WO2011147160A1

WO2011147160A1 - Terminal and method for searching cell frequency

Info

Publication number: WO2011147160A1
Application number: PCT/CN2010/078334
Authority: WO
Inventors: 邱宁; 李强; 曾文琪; 于天昆; 刘中伟; 邢艳楠; 梁立宏; 李立文; 林峰; 褚金涛
Original assignee: 中兴通讯股份有限公司
Priority date: 2010-05-24
Filing date: 2010-11-02
Publication date: 2011-12-01
Also published as: CN102264120B; CN102264120A

Abstract

The present invention provides a terminal and method for searching cell frequency, and the method includes: for each frequency to be searched in a region of search, collecting a number of samples meeting one sub-frame, dividing the number of samples meeting one sub-frame into n signal segments, calculating the power spectral density of each signal segment, then based on the power spectrum density of each signal segment, deriving the characteristic value of the signal segment, reserving the maximum characteristic value in the n derived characteristic values as the characteristic value of the frequency to be searched, grouping the characteristic values of all of the frequencies to be searched in the whole region of search into a list of characteristic values, and considering the frequency corresponding to the maximum characteristic value in the list of characteristic values as the preferred frequency; wherein n≥1. With the technical scheme of the present invention, an ACG trying in the prior art can be avoided, so as to prevent a series of problem arising therefrom, and the cross-coupling with peripheral modules can be reduced, the miss probability and error probability can be lowered greatly, as well as the performance of the frequency search can be improved.

Description

Search terminal and method for cell frequency point

Technical field

The invention belongs to the field of communication and information technology intermediate frequency point search, and particularly relates to a search terminal and method for a cell frequency point.

Background technique

TD-SCDMA (Time Division-Synchronous Code Division Multiple Access) is one of the three mainstream standards of 3G (3rd generation) and has broad application prospects. The purpose of the frequency search by the TD-SCDMA terminal is to find the frequency used by the neighboring base stations, and then the terminal attempts to connect. Through the frequency search, the cell search can establish a normal processing flow at the selected working frequency.

The frequency allocation scheme in the standard is as follows:

Channel frequency spacing: 200 kHz, ie all carrier frequencies are integer multiples of 200 kHz

The UTRA Absolute Radio Frequency Channel Number (UARFCN) is defined as N _t = 5*F 0.0 MHz ≤ F ≤ 3276.6 MHz;

Where F is the carrier frequency, in MHz.

1.28 Mcps Time Division Duplexing (TDD) UARFCN Range: Table 1: 1.28 McPS TDD UTRA Absolute Radio Frequency Number

The existing frequency point search methods are mostly based on RSSI (Received Signal Strength Indicator) statistics, and the RSSI values are counted in the time domain in steps of 200 kHz, and the values are sorted as the order of effective frequency point reliability. Since the frequency point may be distributed at intervals of 200 kHz, it is much smaller than The signal bandwidth of 1.6MHz is only less than 0.6dB from the actual power point of 200kHz. Even at 800kHz from the actual frequency point, the signal power is only reduced by about 3dB, so RSSI based. The frequency point search method does not clearly distinguish the true frequency point. Only one possible frequency point list can be reported, and the actual frequency point cannot be guaranteed to appear in the first place. Especially in the existing mobile networking scenarios where the actual frequency points are arranged at 1.6 MHz intervals and in the external field environment, the actual frequency points may be ranked later in the search results.

Due to the excessive frequency of reporting, in order to eliminate the subcarrier frequency, the partial frequency search scheme introduces a primary and secondary frequency discrimination mechanism based on the downlink synchronization code judgment (same as the cell search coarse synchronization process). This introduces a new problem, because when the cell search coarse synchronization process is performed, the subframe synchronization has not been established, and the automatic gain control (AGC) cannot enter the synchronous mode, and is affected by the adjacent mobile station, when going up and down. There may be a huge difference in power between the gaps. In order to obtain a reasonably quantized downlink synchronization code signal and its nearby Guard Period (GP) on the digital baseband, the cell search coarse synchronization has to try a variety of possible AGC gains, and in each AGC gain scenario, Feature window search, the feature window optimal value obtained under all AGC gain scenarios is used as the estimated position of the subframe synchronization code position.

