9600 bits per second source of information which is then converted by an error correcting encoder with no return channel. ratio 1/3 in an output train of 28,800 symbols per second. These symbols are grouped 6 at a time to form 4,800 Walsh symbols per second, each of which selects one of the sixty-four orthogonal Walsh functions that are sixty-four Walsh segments in duration. The Walsh segments are modulated with a user-specific PN sequence generator. User-specific PN modulated data is divided into two signals: a signal that is modulated with an in-phase (I) channel PN sequence and another signal that is modulated with a quadrature-phase (Q) channel PN sequence. Both I channel modulation and Q channel modulation provide four PN segments per Walsh segment with a PN spreading rate of 1.2288 MHz. The I and Q modulated data are combined by off-center quadrature phase shift keying (OQPSK) for transmission.

In the CDMA cellular system described in the aforementioned US Patent No. 4,901,307, each base station provides coverage to a limited geographic area and links the mobile units in its coverage area, through a switch of the cellular system, with the public switched telephone network (PSTN). When a mobile unit reaches the coverage area of ​​a new base station, that user's call routing is transferred to the new base station. The signal transmission path from the base station to the mobile unit is called the forward link and, as already noted, the signal transmission path from the mobile unit to the base station is called the reverse link.

As described above, the PN segment interval defines the minimum separation that two trajectories must have to be combined. In order to demodulate the discrete paths, the relative arrival times (or offsets) of the received signal paths must first be determined. The channel element modem performs this function by "searching" through a sequence of potential path offsets and measuring the received energy at each potential path offset. If the energy associated with a potential shift exceeds a certain threshold, a signal demodulation element can be assigned to that shift. The signal present at said path offset can then be combined with the contributions of other demodulation elements at their respective offsets. A method and apparatus for assigning demodulation elements, based on the energy levels of the seeker demodulation elements, is disclosed in US Pat. No. 5,490,165 pending, entitled "DEMODULATION ELEMENT ASSIGNMENT IN A CAPABLE SYSTEM OF RECEIVING MULTIPLE SIGNALS", published on February 6, 1996 and granted to the assignee of the present invention. Such a diversity or rake receiver provides a powerful digital link, since for the combined signal to degrade it is necessary for all paths to fade at the same time.

Figure 1 shows an example of a group of signals arriving at the base station, coming from a single mobile unit. The vertical axis represents the received power in decibels (dB). The horizontal axis represents the delay in arrival time of a signal, due to multipath delays. The axis (not shown) directed towards the page represents a time segment. Each signal peak of the common plane of the page arrives at a common time, although it has been transmitted by the mobile unit at a different time. In the common plane, the peaks on the right have been transmitted by the mobile unit before the peaks on the left. For example, peak 2 of the ridge located

ES 2 201 123 T3 at the extreme left corresponds to the last transmitted signal. Each signal peak 2 to 7 has traveled a different path and therefore shows a different propagation time and a different amplitude response. The six different signal peaks represented by peaks 2 through 7 are representative of a harsh multipath environment. Typical urban environments provide fewer usable paths. The minimum system noise threshold is represented by the peaks and troughs that have lower energy levels. The task of a finder element is to determine, in the delay measured by the horizontal axis of signal peaks 2 to 7, any potential assignment of demodulation elements. The task of demodulation elements is to demodulate a group of multipath peaks to combine them into a single output. Another task of demodulation elements, once assigned to a multipath ridge, is to track the movement of said ridge over time.

The horizontal axis can also be considered to be composed of units of PN displacements. At any given time, the base station receives, from a single mobile unit, a variety of signals that have traveled a different path and that may each have a different delay. The mobile unit signal is modulated by a PN sequence. A copy of the PN sequence is also generated at the base station, where individual demodulation of each multipath signal is carried out with a PN sequence code aligned with its timing. The coordinates of the horizontal axis are considered to correspond to the offset of the PN sequence code that would be used to demodulate a signal located at those coordinates.

As demonstrated by its irregular layout, each of the multipath ridges varies in amplitude as a function of time. In the limited time shown, there are no major changes to the multipath ridges. Over a longer period of time, multipath peaks disappear and new paths are created. The peaks can also be positioned anteriorly or posteriorly as trajectory distances change, due to the movement of the mobile unit through the coverage area of ​​the base station. Each demodulation element tracks the small variations that occur in its assigned signal. The task of the search procedure is to generate a log of the current multipath environment received by the base station.

In a conventional wireless telephone communication system, the mobile unit transmitter may employ a vocoder system that encodes voice information in a variable rate format. For example, the data transmission speed may decrease due to pauses in vocal activity. The lower data transmission speed reduces the level of interference with other users caused by the mobile unit's transmitter. A vocoder system, located in the receiver or associated with the receiver, is used to reconstruct the voice information. Apart from voice information, the mobile unit can transmit non-voice information or a mixture of both.

A suitable vocoder applicable to this environment is described in pending US patent application WO9222891, entitled "VARIABLE RATE VOCODER" and issued to the assignee of the present invention. This vocoder generates, from digital samples of voice information, data encoded at four different rates (approximately 8,000 bits per second (bit / s), 4,000 bit / s, 2000 bit / s and 1000 bit / s), based on the speech activity during a 20 millisecond (ms) frame. Vocoder data frames are formatted with extra bits as 9,600 bit / s, 4,800 bit / s, 2,400 bit / s, and 1,200 bit / s data frames. The highest transmission rate data frame corresponding to a 9600 bit / s frame is called the "full rate" frame, the 4800 bit / s data frame is called a "half rate" frame, the 2,400 bit / s data is called a "quarter speed" frame and 1,200 bit / s data frame is called a "one eighth speed" frame. No rate information is included in the data in either the encoding procedure or the frame formatting procedure. When the mobile unit transmits data at a rate less than the full rate, the duty cycle of the signal transmitted by the mobile unit is equal to the data rate. For example, at a quarter speed, a signal is transmitted from the mobile unit only a quarter of the time. During the other three-quarters of the time, no signal is transmitted from the mobile unit. The mobile unit includes a random data burst generator. At a given data rate of the signal to be transmitted, the random data burst generator determines for what periods of time the mobile unit transmits and for what periods of time it does not transmit. More information about the random data burst generator is provided in pending US Patent No. 5,535,239 entitled "DATA BURST RANDOMIZER," published September 9, 1996, and issued to the assignee of the present invention.

At the base station, each individual mobile unit signal must be distinguished from the set of received call signals to be converted by demodulation to the original mobile unit call signal. A system and method for demodulating a mobile unit signal received at a base station is described, for example, in US Patent No. 5,103,459. Figure 2 is a block diagram of the base station equipment described in US Pat. No. 5,103,459 for demodulating a reverse link mobile unit signal.

A conventional prior art base station comprises several independent search and demodulation elements. The search and demodulation elements are controlled by a controller. In this exemplary embodiment, to keep the capacity of the system high, the mobile units of the system do not transmit any pilot signals continuously. The absence of a pilot signal on the reverse link increases the time required to poll all possible time offsets with which a mobile unit signal can be received. Typically, a pilot signal is transmitted at a higher power than the signals that generate traffic,

ES 2 201 123 T3 which determines that the signal-to-noise ratio of the pilot signal is higher than that of the received traffic channel signals. Instead, in the best case, each mobile unit transmits a reverse link signal that arrives with a signal level equal to the power level received from the other mobile units and therefore has a signal-to-noise ratio. short. Also, the pilot channel transmits a known sequence of data. When no pilot signal is available, the search process must explore all the data possibilities that could have been transmitted.

For the system of Figure 2, each seeker contains an FHT processor capable of performing an FHT transform for a period of time equal to the period of a Walsh symbol. The FHT processor is subordinate to "real time", in the sense that at each Walsh symbol interval it receives and provides a value of the FHT. Therefore, to provide a quick search procedure, more than one search item should be used. The search elements perform a continuous scan to find the information signal of a particular mobile station, under control of the system controller. The search elements scan a set of time offsets around the nominal arrival of the signal to find the multipath signals that have been generated. Each of the search items sends the results of the search it performs to the controller. The controller arranges these results in tables for use in assigning demodulation elements to input signals.

Figure 2 shows an exemplary embodiment of a prior art base station. The base station of Figure 2 has one or more antennas 12 that receive CDMA signals from reverse link mobile unit 14. Typically, the coverage area of ​​an urban base station is divided into three sub-areas called sectors. With two antennas per sector, a conventional base station has a total of six receive antennas. The received signals are subjected to frequency reduction and converted into baseband signals by the analog receiver 16, which quantizes the I and Q channels of the received signals and sends these digital values ​​to the channel element modem 20, through signal lines 18. Each channel element modem 20 serves a single user. The modem contains a number of digital data receivers, or demodulation elements 22 and 24 and a number of paging receivers 26. The microprocessor 34 controls the operation of the demodulation elements 22 and 24 and the pagers 26. The PN code of the user that is set for each demodulation element and page is that of the mobile unit assigned to that channel element. The microprocessor 34 presents to the finders 26 a set of offsets, called a search window, which are likely to contain peaks of multipath signals suitable for the assignment of demodulation elements. The finder 26 indicates to the microprocessor 34 the energy it finds in each displacement. The microprocessor 34 then assigns demodulation elements 22 and 24 to the paths determined by the finder 26 (ie, the reference timing of the PN generators is moved to align with that of the found path). Once one of the demodulation elements has latched onto the signal in its assigned offset, it tracks that trajectory without the supervision of the microprocessor, until the trajectory vanishes or until the microprocessor assigns it to a better trajectory.

Although the internal structure of a demodulation element 22 only is shown in Figure 2, it should be understood that this can also be applied to the demodulation element 24 and to the finders 26. Each demodulation element 22 and 24 or finder 26 of the element modem The channel has a corresponding I and Q PN sequence generator 36 and 38 and user-specific PN sequence generator 40 that is used to select a particular mobile unit. The supplied user-specific PN sequence 40 is exclusive XORed by the exclusive XOR gates 42 and 44 with the output of the PN I and PN Q sequence generators 36 and 38 to generate the PN-I 'and PN- sequences. Q 'that are provided to the spreader 46. The reference timing of the PN 36, 38 and 40 generators is adjusted to the offset of the assigned signal, so that the spreader 46 correlates the received I-channel and Q-channel antenna samples with the PN-I 'and PN-Q sequences. 'according to the offset of the assigned signal. Four of the outputs of the spreader, corresponding to the four PN segments of each Walsh segment, are summed by accumulators 48 and 50 to obtain a single Walsh segment. The accumulated Walsh segment is then input into the fast Hadamard transform processor (FHT) 52. The FHT processor 52 maps the group of sixty-four received Walsh segments to each of the sixty-four possible transmitted Walsh functions. and provides a matrix of sixty-four programmable decision data inputs. Combiner 28 then combines the FHT processor 52 output of each demodulation element with the outputs of other demodulation elements. The output of combiner 28 is a demodulated "programmable decision" symbol. The programmable decision data consists of the chosen demodulated symbol weighted by the confidence that it correctly determines the initially transmitted Walsh symbol. The programmable decision data is then passed to the back-channelless error correction decoder 29 for further processing and recovery of the original ringing signal. This call signal is then transmitted through digital link 30 which routes the call to the public switched telephone network (PSTN) 32.

Like each demodulation element 22 and 24, each finder 26 contains a complete demodulation data path. Finder 26 differs from demodulation element 22 only in the way that its output is used and that it does not provide time tracking. Each seeker 26 searches for the correlation energy in each processed displacement, carrying out the de-spreading of the antenna samples, accumulating them in Walsh segments that are fed into the FHT processor, carrying out the FHT transform and adding the output energy Maximum FHT for each of the Walsh symbols examined by the seeker in an offset. The final sum is sent to microprocessor 34. Typically, the microprocessor 34 presents the search window in sequence to the searchers 26 in the group, which are half a PN segment apart from each other. Thus,

ES 2 201 123 T3 there will be enough correlation energy in each possible maximum quarter segment offset error to ensure that no path is missed, due to improper correlation of the seeker with a path offset. Once the microprocessor 34 has presented the search window in sequence to the search engines 26, it evaluates the received results and searches for strong trajectories for the assignment of demodulation elements, as described in pending US Patent No. 5,490,165. mentioned above.

The multipath environment is constantly changing as the mobile unit moves through the coverage area of ​​the base station. The number of searches to be carried out is determined by the need to find multiple paths quickly enough for the demodulation elements to be able to use the paths correctly. On the other hand, the number of necessary demodulation elements is a function of the number of valid paths generally found at a given moment. To meet these needs, the system in Figure 2 has two finders 26 and one demodulation element 24 for each of the four demodulator ICs used, for a total of four demodulation elements, and eight finders per modem. channel elements. Each of these twelve processing elements contains a complete demodulation data path, including the FHT processor, the execution of which on an integrated circuit requires a relatively large and expensive amount of area. In addition to the four demodulator ICs, the channel element modem also has a modulator IC and a return channel error correction decoder IC for a total of 6 IC chips. A powerful and expensive microprocessor is required to direct and coordinate the demodulation elements and the finders. As shown in the modem of Figure 2, these circuits are completely independent and require the close cooperation of the microprocessor 34 for the sequential display of the correct offsets and the processing of the FHT outputs. Each Walsh symbol microprocessor 34 receives an interrupt signal to process the FHT outputs. This frequency of interrupt signals already requires the use of a high-power microprocessor.

It will be advantageous to reduce the six ICs required for a modem to a single IC requiring less microprocessor assistance, thereby reducing the direct cost of the IC and the cost of production at the modem card level, and allowing migration to a microprocessor of lower cost (or to a single high-power microprocessor that can use multiple channel element modems at the same time). Neither reducing the size of the devices in the IC manufacturing process nor putting six chips together on a single chip is sufficient. It is necessary to redesign the fundamental architecture of the demodulator for a modem truly profitable single chip. From the foregoing, the need for a signal receiving and processing apparatus that can demodulate a spread spectrum call signal, of lower cost and more efficient architecture, is evident.

The present invention relates to a single integrated search processor that can rapidly evaluate a large number of offsets that potentially contain multiple paths of the received ringing signal. For the system of Figure 2, each seeker contains an FHT processor capable of performing a Walsh symbol FHT transform. To obtain more search processing power in the system of Figure 2, additional discrete search elements should be added that have their own FHT processor. A fundamental aspect of the present invention is the decoupling between the sequences of the FHT processor and the real time, and the use of a single FHT processor with time distribution to share between the demodulation and search procedures. To take full advantage of fast FHT processing, it is necessary to provide the FHT processor with a fast data stream. The present invention includes an efficient mechanism for providing data to the FHT processor.

Summary of the invention

In accordance with the present invention, a signal demodulator for a spread spectrum communication system uses a single integrated search processor to rapidly evaluate a large number of offsets potentially containing multiple received signal paths. After completing an assigned search, the integrated search processor presents a summary of the best potential paths for the assignment of demodulation elements.

The operation of the integrated search processor is based on the demodulation of antenna samples with Walsh coding, using a fast Hadamard transform (FHT) processing engine. The FHT processing engine can run at a speed that is many times the real-time speed at which data is received. For example, in the preferred embodiment, the FHT processing engine can generate 32 Walsh symbol correlation results at the time the system receives an amount of data equivalent to one Walsh symbol.

Data must be fed at a suitably high speed to the fast FHT processing engine to take advantage of it. In the preferred embodiment, the antenna samples are spread spectrum modulated and then de-spread before passing to the FHT processing engine.

Two buffers are required to provide input to the spreader. The first buffer is used to store the antenna data samples and the second buffer is used to store the PN sequence samples. Because there are more data bits associated with antenna samples than PN sequences, it is advantageous to limit the number of antenna data samples that are stored, although this

ES 2 201 123 T3 involves expanding the number of PN sequence data to be stored. The antenna sample buffer in the preferred embodiment can store an amount of data equivalent to two Walsh symbols. In this buffer memory, writing and reading are carried out in a circular fashion. The PN sequence buffer contains an amount of data equivalent to four Walsh symbols in the preferred embodiment.