The feature window coarse synchronization method based on AGC attempts has the following problems:

1) When the AGC gain is low and the actual signal power is small, most of the data does not obtain valid quantization bits. The division with the undersized data leads to the frequent occurrence of abnormal eigenvalues, which affect the normal eigenvalue estimation.

2) Excessive AGC attempt types increase the probability of misjudging the timing position and reduce the overall performance of the cell search.

3) The interval and range of AGC attempts depend on various factors such as the RF device, the analog to digital converter (ADC) bit width, and the dynamic range of the downstream signal, increasing the cross-linking coupling between multiple module designs.

4) In order to guarantee the performance of coarse timing, especially the reliability in low speed environment, often single time

The AGC gain must go through enough subframes, and multiple AGC attempts will greatly increase the processing time of the cell search.

In summary, there is a certain search method based on RSSI or AGC to try feature window assisted RSSI. The inherent flaws, and the actual frequency is often after a large number of false frequencies. In the case where the cell search process takes a long time, such frequent false positives greatly increase the time of the search process.

Summary of the invention

The invention provides a search terminal and method for a cell frequency point, avoids a series of problems introduced by the AGC attempt in the prior art, reduces cross-linking coupling with peripheral modules, and greatly reduces false negatives and false positives. Improved frequency search performance.

A cell frequency search method includes:

For each frequency to be searched in the search interval: a sample that satisfies the number of one subframe, the sample that satisfies the number of one subframe is divided into n signal segments, and the power of each signal segment is calculated. Spectral density, then obtaining the characteristic value of the signal segment according to the power spectral density of each signal segment, and retaining the largest eigenvalue of the obtained n eigenvalues as the eigenvalue of the frequency to be searched;

And characterizing the feature values of all the frequency points to be searched in the entire search interval into a feature value list, and using the frequency point corresponding to the largest feature value in the feature value list as the preferred frequency point;

The η > 1.

The step of determining the characteristic value of the signal segment according to the power spectral density of each signal segment may include: for each signal segment, setting an average power spectral density of the signal in the passband of the signal segment to T1, and the power of the signal in the transition band The maximum value of the spectral density average is T2, and the characteristic value T=T1 - Τ2 of the signal segment is calculated.

In the step of calculating the power spectral density of each signal segment, the power spectral density exceeding 12 dB can be recorded as 12 dB.

The method may further comprise: performing hard decision processing on data of each sample of the set.

In the step of dividing the collected samples into n signal segments, each signal segment may contain less than or equal to 2048 samples.

The method may further include: if the eigenvalue of the preferred frequency point is greater than the eigenvalue threshold, performing the following steps: (a) selecting the second largest and the eigenvalue of the frequency to be searched in the range of the preferred frequency point plus or minus 1.6 MHz The frequency of the to-be-searched corresponding to the three eigenvalues is used as the candidate frequency point corresponding to the preferred frequency point; (b) the eigenvalue of the frequency to be searched in the range of the preferred frequency point plus or minus 1.6 MHz is cleared; The eigenvalues of the frequency to be searched outside the positive and negative 1.6MHz range of the selected preferred frequency point are greater than the eigenvalue. The characteristic value of the threshold is selected as the preferred frequency point corresponding to the largest eigenvalue of the eigenvalue of the frequency to be searched outside the range of plus or minus 1.6 MHz of the preferred frequency point, and returns to step (a) until this time. There is no eigenvalue greater than the eigenvalue threshold in the eigenvalues of the frequency to be searched outside the range of plus or minus 1.6 MHz of the selected preferred frequency point.

The invention also provides a search terminal of a cell frequency point, comprising a sample collection module, a segmentation module, a calculation module and a selection module; the sample collection module is set to each frequency to be searched in the search interval. Collecting samples that satisfy the number of one subframe;

The segmentation module is configured to divide the sample points satisfying the number of one subframe into n signal segments;

The calculating module is configured to calculate a power spectral density of each of the n signal segments, and then obtain a characteristic value of the signal segment according to a power spectral density of each signal segment, and retain the obtained n features a maximum eigenvalue in the value as a feature value of the frequency to be searched;

The selection module is configured to select, as a preferred frequency point, a frequency point corresponding to the maximum feature value from a list of feature values consisting of feature values of all the frequency points to be searched in the entire search interval;

The η > 1.