To facilitate the circular mode of operation of the antenna sample buffer, the operation of the integrated search processor is subdivided into groups of discrete searches. Each group of discrete searches is called a search rake. Each discrete search is called a rake item. Each rake item corresponds to an amount of data equivalent to a Walsh symbol and a transform operation of the FHT processing engine. In the circular buffer, each consecutive rake element of a search rake has an offset from the previous rake element equal to half a segment of the PN sequence and an offset equal to half the time. In this configuration, each rake element of a common search rake is mapped to the same PN sequence.

Search rake groups can be displayed in a search window. Search window groups can be displayed as antenna search groups. A microprocessor can determine the hunt groups of an antenna, designating some parameters. The integrated search engine then performs the indicated searches and provides the results to the microprocessor without any additional input from the microprocessor. In this way, the integrated search processor performs a plurality of searches quickly and with minimal processor interaction.

Brief description of the drawings

The characteristics, objectives and advantages of the present invention will be better understood from the detailed description provided below, illustrated by the accompanying drawings, in which equivalent reference numerals are used for equivalent elements and in which:

Figure 1 represents an extreme case example of multipath signals;

Figure 2 is a block diagram of a prior art communication network demodulation system;

Figure 3 represents an example of a CDMA telecommunications system constructed in accordance with the present invention;

Figure 4 is a block diagram of a channel element modem constructed in accordance with the present invention; Figure 5 is a block diagram of the search processor;

Figure 6 illustrates the circularity of the antenna sample buffer using a first offset;

Figure 7 illustrates the circular character of the antenna sample buffer for a second accumulation of the first offset of Figure 6;

Figure 8 illustrates the circular character of the antenna sample buffer for a second offset; Figure 9 is a graph showing how the browser processes the receiver input as a function of time; Figure 10 is a block diagram of the front of the finder;

Figure 11 is a block diagram of the seeker spreader;

Figure 12 is a block diagram of the search engine results processor;

Figure 13 is a block diagram of the browser sequence control logic;

Figure 14 is a timing diagram showing the processing sequence shown in Figure 5, with the corresponding states of certain elements of the control logic presented in Figure 13; and Figure 15 is an alternative block diagram of the search processor.

Description of the preferred embodiment

The present invention can be implemented in a wide variety of data transmission applications and, in the preferred embodiment illustrated in Figure 3, is implemented within a system 100 for the transmission of voice and data, in which a controller and system switch, also called mobile telephone exchange (MTSO) 102, performs interface and control functions to allow calls between units

ES 2 201 123 T3 mobiles 104 and base stations 106. MTSO 102 also controls the routing of calls between the public switched telephone network (PSTN) 108 and base stations 106 for transmission to mobile units 104 and vice versa.

Figure 4 illustrates the 1 10 channel element modem and other elements of the base station infrastructure operating in accordance with CDMA procedures and the data formats described in the patents listed above. A plurality of antennas 112 provide the reverse link signal 114 to analog transceiver 116. Analog transceiver 116 lowers the frequency of the link signal to a baseband signal and samples the waveform at eight times the frequency of PN segments. The analog transceiver 116 provides the digital samples to the channel element modem 110 via the motherboard RX signal from the base station 118. When assigned to an active call, the front end of the demodulator 122 and the integrated search processor 128 isolate a signal of a particular call from the plurality of call signals contained in the reverse link signal, using the PN sequences as described. in the patents listed above.

The channel element modem 110 of Figure 4 includes a single integrated search processor 128 for determining multipath signals from the received signal. The 1-channel element modem 10 contains a single time-split Hadamard fast transform (FHT) processing engine 120 that serves both the integrated search processor 128 and the front end of the demodulator 122. The FHT processing engine 120 maps the input data to each of the possible Walsh symbols. In this exemplary embodiment, there are 64 possible Walsh symbols. The FHT processing engine 120 provides an energy level corresponding to each of the 64 possible Walsh symbols. The higher the energy level, the greater the probability that the corresponding symbol is the symbol that was actually transmitted. The maximum energy detector 160 determines which is the highest output of the sixty-four outputs from each input of the FHT processing engine 120. The maximum energy and the Walsh symbol index are then passed to the integrated search processor 128 and to the segmented demodulator processor 126. The segmented demodulator processor 126 contains functions contained in prior art non-integrated demodulation elements that are not executed on the front end of the demodulator 122 and that can be shared with the same time allocation as the processing engine FHT120. The segmented demodulator processor 126 also time-aligns and combines the symbol data received at different offsets into a single stream of demodulated "programmable decision" symbols that is weighted to obtain the best performance from the error-correcting de-interleaver decoder. no return channel 130. In addition, the segmented demodulator processor 126 calculates the power level of the signal being received. From the received power level, a power control indication is created to command the mobile unit to raise or lower its transmit power. The power control indication is passed through modulator 140, which adds the indication to the signal transmitted by the base station that will be received by the mobile unit. This power control loop operates according to the procedure described in US Patent No. 5,056,109 noted above.

The programmable decision token is transmitted to error correction decoder / deinterleaver with no return channel 130, where it is deinterleaved and decoded. The channel element microprocessor 136 supervises the entire demodulation procedure and obtains the call signal recovered from the error correction decoder / deinterleaver with no return channel via the microprocessor 134 bus interface. The ringing signal is then routed, through the digital backhaul network link 121, to the MTSO102 which connects the call via the PSTN 108.

The data path of the forward link takes place in reverse to the functions indicated for the reverse link. The signal is transmitted from PSTN 108, passes MTSO 102 and arrives at digital backward network 121. Digital backward network 121 provides input to encoder / interleaver 138 through channel element microprocessor 136. After encoding and interleaving the data, the encoder / interleaver 138 passes the data to modulator 140, where it is modulated in the manner disclosed in the patents listed above. The output of the modulator is passed to the transmit adder 142, where it is added to the outputs of the other channel element modems before being subjected to frequency increase from baseband and amplification in the analog receiver-transmitter 116. At the US patent No. 5,724,385 pending, entitled "SERIAL LINKED INTERCONNECT FOR THE SUMMATION OF MULTIPLE WAVEFORMS ON A COMMON CHANNEL", published on March 3, 1998 and granted to the assignee of the present invention, an addition procedure is presented. As indicated in the aforementioned application, the transmit adder corresponding to each element 1 10 can be cascaded into a daisy chain that results in a final sum that is provided to the analog transceiver for broadcast.

Figure 5 shows the elements that the integrated search processor 128 comprises. The core of the search procedure is the FHT processing engine with time division 120 which, as indicated above, is shared by the integrated search processor 128 and demodulation front 122 (not shown in Figure 5). Except for sharing the FHT 120 processing engine and 160 maximum energy detection block, the 128 embedded search processor is self-contained, self-controlled, and independent. The FHT processing engine 120 can perform, in a manner described below, Walsh symbol transformations 32 times faster than the FHT 52 processor of Figure 2. This fast transform capability enables time-sharing operation of the 110-channel element modem.

ES 2 201 123 T3

In the preferred embodiment, the FHT 120 processing engine is constructed using a six-stage butterfly network. Such butterfly network architectures, which are well known in the art, provide an efficient mechanism to carry out an FHT, both in terms of minimizing gates and operations and in terms of the number and speed of the devices. clock cycles required to carry out the transformation.

A butterfly lattice can be used to create an inverse transform, taking into account the symmetry used to create the Walsh symbols. A Walsh function of order n can be defined recursively as follows:

W (n) =

W (n / 2),

W (n / 2),

W (n / 2)

W '(n / 2) where W' denotes the logical complement of W and W (1) = | 0 |.

In the preferred embodiment, a Walsh sequence is generated where n = 6 and thus 6-stage convolutional butterfly coding is used to correlate the sample of 64 inputs to each of the 64 Walsh functions. possible. Convolutional butterfly coding is made up of a series of 6 parallel adders.

To reap the benefits of the FHT 120 processing engine, whose total performance is thirty-two times that of its subordinate real-time equivalent, it is necessary to provide the FHT 120 processing engine with high-speed input data for processing. The antenna sample buffer 172 has been tailored to these particular needs. In the antenna sample buffer 172, writing and reading is performed in a circular fashion.

The search procedure is grouped into single displacement search groups. The highest grouping level is the hunt group of the antennas. Each antenna search group is made up of a plurality of search windows. Typically, each search window of the antenna search group is a group of identically performed searches, where each search window receives data from a different antenna. Each search window is made up of a series of search rakes. A search rake is a group of consecutive search movements that take place in a time equal to the duration of a Walsh symbol. Each search rake consists of a set of rake items. Each rake item represents a single search at a given offset.

At the beginning of the search procedure, the channel element microprocessor 136 sends parameters indicating a search window that may be part of an antenna search set. The width of the search window can be indicated by PN segments. The number of search rakes required to perform the search window varies with the number of PN segments indicated in the search window. The number of rake elements in each search rake can be indicated by the channel element microprocessor 136 or can be set to a certain constant.

Referring again to Figure 1, which shows an example of a group of signals arriving at the base station from a single mobile unit, the relationship between the search window, the search rake, and the search window can be seen more clearly. rake element. The vertical axis of Figure 1 represents the received power in decibels (dB). The horizontal axis represents the delay in arrival time of a signal, caused by multipath delays. The axis (not shown) directed towards the page represents a time segment. The signal peaks of the common plane of the page have arrived at the same time but have been transmitted by the mobile unit at different times.

The horizontal axis can be considered to be composed of PN segment displacement units. At any given time, the base station perceives a variety of signals from a single mobile unit, which have traveled a different path and each may have a different delay. The mobile unit signal is modulated by a PN sequence. A copy of the PN sequence is also generated at the base station. For individual demodulation of multipath signals at the base station, a PN sequence code aligned with each signal timing is needed. Each of said aligned PN sequences lags behind the base station's zero reference offset due to the delay. The number of PN delay segments of the PN sequence aligned to the base station zero reference offset can be projected onto the horizontal axis.

In Figure 1, time slot 10 represents a group of PN segment offsets for processing from a search window. Time segment 10 is divided into five different search rakes, such as time segment 9 of the search rake. Each search rake, in turn, is made up of several rake elements that represent the specific offsets at which the search is performed.

ES 2 201 123 T3

For example, in Figure 1, each search rake is made up of 8 different rake elements, such as the rake element indicated by arrow 8.

To process a single rake element, as indicated by arrow 8, a group of samples taken over time at that offset is needed. For example, to process the rake element indicated by arrow 8, the de-widening procedure requires the set of samples of the displacement indicated by arrow 8, taken over time in the direction toward the inside of the page. The de-spreading procedure also requires a corresponding PN sequence. The PN sequence can be determined taking into account the arrival time of the samples and the offset to be processed. The desired offset can be combined with the arrival time to determine the corresponding PN sequence to be correlated with the received samples.

When the rake element is de-spread, the received antenna samples and the PN sequence take on a series of values ​​over time. It should be noted that the received antenna samples are the same for all offsets shown in Figure 1, and that peaks 2 to 7 show examples of multipath peaks arriving at the same time and differing only from one another by the procedure de-widening.

In the preferred embodiment described below, each rake element is offset in time from the preceding rake element by half a PN segment. This means that if the rake element corresponding to arrow 8 is correlated starting from the sectional plane shown and progressing through time (towards the page, as shown), then the rake element located at the left of the one corresponding to arrow 8 will begin to use the samples by going back in time half a segment along the sectioned plane shown. This progression in time allows each rake element of a common search rake to be correlated with the same PN sequence.

Each mobile unit receives the signal transmitted by the base station with a certain delay, due to the delay of the path through the terrestrial environment. In the mobile unit, the same generation of short and long codes is also carried out. The mobile unit generates a reference time based on the reference time it perceives from the base station. The mobile unit uses the time reference signal as input for the short and long code generators. The information signal received at the base station from the mobile unit therefore suffers from the round-trip delay of the signal path between the base station and the mobile unit. Therefore, if the timing of the PN 202, 204 and 206 generator used in the search procedure is subordinate to the base station's zero offset reference timing, the output of the generators will always be available before the corresponding one is received. signal from the mobile unit.

In an OQPSK signal, the I channel data and the Q channel data are shifted in time from each other by half a segment. Accordingly, in the OQPSK de-spreading used in the preferred embodiment, it is necessary to sample data at twice the segment frequency. The search procedure also works best when data samples are taken at half the segment frequency. Each rake element of a search rake is offset one-half segment from the previous rake element. The resolution of half-segment rake elements ensures that no multipath ridge signals are missed when performing detection. For these reasons, the antenna sample buffer 172 stores data from samples taken at a frequency equal to twice the frequency of PN segments.

In antenna sample buffer 172, an amount of data equivalent to one Walsh symbol is read to process a single rake element. Each successive rake element is obtained from antenna sample buffer 172 offset by one half PN segment relative to the previous rake element. Each rake element is de-spread with the same PN sequence obtained from PN sequence buffer 176 by the de-spread. The antenna sample buffer 172 is used for all rake elements of the search rake.

The antenna sample buffer 172 is two Walsh symbols deep and is repeatedly read and written throughout the search procedure. In each search rake, the first item that is processed is the one with the last offset in time. The last shift corresponds to the signal that has traveled the longest path from the mobile unit to the base station. The time the seeker begins to process a search rake is within the limits of the Walsh symbol associated with the rake element that has the last search rake offset. Once all the necessary samples are available in the antenna sample buffer 172, a time enable signal, called the Walsh symbol offset limit, indicates the earliest time that the search procedure can begin at the first element of the search rake.

The simplest way to illustrate the operation of the antenna sample buffer 172 is by considering its circular character. Figure 6 shows an illustrative diagram of the operation of the antenna sample buffer 172. In Figure 6, the thick circle 400 represents the antenna sample buffer 172. The antenna sample buffer 172 contains memory locations for an amount of data equivalent to two Walsh symbols. The write pointer 406 circulates around the antenna sample buffer 172 in the indicated direction, in real time, which means that the write pointer 406 rotates around

ES 2 201 123 T3 of the antenna sample buffer of two Walsh symbols of depth 172 during the time that a number of samples equivalent to two Walsh symbols is transmitted to the front of the seeker 174. As the samples are transmitted As they write to the antenna sample buffer 172 at the memory location indicated by the write pointer 406, previously stored values ​​are overwritten. In the preferred embodiment, the antenna sample buffer 172 contains 1024 antenna samples, because each of the two Walsh symbols contains 64 Walsh segments, each Walsh segment contains 4 PN segments, and each PN segment is sampled two. times.

The operation of the search procedure is divided into discrete "time slots". In the preferred embodiment, a time division is equal to 1/32 of the duration of the Walsh symbol. The 32 time divisions per Walsh symbol option is derived from the available clock frequency and the number of clock cycles required to perform an FHT. 64 clock cycles are required to perform an FHT for a Walsh symbol. In the preferred embodiment, a clock is provided that operates at eight times the PN segment rate, providing the necessary level of performance. Eight times the frequency of PN segments multiplied by the 64 clock cycles required equals the time it takes to receive an amount of data corresponding to two Walsh segments. Because there are 64 Walsh segments in each half of the buffer, it takes 32 time slots to read a complete Walsh symbol.

In Figure 6, a group of concentric arcs is shown outside the thick circle 400, representing the read and write operation to the antenna sample buffer 172. (The arcs within the thick circle 400 are used to to facilitate description, but they do not correspond to read or write operations.) Each arc represents a read or write operation during a time division. The arc closest to the center of the circle is the first in time and the consecutive arcs represent operations performed in subsequent consecutive time divisions, indicated by the time arrow 414. Each of the concentric arcs corresponds to a section of memory antenna sample buffer 172, represented by the thick circle 400. If imaginary radii are drawn from the center of the thick circle 400 to the end points of each of the concentric arcs, the part of the thick circle 400 between the intersection of the radii and the thick circle 400 will represent the memory locations to which has been accessed. For example, during operation of the first time division shown, 16 antenna samples are written to the antenna sample buffer 172 represented by arc 402A.

In Figures 6, 7 and 8, the following search parameters are adopted for the illustrative search window:

Search window width = 24 PN segments

Search offset = 24 PN segments

Number of symbols to accumulate = 2

Number of rake items per search rake = 24

Figure 6 also assumes that antenna sample buffer 172 contains an amount of valid data equivalent to nearly a full Walsh symbol prior to the writing indicated by arc 402A. During subsequent time divisions, writing is performed for arc 402B and arc 402C. During the 32 time slots available in an amount of time equivalent to one Walsh symbol, write operations for arcs 402A to 402FF are continued, most of which are not shown.