The calculation module may be configured to determine the characteristic value of the signal segment according to the power spectral density of each signal segment as follows: For each signal segment, the average power spectral density of the signal in the passband of the signal segment is T1, the transition band The maximum value of the average power spectral density of the internal signal is T2, and the characteristic value T=T1_T2 of the signal segment is calculated.

The calculation module can be set to record a power spectral density of more than 12 dB as 12 dB when calculating the power spectral density of each signal segment.

The terminal may further include a processing module, and the processing module may be configured to perform hard decision processing on data of each sample collected by the sample collection module.

Among the n signal segments, each signal segment may contain less than or equal to 2048 samples.

Further, the selecting module may be further configured to: when the feature value of the selected preferred frequency point is greater than the feature value threshold, perform the following operations:

(a) the second of the feature values of the frequency to be searched in the range of the preferred frequency point plus or minus 1.6 MHz The frequency to be searched corresponding to the large and third largest eigenvalues is used as the candidate frequency point corresponding to the preferred frequency point;

(b) clearing the eigenvalue of the frequency of the to-be-searched frequency in the range of plus or minus 1.6 MHz of the preferred frequency point; if the eigenvalue of the frequency to be searched outside the range of plus or minus 1.6 MHz of the selected preferred frequency point is selected If there is an eigenvalue greater than the eigenvalue threshold, the frequency of the eigenvalue corresponding to the largest eigenvalue of the frequency to be searched outside the range of plus or minus 1.6 MHz of the preferred frequency point is selected as the preferred frequency point, and is repeated. In step (a), there is no feature value greater than the feature value threshold in the feature values of the frequency to be searched outside the range of plus or minus 1.6 MHz of the selected preferred frequency point.

In summary, the present invention provides a search terminal and method for cell frequency points. Starting from constructing a more reliable estimation factor, a frequency domain frequency point search scheme based on hard decision is proposed, and the estimation method does not need to involve radio frequency. Many factors such as the device, ADC bit width, and dynamic range of the downstream signal. Theoretical analysis and simulation show that the frequency search process constructed based on the present invention has excellent performance and can work stably in various scenarios. Compared with the prior art, with the technical solution of the present invention, the actual frequency points can be more accurately listed in the search results, and the prior art needs to greatly expand the range of candidate frequency points to reliably list the actual frequency points. problem. The processing time and the amount of calculation for eliminating the false frequency point in the subsequent step cell search are reduced.

BRIEF abstract

Figure 1 is the amplitude-frequency response of the received signal;

Figure 2 is a flow chart of the method of the present invention;

Figure 3 is a single-slot single-slot scene frequency search performance on an AWGN channel;

Figure 4: Single-frequency point two-slot scene frequency point search performance under AWGN channel;

Figure 5 Single-frequency point full-slot scene frequency search performance under AWGN channel;

Figure 6: Frequency performance of two isolated frequency points in the AWGN channel;

Figure 7: Frequency search performance of two adjacent frequency points in the AWGN channel;

Figure 8: Frequency search performance of three adjacent frequency points in the AWGN channel;

Figure 9: Single-frequency point scene frequency search performance in the easel channel;

Figure 10 Single-frequency point scene frequency search performance under the case2 channel; Figure 11 Single-frequency point scene frequency search performance under the case3 channel;

Figure 12 Single-frequency point scene frequency search performance under the case3 channel;

Figure 13 Single-frequency point scene frequency search performance under the case3 channel.

Preferred embodiment of the invention

The present invention provides a search terminal and method for a cell frequency point. For each frequency point to be searched in a search interval, a sample that satisfies the number of one subframe is collected, and the sample that satisfies the number of one subframe is sampled. Dividing into n signal segments, calculating the power spectral density of each signal segment, then obtaining the characteristic value of the signal segment according to the power spectral density of each signal segment, and retaining the largest feature among the eigenvalues of the n signal segments The value is used as the feature value of the frequency to be searched, and the feature values of all the frequency points to be searched in the entire search interval are grouped into a feature value list, and the frequency point corresponding to the largest feature value in the feature value list is used as the preferred frequency point; η > 1.