The 32 time divisions represented by arcs 402A to 402FF correspond to the time taken to perform a search rake. Using the parameters provided above, the search rake starts with a 24 PN segment offset from zero or "real time" reference offset and contains 24 rake elements. The 24 PN segment offset corresponds to a rotation of 16.875 degrees around the thick circle 400, from the beginning of the first writing indicated by arc 402A (calculated by dividing the 24 PN segment offset by the total number of 256 segments in half antenna sample buffer 172 and multiplying by 180 degrees). The 16.875 degree arc is illustrated by the 412 arc. The 24 rake elements correspond to the readings indicated by arcs 404A through 404X, most of which are not shown. In scrolling, the first read for arc 404A starts some time after the write for arc 402C, and thereby a contiguous data set is obtained. Each successive read, such as 404B, is offset from the previous one by a single memory location, corresponding to 1/2 PN time segment. During the displayed search rake, the readings are shifted to previous time offsets (indicated by arcs 404A to 404X) over time, counterclockwise and in a direction of rotation opposite to that indicated by write pointer 406. The 24 reads represented by arcs 404A through 404X traverse the arc indicated by 418. Advancing reads to previous samples has the advantage of providing a homogeneous search within a search window when the search rakes are run. This advantage will be described in more detail later.

Each of the reads corresponding to arcs 404A through 404X passes to the spreader 178 an amount of data equivalent to one Walsh symbol. Therefore, the reading is equivalent to traversing 180 degrees of the thick circle 400. It should be noted that, in the search rake shown in Figure 6, the last writing corresponding to the

ES 2 201 123 T3 arc 402FF and the last reading for arc 404X do not include any common memory locations to ensure contiguity of valid data. However, hypothetically, if you continue with this pattern of reads and writes, they will reach a point of intersection and no valid data will be obtained.

Under most signaling conditions, the result of an amount of data equivalent to one rake element collected over an amount of time equivalent to one Walsh symbol is not sufficient to provide accurate information about the location of the various signals. In these cases, the search rake can be repeated several times. The results of the rake elements of successive search rakes in a common offset are accumulated by the search result processor 162, as explained in detail below. In this case, the search parameters provided above indicate that the number of symbols to accumulate at each shift is two. Figure 7 shows the search rake of Figure 6 repeated at the same offset for the next amount of data equivalent to one Walsh symbol. It should be noted that the antenna sample buffer 172 contains an amount of data equivalent to two Walsh symbols, so that the data that is needed for processing during the search rake indicated in Figure 7 has been written during the rake. search shown in Figure 6. In this configuration, memory locations 180 degrees apart from each other represent the same PN offset.

After completion of the two accumulated search rakes of Figures 6 and 7, the search procedure proceeds to the next scroll of the search window. The advanced distance equals the width of the processed search rake, which in this case is 12 PN segments. As indicated in the search parameters, the width of the search window is 24 PN segments. The width of the window determines how many search rake offsets are required to carry out the search window. In this case, two different offsets are needed to span the width of the 24 PN segment window. In Figure 8, the width of the window is indicated by arc 412. The second offset of this search window begins at the offset that follows the last offset of the previous search rake and continues around the nominal zero offset point determined by the first write start location indicated by arc 430A. Again, the search rake is comprised of 24 rake elements indicated by arcs 432A through 432X, most of which are not shown. Also, the 32 writes are indicated by the arcs 430A to 430FF. Therefore, the last write, indicated by arc 430FF, and the last read, indicated by arc 432X, abut each other in antenna sample buffer 172, as indicated by reference arrow 414.

The search rake shown in Figure 8 is repeated on the opposite side of the antenna sample buffer 172, in the same way that the search rake of Figure 6 is repeated in Figure 7, because the parameters of search indicate that each symbol is accumulated twice. After the second accumulation of the second search rake has completed, the integrated search processor 128 is ready to begin another search window. The subsequent search window may have a different offset, it may indicate a new antenna, or both.

In Figure 8, the boundary between the read half and the write half of the buffer is marked with the label 436. In Figure 6, the boundary is marked with the label 410. The signal indicating the point at the The time corresponding to the tag 436 is called the Walsh symbol shift enable signal and also indicates that there is another number of samples available equivalent to one Walsh symbol. As the search rakes in a window advance to previous offsets, the boundary between the read and write halves of the buffer rotates step by step counterclockwise, as shown in Figure 8. If after completion of the present search window a large change in the offset being processed is desired, the Walsh symbol offset enable signal can be placed much further along the circumference of the circle.

Figure 9 is a search event line further illustrating the browser processing in a graphical manner. Time is represented along the horizontal axis in units of Walsh symbols. The address of the antenna sample buffer 172 and the addresses of the PN sequence buffer 176 are shown along the vertical axis, also in units of Walsh symbols. Because the antenna sample buffer 172 has a depth of two Walsh symbols, the addresses of the antenna sample buffer 172 overflow at the boundaries of the even Walsh symbols, although for illustrative purposes, the addresses in Figure 9 are shown before stacking on top of each other. The samples are written to the antenna sample buffer 172 in an address derived directly from the time they were taken; therefore, the write pointer 181 in the antenna sample buffer 172 is a straight line with a 45 degree tilt. The offset being processed is mapped to a base address of the antenna sample buffer 174 address to initiate a read of the samples corresponding to a Walsh symbol for a single rake element. The rake elements are illustrated in Figure 9 as nearly vertical read pointer line segments 192. Each rake element maps to a Walsh symbol in height relative to the vertical axis.

The vertical gaps between the rake elements of a search rake are due to the interruption of the search procedure by the demodulation front 122 to use the processing engine FHT 120. The demodulation front 122 operates in real time and has the first priority of use of the engine

ES 2 201 123 T3 FHT 120 processing, as long as you have a current or queued data set to process. Accordingly, the use of the FHT processing engine 120 is typically ceded to demodulation front end 120 at each Walsh symbol boundary corresponding to the PN offset that is demodulated by demodulation front end 122.

Figure 9 shows the same search rakes shown in the Figures. 6, 7, and 8. For example, search rake 194 has 24 rake elements, each corresponding to one of the read arcs 404A through 404X of Figure 6. In Figure 9, for the search rake search 194, pointer 410 indicates that the Walsh symbol shift enable signal corresponds to the equivalent pointer of Figure 6. To read current samples, each rake item must be below write pointer 181. The downward slant of rake items on a search rake indicates progress toward previous samples. Search rake 195 corresponds to search rake shown in Figure 7 and search rake 196 corresponds to search rake shown in Figure 8.

In the search window delimited by the above parameters, only 24 rake items are indicated for each search rake, even though the search rake has 32 available time slots. Each rake item can be processed in a time slice. However, in practice it is not possible to increase the number of rake items per search rake to 32 to match the number of time slots available during a search rake. The demodulation front end 122 uses some of the available time slots from the FHT processor, such as the four time slots used to process the signals in box 178 of Figure 5. Also, there is a time delay associated with the advance of the rake, since the read procedure must wait for the write procedure to fill the buffer with valid data in the advanced movement. In addition, a certain margin is needed for synchronization with a time division processing limit after observing the Walsh symbol shift enable signal. All of these factors effectively limit the number of rake items that can be processed in a single search rake. In certain cases, the number of rake elements per search rake can be increased (for example, when demodulation front end 122 has only one demodulation element assigned to it and only interrupts the processing engine FHT 120 once for each search rake ). Therefore, in the preferred embodiment, the number of rake elements per search rake can be controlled by the channel element microprocessor 136. In alternative embodiments, the number of rake elements per search rake can be a fixed constant.

Also, significant extra delay can occur when switching between source antennas at the input of the sample buffer or when changing the start point or width of the search window between searches. If a rake requires a particular group of samples and the next rake from a different antenna needs to use an overlapping portion of the buffer, the next rake should postpone processing until the next Walsh symbol offset limit appears, at which point which will have the number of samples corresponding to a complete Walsh symbol for the new source antenna. In Figure 9, it is shown how search rake 198 processes data from a different antenna than search rake 197. Horizontal line 188 indicates the memory location corresponding to the input samples from the new antenna. It should be noted that search rakes 197 and 198 do not use any common memory locations.

For each time division, a number of samples corresponding to two Walsh segments must be written into the sample buffer and a number of samples corresponding to one complete Walsh symbol can be read from the sample buffer. In the preferred embodiment, each time division is made up of 64 clock cycles. A sample quantity corresponding to a complete Walsh symbol comprises four sample groups: spot I channel samples, delayed I channel samples, spot Q channel samples, and delayed Q channel samples. In the preferred embodiment, each sample is four bits. Therefore, sixty-four bits per clock of antenna sample buffer 172 are required. If a single port RAM is used, the simplest buffer design doubles the width of the word to 128 bits, and splits the buffer into two 64-bit wide odd and even Walsh segment buffers, 64 words and independent read and write 168 and 170. Buffer writes, which are much less frequent, are multiplexed between reads that alternate between the two banks on consecutive clock cycles.

The samples corresponding to a Walsh segment read from the odd and even Walsh segment buffers 168 and 170 have an arbitrary alignment with the physical alignment of the RAM words. Therefore, at the first reading of a time division, both halves are fed into the spreader 178 to generate a two-segment wide Walsh window from which the single Walsh segment with offset alignment is obtained. current. For even Walsh segment search offsets, the odd and even Walsh segment buffer addresses for the first read are the same. For odd Walsh segment offsets, the even address for the first read is advanced one position from the odd address to provide a consecutive Walsh segment starting from the odd half of the sample buffer. Additional Walsh segments needed by the despread 178 can be transmitted to it by reading a single Walsh segment buffer. With successive reads, it is ensured that there is always a renewed width two Walsh segment window to obtain an amount of data corresponding to a offset-aligned Walsh segment that is currently being processed.

Referring again to Figure 5, for each rake element of the processed search rake, the

ES 2 201 123 T3 same amount of PN sequence data corresponding to a Walsh symbol in PN sequence buffer 176 in the de-spreading procedure. For each clock cycle of a time division, four pairs of PN-I 'and PN-Q' are needed. If a single port RAM is used, the width of the word is doubled and it is read at half the frequency. Then, the only writing to the PN 176 sequence buffer required in each time slot takes place in a cycle not used for reading.

Because the search procedure can indicate search PN offsets of up to two Walsh symbols behind the current time, it is necessary to store an amount of PN sequence data equivalent to four Walsh symbols. In the preferred embodiment, the PN sequence buffer 176 is one hundred and twenty-eight words by sixteen bits RAM. Four Walsh symbols are required, because the initial offset can vary by 2 Walsh symbols, and once the initial offset is chosen, a number of PN sequences equivalent to one Walsh symbol are needed for correlation, which means that an amount of data equivalent to three Walsh symbols for the de-spreading procedure. Because the same PN sequence is used repeatedly, the PN176 sequence buffer data cannot be overwritten during the de-spreading procedure for a single search rake. Consequently, an amount of memory equivalent to an additional Walsh symbol is required to store the generated PN sequence data.

The data that is written to PN sequence buffer 176 and antenna sample buffer 172 is provided by the face of the finder 174. In Figure 10, a block diagram of the face of the finder 174 is shown. The front of the browser 174 includes PNI and Q short code generators 202 and 206 and the PN user long code generator 204. The values ​​provided by short code PN I and Q generators 202 and 206 and user long code PN generator 204 are determined by the time of day. Each base station has a universal timing standard, such as GPS timing, to create a timing signal. Base stations also transmit their timing signals to mobile units, over the air. The base station reference timing is assigned zero offset, because it is aligned with the universal reference.

The output of user PN long code generator 204 is XOR exclusive logic function with the output of PN I and Q short code generators 202 and 206 by exclusive XOR gates 208 and 210, respectively. (This process is carried out in the mobile unit as well, and the output is used to modulate the signal transmitted by the mobile unit.) The output from the exclusive XOR gates 208 and 210 is stored in the serial / parallel shift register 212 . The serial / parallel shift register 212 temporarily stores the sequences until it fills the full width of the PN 176 sequence buffer. The output of the serial / parallel shift register 212 is then written to the PN 176 sequence buffer at an address obtained from the zero offset reference time. Thus, the browser face 174 provides the PN sequence data to the PN sequence buffer 176.

Finder front 174 also provides antenna samples to antenna sample buffer 172. Received samples 118 are selected from one of several existing antennas via MUX 216. Received samples selected via MUX 216 are passed to latch. 218, where they undergo decimation, meaning that a quarter of the samples are selected for use in the search procedure. The received samples 118 have been taken at a frequency that is eight times the PN segment frequency by the analog transceiver 116 (of Figure 4). The search algorithm processing is designed for samples taken at half the segment frequency. Therefore, it is only necessary to pass a quarter of the received samples to the antenna sample buffer 172.

The output from latch 218 is applied to the serial / parallel shift register 214, which temporarily stores the samples until it fills the width of the antenna sample buffer 172. The samples are then written to the odd and even Walsh segment buffers 168 and 170 in addresses that have also been obtained from of the zero offset reference time. In this way, the spreader 178 can align the antenna sample data with a known offset from the PN sequence.

Referring again to Figure 5, for each clock cycle of a time division, the spreader 178 obtains an amount of antenna samples equivalent to a Walsh segment of the antenna sample buffer 172 and a corresponding group of PN sequence values ​​from PN sequence buffer 176 and transmits an I and Q channel Walsh segment to the FHT processing engine 120, via MUX 124.

Figure 11 shows a detailed block diagram of the spreader 178. The even Walsh segment latch 220 and the odd Walsh segment latch 222 retain data from the even Walsh segment buffer 168 and the even-numbered segment buffer. Walsh odd 170, respectively. The MUX bank 224 extracts the number of samples equivalent to the Walsh segment to be used from the number of samples equivalent to two Walsh segments presented by the odd and even Walsh segment latches 220 and 222. The selection logic del MUX 226 defines the boundary of the selected Walsh segment, based on the offset of the rake element being processed. A Walsh segment is transmitted to the dedicated XOR function bank of the OQPSK 228 decoder.

PN sequence values ​​in PN sequence buffer 176 are held by the

ES 2 201 123 T3 PN 234 sequences. The combinatorial shift register 232 rotates the output of the PN 234 sequence latch, based on the offset of the rake element being processed, and passes the PN sequence to the exclusive XOR function bank of the OQPSK 228 decoder, which conditionally inverts antenna samples based on PN sequence. The values ​​subjected to the exclusive XOR function are then added through the adder tree 230, which performs the addition operation on the OQPSK de-spreader and then the sum of four de-spread segment outputs to generate a Walsh segment and transmit it. to the FHT 120 processing engine.

Again referring to Figure 5, the FHT 120 processing engine obtains sixty-four Walsh segments that it has received from the spreader 178, through the MUX 124, and by 6-stage butterfly convolutional coding, correlates the sixty-four Walsh segments. four input samples with the sixty-four Walsh functions in a time division of sixty-four clock cycles. The maximum energy detection block 160 can be used to find the highest correlation energy provided by the FHT processing engine 120. The output of the maximum energy detection block 160 is passed to the search results processor 162 which is part of the built-in search engine 128.

In Figure 12, the search results processor 162 is shown in detail. The search results processor 162 also operates with time division. The control signals provided to said processor 162 are delayed by segmentation to match the delay of two time slots from the start of the Walsh segment transmission to the FHT processing engine 120 to obtain the maximum output power. As noted above, a set of search window parameters can indicate that the amount of data equivalent to a certain number of Walsh symbols has accumulated before the results of the chosen shift are processed. In the parameters used in the example of Figures 6, 7, 8 and 9, the number of symbols to be accumulated is 2. The search results processor 162 performs the addition function in conjunction with other functions.

As search results processor 162 performs sums for consecutive Walsh symbols, it must store a cumulative sum for each rake item in the search rake. These cumulative sums are stored in Walsh symbol accumulation RAM 240. The results of each rake element of each search rake are passed from maximum energy detector 160 to adder 242. The adder 242 sums the present result with the corresponding intermediate value available in the Walsh symbol accumulation RAM 240. After the last accumulation of Walsh symbols from each rake element, the intermediate result is read from the symbol accumulation RAM. Walsh 240 and is added, at adder 242, to the final energy of said rake element to generate a final search result for said rake element offset. The search results are then compared to the best results found so far, as described below.