First, we will pay attention to the frequency domain characteristics in the bandwidth. Figure 1 shows the amplitude of the received signal in the case of the Additive White Gaussian Noise (AWGN) channel when the soft and hard decisions are used in the range of 0 to 800 kHz. response. In order to ensure the reception performance of high-order modulation, the bandwidth of the RF anti-aliasing filter is usually greater than 800 kHz. The characteristics here are basically the power spectrum of the raised cosine of the transmitting end in addition to the frequency selectivity of the channel. The waveform of the power spectral density at 800 kHz to 0 is symmetrical with the waveform in Fig. 1.

Two conclusions can be drawn from Figure 1:

(1) The root-raised cosine power spectrum (ie, power spectral density) begins to decay at 480 kHz, with a 3 dB bandwidth at 640 kHz, and a large attenuation range from 640 kHz to 800 kHz, but after oscillating 200 kHz, the in-band attenuation is 440 kHz to 600 kHz. Therefore, the relationship between the average power spectral density in the band of -640kHz to 640kHz (ie, in the passband) and the average power spectral density of -800kHz to 640kHz and 640kHz to 800kHz in-band (ie, in the transition band) will be a performance. Good criteria.

(2) The distortion caused by the hard decision to the signal power spectral density in the 800 kHz range is not obvious, and a hard decision-based method scheme can be used.

The difference between soft decision and hard decision: The baseband signal of the RF output is itself an analog signal, which is converted into a multi-bit quantized digital signal by A/D (analog/digital) conversion. The multi-bit quantization here is soft. Judgment. The meaning is that in addition to the symbol, there is information about the magnitude of the data. For example, 107.152 is quantized and becomes 107, -13.87 is quantized and becomes -14; hard decision means that only the sign bit of the data is taken as the output, except for the symbol, there is no other information, for example, 107.152 is quantized and becomes 1 - 13.87 is quantized and becomes -1.

The embodiment provides a search terminal of a cell frequency point, which includes a sample collection module, a segmentation module, a calculation module, a processing module, and a selection module. The sample collection module is set to each search frequency in the search interval. Point sets a sample that satisfies the number of one subframe;

The segmentation module is configured to divide the sample that satisfies the number of one subframe into n signal segments; the calculation module is configured to calculate a power spectral density of each signal segment, and then determine the power spectral density according to each signal segment. a feature value of the signal segment, and retaining a maximum feature value of the feature values of the n signal segments;

The selection module is configured to select a frequency point corresponding to the maximum eigenvalue as a preferred frequency point from a list of feature values consisting of eigenvalues of all the frequency points to be searched in the entire search interval;

n > l„

The calculation module is configured to determine the characteristic value of the signal segment according to the power spectral density of each signal segment as follows: For each segment of the signal, the average power spectral density of the signal in the passband is T1, and the power of the signal in the transition band The maximum value of the spectral density average is T2, and the characteristic value T=T1 _ Τ2 of the signal segment is calculated.

Further, when the calculation module calculates the power spectral density of each signal segment, the power spectral density exceeding 12 dB is recorded as 12 dB.

The processing module is arranged to perform hard decision processing on the data of each sample of the sample collection module.

Further, when the segmentation module divides the sample that satisfies the number of one subframe into n segments, the number of samples included in each segment of the signal is less than or equal to 2048.

The selection module is further configured to: when the selected feature value of the preferred frequency point is greater than the feature value threshold, select an alternate frequency point corresponding to the preferred frequency point; and when the last selected preferred frequency point is within a range of plus or minus 1.6 MHz When there is a feature value greater than the feature value threshold in the feature value of the frequency to be searched, the first time is selected. The frequency of the to-be-searched corresponding to the largest eigenvalue of the to-be-searched frequency point outside the range of 1.6 MHz of the frequency selection point is used as the preferred frequency point, and the candidate frequency corresponding to the selected preferred frequency point is selected. The selection module is configured to select an alternative frequency point corresponding to the preferred frequency point by: selecting the second largest and third largest eigenvalues of the eigenvalues of the frequency to be searched within the range of the preferred frequency point plus or minus 1.6 MHz. The corresponding frequency to be searched is used as an alternative frequency point corresponding to the preferred frequency point;

The selection module is further configured to clear the eigenvalues of the frequency points to be searched within the range of the preferred frequency points of the selected frequency points in the range of 1.6 MHz after selecting the candidate frequency points.