In the pending US Patent No. 5,490,165 referred to above, entitled "DEMODULATION ELEMENT ASSIGNMENT IN A CAPABLE SYSTEM OF RECEIVING MULTIPLE SIGNALS," the preferred embodiment assigns demodulation elements based on the best search results. In the present preferred embodiment, the eight best results are stored in the best results record 250. (In other embodiments, a greater or lesser number of results may be stored.) Intermediate result register 164 stores the maximum values ​​and their corresponding category. If the energy of the current search result exceeds at least one of the energy values ​​in intermediate results record 164, the control logic of search results processor 254 rejects the eight best results in intermediate results record 164 and inserts the new result, along with the appropriate category, the PN offset, and the antenna corresponding to the rake element result. All lower category results are “downgraded” one level. In the art, there are a large number of well-known methods for providing such a sorting function, any of which can be used within the scope of the present invention.

The search results processor 162 has a local peak filter consisting basically of a comparator 244 and a pre-power latch 246. If the local peak filter is enabled, updating of the intermediate results record 164 is prevented, even though any of the energies in the search results is of the appropriate level to be included, unless the search result represents a local multipath ridge. In this way, the local peaks filter prevents intense and widely "spread out" multipaths from occupying multiple intermediate result record entries 164, leaving no room for less intense, yet distinct multipaths that may be better options for demodulation.

Running the local peak filter is straightforward. The energy value of the sum of the previous rake element is stored in the previous energy latch 246. The comparator 244 compares the sum of the present rake element with the stored value. The output of the comparator 244 indicates which of its two inputs is higher and this is retained in the control logic of the search results processor 254. If the previous sample represents a local maximum, the control logic of the search results processor 254 compares the previous power result with the data stored in the intermediate results register 164 in the manner indicated. If the local peak filter is disabled by the channel element microprocessor 136, then the comparison with the intermediate result register 164 is always enabled. If the first or last rake element at the edge of the search window has a slope, then the slope lock is set so that the edge limit value can also be considered a ridge.

ES 2 201 123 T3

Simple execution of this local peak filter is facilitated by advancing reads back to previous symbols within a search rake. As illustrated in Figures 6, 7, 8 and 9, within a search rake, each rake element advances toward the signals that arrive earlier. This advance determines that, within a search window, the last rake element of a search rake and the first rake element of the subsequent search rake have contiguous offsets. Therefore, the operation of the local peak filter does not have to change and the output of comparator 244 remains valid when moving from one search rake boundary to another.

At the end of the processing of a search window, the values ​​stored in the intermediate results register 164 are transferred to the best results register 250 which can be read by the channel element microprocessor 136. The search result processor 162 therefore absorbs much of the workload from the channel element microprocessor 136 which, in the system of Figure 2, must process each rake element result independently.

In the previous sections, attention has been focused on the data processing path of the integrated search processor 128 and it has been indicated in detail how the original antenna samples 118 are converted to a multipath summary in the output of the data log. best results 250. The following sections describe in detail how each element of the search data processing path is controlled.

The search control block 166 of Figure 5 is shown in detail in Figure 13. As noted above, the channel element microprocessor 136 indicates a group of search parameters including the group of antennas to be searched stored in the antenna selection buffer 348, the initial stored offset in search offset buffer 308, the number of rake elements for each search rake stored in rake width buffer 312, the width of the search window stored in the search width buffer 314, the number of Walsh symbols to be accumulated stored in the Walsh symbol accumulation buffer 316 and a control word stored in the buffer of control words 346.

The initial offset stored in search offset buffer 308 is indicated with eight segment resolution. The initial offset controls which samples are eliminated by decimation by the latch 218 of Figure 10 on the front of the finder 174. Since in this embodiment the antenna sample buffer 172 is two Walsh symbols wide, the largest value of the initial offset is half a PN segment and less than two full Walsh symbols.

So far, the generic configuration for conducting a search has been set out. Actually, there are several kinds of predefined searches. When a mobile unit tries to access the system for the first time, it sends a beacon signal called a "preamble" using the Walsh symbol zero. The Walsh symbol zero is the Walsh symbol that is made up entirely of logical zeros, instead of half ones and half zeros as described above. When a preamble search is performed, the seeker checks if there is any mobile unit that is sending a zero Walsh symbol beacon signal on an access channel. The result of a preamble search is the energy for the zero Walsh symbol. When performing an acquisition mode access channel search, the maximum energy detector 160 provides the zero Walsh symbol energy, regardless of the detected maximum output energy. The control word stored in control word buffer 346 includes a preamble bit that indicates when a preamble search is being performed.

As described above, the power control mechanism of the preferred embodiment measures the signal level received from each mobile unit and creates a power control indication to command the mobile unit to raise or lower its transmit power. . The power control mechanism operates in relation to a group of Walsh symbols called a power control group, during the operation of the traffic channel. (Traffic channel operation occurs after access channel operation and includes operation during an active call.) All Walsh symbols in a single power control group are transmitted using the same power control indication command. mobile unit power.

Also, as described above, in the preferred embodiment of the present invention, the signal transmitted by the mobile unit is of a variable speed during the operation of the traffic channel. The speed used by the mobile unit to transmit the data is unknown at the base station during the search procedure. During the accumulation of consecutive symbols, it is essential that the transmitter is not deactivated. Consecutive Walsh symbols in a power control group are turned on as a group, which means that the 6 Walsh symbols that comprise a power control group, in the preferred embodiment, turn on or off at the same time.

Therefore, when the search parameter indicates that a plurality of Walsh symbols have accumulated during the operation of the traffic channel, the search procedure must align each search rake to start and end within a single control group. power. The control word stored in control word buffer 346 includes a power control group alignment bit. With the power control group alignment bit set to one, indicating a traffic channel search, the pro18

ES 2 201 123 T3 paging is synchronized with the next power control group limit, instead of only synchronizing with the next Walsh symbol offset limit.

The control word stored in control word buffer 346 also includes the peak detection filter enable bit described above in connection with Figure 8.

The seeker operates in either continuous or single-stage mode, depending on the value of the continuous / single-stage bit of the control word. In the single stage mode, after the search is performed, the integrated search processor 128 returns to the idle state and waits for further instructions. In streaming mode, the integrated search processor 128 is always searching, and by the time the channel element microprocessor 136 is indicated that the results are available, the integrated search processor 128 has already started the next search.

The search control block 166 generates the timing signals used to control the search procedure carried out by the integrated search processor 128, sends the zero offset reference timing to the short code PN I and Q generators 202, and 206 and to the PN generator of long codes of the user 204, it sends the enable signal to the decimation latch 218 and the selection signal to the MUX 216 on the front of the search engine 174, provides read and write addresses for PN sequence buffers 176 and odd and even Walsh segment buffers 168 and 170, provides current offset to control the operation of the spreader 178, provides reference timing of splits intratime for processing engine FHT 120, determines whether the search procedure or demodulation procedure uses the processing engine FHT 120, controlling the input MUX of the FHT 124, it provides various segmentation delayed versions of certain internal timing enable signals to the control logic of the search results processor 254 of Figure 12 to allow summation of the search results of all a rake of shifts for various accumulations of Walsh symbols and provides, to the record of best results 250, the segmented displacement and antenna information corresponding to the accumulated energy in the record of best results 250.

In Figure 13, the system time counter 342 is subordinate to the zero offset reference time. In the preferred embodiment, as noted above, the system clock operates at eight times the PN segment rate. A Walsh symbol contains 256 PN segments, and a power control group contains 6 Walsh symbols for a total of 6 x 256 x 8 = 12,288 system clocks per power control group. Therefore, in the preferred embodiment, the system timer 342 consists of a fourteen-bit counter that counts the 12,288 system clocks. The system counter 342 is subordinate to the zero offset reference time enable signal for the base station. The input reference for the PN I and Q short code generators 202 and 206 and the PN user long code generator 204 on the front of the browser 174 of Figure 10 is obtained from the system timer 342. (The output of the User Long Code Generator PN 204 also relies on a larger system reference width that does not repeat for approximately 50 days. The largest reference width in the system is not controlled by the search procedure and acts as a preset value. Continued operation based on preset value is controlled by system timer 342.) Addresses for PN sequence buffer 176 and odd and even Walsh segment buffers 168 and 170 are obtained from timer system 342. System timer 342 is latched by latch 328 at the beginning of each time division. The output of latch 328 is selected by the address MUXs 330, 332, and 334 which provide the write addresses corresponding to the current time slot for writing these buffers at some later time within the time slot.

Offset accumulator 310 tracks the offset of the currently processed rake element. The initial offset stored in search shift buffer 308 is entered into shift accumulator 310 at the beginning of each search window. The count of the shift accumulator 310 is reduced with each rake element. At the end of each search rake to be repeated for subsequent accumulations, the number of rake elements per search rake stored in rake width buffer 312 is added back to the offset accumulator to be included again in the first scroll search rake. In this way, the search procedure performs a new sweep of the same search rake to obtain another accumulation of Walsh symbols. When the search procedure has swept the entire current search rake for the last accumulation of Walsh symbols, the offset accumulator 310 count is reduced by one by selecting the "-1" input of the repeat MUX. rake 304 generating the offset of the first rake element from the next search rake.

The output of the offset accumulator 310 always represents the offset of the current rake element being processed and is therefore used to control the data input to the spreader 178. The output of shift accumulator 310 is added by adders 336 and 338 to the intratime slot timing output of system time counter 342 to generate the address sequence of a time slot corresponding to a rake element. The output of the adders 336 and 338 is selected by the address MUXs 330 and 332 to provide the read addresses of the antenna sample buffer 172.

ES 2 201 123 T3

The output of the offset accumulator 310 is also compared by the comparator 326 with the output of the system timer 342 to generate the Walsh symbol offset enable signal indicating that the antenna sample buffer 172 has sufficient data. valid for the search procedure to begin.

Search rake counter 320 keeps track of the number of rake items remaining to be processed in the current search rake. Search rake counter 320 is provided with the width of the search window stored in search width buffer 314 at the beginning of a search window. The search rake counter 320 is incremented after the processing of the last accumulation of Walsh symbols from each search rake is completed. When the counter reaches its terminal count, all of the search window offsets will have been processed. To indicate that the end of the current search window is imminent, the output of the search rake counter 320 is summed by the adder 324 with the output of the rake width buffer 3 12. The end of search window indication marks the time that antenna sample buffer 172 can begin to fill with data samples from an alternate antenna, to prepare for the next search window without altering the content needed to the current search window.

When the channel element microprocessor 136 indicates a search window, it may also indicate that the search is performed for a plurality of antennas. In such a case, the same search window parameters are repeated using samples from a series of antennas. Such a group of search windows is called an antenna search group. If channel element microprocessor 136 indicates an antenna hunt group, the set antenna is programmed by the value stored in antenna selection buffer 348. After completion of an antenna hunt group, the antenna element microprocessor Channel 136 receives the corresponding announcement.

The rake item counter 318 contains the number of rake items remaining to be processed in the current search rake. The rake item counter 318 is incremented once for each rake item processed and receives the output from the rake width buffer 312 when the seeker is in the idle state or after a search rake is completed.

The Walsh symbol accumulation counter 322 counts the number of Walsh symbols remaining to accumulate for the current search rake. The number of Walsh symbols remaining to accumulate stored in the Walsh symbol accumulation buffer 3 16 is provided to the counter when the seeker is in the idle state or after a search rake sweep has completed in the last accumulation of Walsh symbols. Otherwise, the counter increments at the end of each search rake.

The valid input counter 302 receives an input each time the input alignment of the antenna or decimator changes. Specifically, the counter receives the minimum number of samples that the seeker needs to process a search rake, based on the output of the rake width buffer 312 (that is, a number of samples equivalent to one Walsh symbol plus a rake width). Each time an antenna sample is written to the antenna sample buffer 172, the valid entry counter 302 is incremented. When the counter reaches its terminal count, it sends an enable signal that allows the search procedure to be started. Valid entry counter 302 also provides the mechanism to hold search processing when consecutive search window offsets prevent continuous data processing.

The browser can be in an idle state, a synchronization state, or an active state. The browser sequence control 350 maintains the current state. When the channel element modem 110 is reinitialized, the integrated paging processor 128 goes into an idle state. During the idle state, all of the search control block 166 counters and accumulators obtain the associated search parameters noted above. Once the channel element microprocessor 136 commands the search procedure to begin a continuous or single-stage search by means of the control word, the integrated search processor 128 enters the synchronization state.

In the synchronized state, the seeker always waits for a Walsh symbol offset limit. If the data in antenna sample buffer 172 is still invalid or if the power control group alignment bit has been set and the Walsh symbol is not a power control group limit, then the processor built-in search engine 128 remains in the synchronization state until the appropriate conditions are met in a subsequent Walsh symbol offset limit. With a properly enabled Walsh symbol offset, the seeker can go to the active state.

The integrated search processor 128 remains in the active state until it has processed a search rake and then normally returns to the synchronization state. If the integrated search processor 128 is in single stage mode, it can go from the active state to the inactive state after the last rake element of the last accumulation of Walsh symbols of the last search rake of the search window completes. Integrated search processor 128 then waits for channel element microprocessor 136 to begin another search. Instead, if the built-in search processor 128 is in streaming mode, then it enters the new set of search parameters and returns to the sync state waiting for the scroll to scroll.

ES 2 201 123 T3 Walsh symbol of the initial offset is processed in the new search. The active state is the only state in which the antenna data samples are processed. In the idle or sync state, the pager only tracks time with the system timer 342 and continues writing to the PN sequence buffer 176 and antenna sample buffer 172 so that when the pager goes to the active state, these buffers are ready to be used.

Figure 14 is an enlarged view of the first accumulation of Walsh symbols from the second search rake of a search window, such as the search rake 196 shown in Figure 9. The third Walsh symbol shown in relation to the clock reference time of the zero displacement system is divided into thirty-two time divisions. The synchronization state of seeker 372 changes to active when the Walsh symbol offset limit indication indicates that antenna sample buffer 172 has valid samples ready to be processed at that offset. During the next available time slot, the first rake item of the search rake is processed. The finder continues to use each time slot to process a rake element (indicated by an "S" in time slots 374), unless the front end of demodulator 122 uses the FTH 120 processing engine (indicated by a " D ”in time divisions 374). The finder finishes processing all rake items in the rake and returns to the sync state before the next Walsh symbol shift boundary. Also, it is shown how the state of the search rake counter 362 is incremented during the active state until it reaches the terminal state, which means that the entire search rake has been processed. It also shows how the state of the offset counter 364 is incremented between each time slot corresponding to a rake element, so that it can be used to obtain the offset read address of the sample buffer during the time slot. The state of the offset counter 364 is delayed by segmentation to generate the offset count for the intermediate result record 164. The shift counter 368 is incremented in the last Walsh symbol accumulation pass 370.

Thus, a single integrated search engine, using antenna sample buffering and a time-split transform processor, can independently and sequentially perform a configured search using a group of search parameters, analyze the results, and present a summary of the best trajectories to use in the reassignment of demodulation elements. This reduces the workload related to the finder of the microprocessor and thus it is possible to use a less expensive microprocessor, as well as reduce direct IC costs by including a full channel element modem in a single IC.

The general principles described here can be used in systems employing alternative transmission systems. The above description is based on the reception of a reverse link signal, in which a pilot signal is not available. In the forward link of the preferred embodiment, the base station transmits a pilot signal. The pilot signal is a signal that has known data, and therefore the FHT procedure used to determine what data has been transmitted is not necessary. An integrated search processor for receiving the signal comprising a pilot signal constituting an embodiment of the present invention will contain neither the FHT processor nor the peak energy detection function. For example, the FHT processing engine 120 and the maximum energy detection blocks 160 in Figure 5 can be replaced by a simple accumulator 125 as shown in Figure 15. When a pilot signal is available, the search operation is analogous to an access channel search operation in acquisition mode as described above.