The present embodiment provides a method for searching for a cell frequency point. For each frequency point to be searched in the search interval, a sample that satisfies the number of one subframe is collected, and the sample that satisfies the number of one subframe is divided. For the n-segment, calculate the power spectral density of each segment of the signal, and then obtain the eigenvalue of the segment signal according to the power spectral density of each segment of the signal, and retain the maximum eigenvalue of the eigenvalue of the n-segment signal as the frequency of the to-be-searched frequency The feature value is used to form a feature value list of the feature values of all the frequency points in the entire search interval, and the frequency to be searched corresponding to the largest feature value in the feature value list is used as the preferred frequency point.

The specific implementation steps of the present invention are as shown in FIG. 2, and include the following steps:

Step S1: selecting a frequency to be searched at intervals of 200 kHz in a specified search interval. For each frequency to be searched, the set of samples may form a sample of one subframe, that is, a set of 6400 x 4 samples;

Further, the data of each sample of the set is subjected to hard decision processing, as shown in formula (1);

Signal =complex((real(SigT)>=0)*2-l,(imag(SigT)>=0)*2-l) (1) In equation (1), Signal is the input signal of the sample point. Real is the real part of the signal, imag is the imaginary part of the signal. The meaning of the whole formula is that the input complex number retains only the sign bits of the real part and the imaginary part as output, and is reassigned to the signal, for example, 107.152-13.87j After the calculation, it is +lj.

Further, the LNA (Low Noise Amplifier) is always turned on during the search in this step. The LNA functions to perform low noise amplification on the RF signal received by the antenna, and fixedly configure the VGA (Video Graphics Array) to the maximum. Gain, select the sign bit of the ADC input data as the input data of the frequency point. The sampling rate is 4 times the chip rate and bypass the digital baseband's root raised cosine filter.

Step S2: For the sequence of samples of each sub-frame, the sequence of the tail is complemented to satisfy an integer multiple of 2 ¹ , preferably 9; this embodiment is described by taking i = 9 as an example; Further, it is also possible to normalize each sample point, that is, formula (2)

Signal = [Signal;Signal(l:256*4)]/sqrt(2) (2) Step S3: Segment the virtual sub-frame in units of 2 ¹ chips (in this embodiment, take 512 chips as an example), Time slot 0 - a signal is present, this step ensures that at least one segment of the signal characteristic condition is satisfied after segmentation.

Segmentation of the sample points within the sub-frame can reduce the interference of the unsignaled time slot, that is, the power spectral density formed by the segment that ensures the signal is more realistic.

The length of each sub-frame is 6400 chips. Since the start position of the sub-frame is not determined here, it is arbitrarily assumed here that a chip takes data for the start position of the sub-frame, and each 6400 chips is defined as one. Virtual sub-frame.

Signal = reshape(Signal, 512*4, []) (3) The formula (3) consists in segmenting the virtual sub-frame data in units of 512 chips.

Step S4: transform each segment of data into a frequency domain, and obtain a spectrum of each segment of the signal, that is, perform FFT transformation on each segment of the signal signal, and normalize the transformed result;

Signal = fft(Signal)/sqrt(512*4) (4) In equation (4), fft(Signal) is used to calculate the spectrum of each segment of the signal, and sqrt(512*4) is to normalize the calculated spectrum.

Step S5: further obtaining a power spectral density for each segment of the signal;

Further, when calculating the power spectral density, the power word exceeding the average power spectral density of 12 dB is also limited, and the limiting method is to retain the original value for the power spectral density not exceeding 12 dB, and for the power spectral density exceeding 12 dB. 12dB, this step reduces the effect of mono or narrowband interference on the frequency search.

Signal = min(abs(Signal). ^A 2, 16) ( 5 ) Equation (5) is to calculate the power spectral density, abs is to calculate the absolute value, ^Λ 2 is the square operation, min is the limit operation, that is, over The 12dB power density is limited to 12dB (equivalent to 16x).

Step S6: Calculate a feature value of the segment for each segment in a virtual sub-frame;

Let the average power density of each band in the passband (ie -640kHz~640kHz) be T1, and the average of the power spectral density of the signal in the transition band (-800kHz-640kHz and 640kHz~800kHz) The value is T2, and the eigenvalue of the segment is Τ=Τ1 - Τ2; as shown in equation (6):

Deffisti=sum([Signal(end-255:end,:);Signal(l:256,:)])/512- max(sum(Signal(end-319:end-256,:)),sum(Signal (257:320,:)))/64 (6) Step S7: The largest eigenvalue in the virtual sub-frame is reserved.

Deffisti = max(Deffisti) (7) Step S8: After completing the above eigenvalue calculation operation for all the frequency points to be searched, a list of feature values of the frequency points to be searched is obtained, and the feature value list has a total of m eigenvalues ( Each search frequency point corresponds to one feature value). After obtaining the list, the effective frequency point is determined by the following rules;

The invention divides the frequency search result into two sets of preferred frequency points and alternative frequency points.

Selecting a maximum eigenvalue from the eigenvalue list, using the frequency point corresponding to the largest eigenvalue as the preferred frequency point, and then performing steps S801 to S802;

Step S801: Searching for the second largest value and the third largest value of the feature value in the feature value of the frequency to be searched in the range of plus or minus 1.6 MHz of the preferred frequency point, and using the frequency point corresponding to the found feature value as The preferred frequency point corresponding to the preferred frequency point;

Step S802: Clearing the eigenvalue of the frequency of the frequency to be searched in the range of the preferred frequency point to the range of 1.6 MHz, if the selected frequency point of the selected frequency point is within the range of plus or minus 1.6 MHz, the eigenvalue of the frequency to be searched is further greater than the eigenvalue. The characteristic value of the threshold DctThd is selected as the preferred frequency point corresponding to the largest eigenvalue of the eigenvalues of the frequency to be searched outside the range of the preferred frequency point and the 1.6MHz range, and returns to step S801 to continue selecting and selecting the preferred frequency point. The alternate frequency point corresponding to the frequency point. There is no characteristic value greater than the eigenvalue threshold DctThd in the eigenvalues of the frequency to be searched until the selected preferred frequency point is within the range of 1.6 MHz. The eigenvalue threshold DctThd can be set according to actual needs, and is taken as 1 in this embodiment.

Further, the selected preferred frequency points may be sorted according to the order in which the preferred frequency points are selected, that is, the size of the feature values on each frequency to be searched is used as the frequency reliability basis.

The following is a simulation comparison of the performance of the frequency point search method in various scenarios with the typical parameter eigenvalue threshold DctThd being 1. The search range is set to 2010MHz~2025MHz (10054-10121), a total of 68 frequency points to be searched.

First, the performance of the hard-decision frequency domain frequency search based on the AWGN channel is analyzed. Figure 3 is a single-frequency (2014MHz, 10070) environment with only the performance of the TS0 signaled scene. In the picture The coordinate is the signal-to-noise ratio of the signal slot, and the ordinate is the probability that the frequency search result is wrong. The missing report is defined as the actual frequency missed in the reported frequency point; the full leak is defined as all the actual frequency points are not reported. The result appears; false positives are defined as the frequency points in the reported frequency points that do not appear in the actual frequency points. When not specified, the above statistics only count the preferred frequency points for reporting, excluding the alternative frequency points. The number of simulations per sample is 1000 global frequency points search, and the starting position of each virtual sub-frame is 0~6399chi.

It can be seen that the method can obtain better performance even if there is only a single time slot in the AWGN channel. When the signal-to-noise ratio is -2dB, the false negative probability is lower than 10%, and when the signal-to-noise ratio is higher than OdB, the false negative report is reduced. To less than one percent. There is no case of false positives for other frequency points.

See Figure 4 and Figure 5 for the performance when there are two signal slots and all slots. Figure 4 shows the performance of the single-frequency two-slot scene frequency search on the AWGN channel. Figure 5 shows the AWGN channel. Single-frequency point full-slot scene frequency point search performance, assuming that the signal power of these time slots is equal. Obviously, the increase in the number of signal slots further improves the performance of the frequency search, and the working area of all time slots is reduced to below -4 dB. In order to determine the lower bound of the performance of the method of the present invention, it is assumed that the performance of the frequency search in other scenarios is followed by using the worst case in which only TS0 has a signal.

When the actual frequency point is two isolated equal frequency points, the frequency point search result of the scene is shown in Fig. 6. Compared with a single frequency point, the full leakage performance is significantly improved at two isolated frequency points. Again, there are no cases of errors.

In practice, there is a case where the center position interval of two frequency points is exactly 1.6 MHz. At this time, since the signal strengths obtained by searching according to the prior art RSSI at a plurality of positions at intervals of 200 kHz are similar, the actual frequency points are not accurately reported. The set of candidate frequency points that do not report a large number of times increases the processing time and computation amount of the subsequent cell search module to exclude false frequency points. The present invention has significant advantages over the prior art in the presence of adjacent frequency points.

When the actual frequency points are adjacent to 1.6MHz, the search performance is slightly reduced due to partial projection of the adjacent frequency point signal to a bandwidth of 640 kHz to 800 kHz. Figure 7 and Figure 8 respectively simulate the search performance of two adjacent frequency points and three adjacent frequency points. Figure 7 shows the performance of the frequency search of two adjacent frequency points in the AWGN channel. Figure 8 shows the performance of the three frequency channels in the AWGN channel. Frequency performance search performance of adjacent frequency points.

It should be added that the method based on the time domain RSSI and DwPTS in the adjacent frequency scene can hardly distinguish the actual frequency point, and there will be a large number of false positives.

Next, the search performance of the present invention under three case fading channels specified by the standard will be analyzed. Figure 9 shows the performance of the frequency search of the single-frequency scene in the easel channel. Figure 10 shows the performance of the frequency search of the single-frequency scene in the case2 channel. Figure 11 shows the performance of the frequency-frequency search of the single-frequency scene in the case3 channel. Under the less easel and case2 channels, the frequency search performance is ideal. However, in _{C ase3} strong frequency selective channels due to multipath dense preferred frequency there a lot of false positives and false negatives. This is because the channel frequency selectivity causes the spectral center of gravity position to be mis-reported as a neighboring frequency point. In addition to providing preferred frequency points, this method provides two alternative frequency points for each preferred frequency point. When the preferred frequency point cannot complete the subsequent process, the candidate frequency points should be tried one by one. Figures 12 and 13 are search performance considering the first alternative frequency point and two alternative frequency points.

Figure 12 shows the performance of single-frequency spot frequency point search with a candidate frequency point in the case3 channel, and Figure 13 shows the performance of the single-frequency point scene frequency search with two alternative frequency points in the case3 channel. It can be seen from the figure that the performance of case3 has been greatly improved after introducing an alternative frequency point. After including two candidate frequency points, the probability of false negatives is greatly reduced, and the search performance fully satisfies the needs of the system.

Due to the uplink and downlink time division multiplexing of TD-SCDMA, and the frequency interval 200 kHz is much smaller than the effective bandwidth 1.6 MHz, the frequency point search method based on the time domain RSSI and AGC attempts encounters various problems in the actual scene. The invention proposes a frequency domain frequency point scheme based on hard decision based on constructing a more reliable estimation factor. The estimation method does not need to involve many factors such as the RF device, the ADC bit width and the dynamic range of the downlink signal. Theoretical analysis and simulation show that the frequency search process constructed based on the present invention has excellent performance and can work stably in various scenarios.

Industrial applicability

Compared with the prior art, the technical solution of the present invention avoids the problem that the prior art needs to greatly expand the range of candidate frequency points to reliably list the actual frequency points, and reduces the processing time and calculation amount of the cell search for eliminating the false frequency points in the subsequent steps. .

Claims

Claim

1. A method for searching for a frequency of a cell, comprising:

For each frequency to be searched in the search interval: a sample that satisfies the number of one subframe, divides the collected samples into n signal segments, calculates the power spectral density of each signal segment, and according to Calculating a characteristic value of the signal segment by a power spectral density of each signal segment, and retaining a maximum feature value of the feature values of the n signal segments as a feature value of the frequency to be searched;

And characterizing the feature values of all the frequency points to be searched in the entire search interval into a feature value list, and using the to-be-searched frequency point corresponding to the largest feature value in the feature value list as the preferred frequency point;

Where η > 1.

2. The method of claim 1, wherein the step of determining a characteristic value of the signal segment based on a power spectral density of each of the signal segments comprises:

For each signal segment: Set the average power spectral density of the signal in the passband to T1, and the maximum value of the average power spectral density of the signal in the transition band to T2, and calculate the characteristic value of the signal segment T=T1 - Τ2.

3. The method of claim 1, wherein in the step of calculating the power spectral density of each of the signal segments, a power spectral density exceeding 12 dB is recorded as 12 dB.

4. The method of claim 1 further comprising:

Hard decision processing is performed on the data of each sample of the set.

5. The method of claim 1, wherein

In the step of dividing the collected samples into n signal segments, each signal segment contains less than or equal to 2048 samples.

6. The method of claim 1, further comprising: if the feature value of the preferred frequency point is greater than a feature value threshold, performing the following steps:

(a) using, as the candidate frequency corresponding to the preferred frequency point, the frequency of the to-be-searched corresponding to the second largest and third largest eigenvalues of the eigenvalues of the frequency points to be searched in the range of the preferred frequency point and the 1.6 MHz frequency range (b) clearing the eigenvalue of the frequency to be searched in the range of plus or minus 1.6 MHz of the preferred frequency point; if the selected frequency point of the selected frequency point is within the range of plus or minus 1.6 MHz, the frequency of the frequency to be searched If there is a feature value greater than the feature value threshold in the feature value, then the preferred frequency point is selected outside the range of plus or minus 1.6 MHz. The frequency of the to-be-searched corresponding point corresponding to the largest eigenvalue of the frequency of the frequency to be searched is used as the preferred frequency point, and the step (a) is repeated until the selected frequency point of the selected frequency point is to be searched outside the range of plus or minus 1.6 MHz. There is no eigenvalue greater than the eigenvalue threshold in the eigenvalues of the frequency points.

7. A search terminal for a cell frequency point, comprising a sample collection module, a segmentation module, a calculation module and a selection module; wherein

The sample collection module is configured to collect, for each of the to-be-searched frequency points in the search interval, a sample that satisfies the number of one subframe;

The calculating module is configured to calculate a power spectral density of each of the n signal segments, and obtain a characteristic value of the signal segment according to a power spectral density of each signal segment, and retain characteristics of the n signal segments a maximum eigenvalue in the value as a feature value of the frequency to be searched;

The selection module is configured to select, as a preferred frequency point, a frequency to be searched corresponding to the maximum feature value from a list of feature values consisting of feature values of all the frequency points to be searched in the entire search interval;

The η > 1.

8. The terminal according to claim 7, wherein the calculation module is configured to determine a characteristic value of the signal segment according to a power spectral density of each signal segment by: for each signal segment, setting the signal The average power spectral density of the signal in the segment passband is T1, and the maximum value of the average power spectral density of the signal in the transition band is T2, and the characteristic value T=T1 - Τ2 of the signal segment is calculated.

9. The terminal of claim 7, wherein

The calculation module is arranged to record a power spectral density of more than 12 dB as i2 dB when calculating the power spectral density of each signal segment.

10. The terminal of claim 7, further comprising a processing module, the processing module configured to perform hard decision processing on data of each sample of the sample collection module.

11. The terminal of claim 7, wherein

Each of the n signal segments includes a number of samples less than or equal to 2048.

12. The terminal of claim 7, wherein The selection module is further configured to: when the feature value of the selected preferred frequency point is greater than the feature value threshold, perform the following steps:

(a) using, as the candidate frequency corresponding to the preferred frequency point, the frequency of the to-be-searched corresponding to the second largest and third largest eigenvalues of the eigenvalues of the frequency points to be searched in the range of the preferred frequency point and the 1.6 MHz frequency range point;