CN115955379B - Low-complexity implementation method of multi-scale configurable narrow transition band channelizer - Google Patents

Abstract

The invention aims to provide a low-complexity implementation method of a multi-scale configurable narrow transition band channelizer, which comprises the following steps: delay and decimate, polyphase filtering, FFT. The channel size of the low-complexity multi-scale channelizer can be configured according to input signals, performance indexes can be set more flexibly, better time and frequency resolution can be achieved, sensitivity is improved, delay is reduced as much as possible, meanwhile, a filter is optimized through a CEM-FRM method, and engineering implementation is easy.

Description

Low-complexity implementation method of multi-scale configurable narrow transition band channelizer

Technical Field

The invention relates to an electronic information and communication technology, in particular to a low-complexity implementation method of a narrow transition band channelizer.

Background

The channelization technology is a scheme widely used for solving the problem of simultaneous reception of multiple signals and multiple channels in a multi-carrier communication system, and is particularly used for receiving and detecting broadband complex electromagnetic signals. On the one hand, it is necessary to ensure the accuracy of signal detection, for example, in order to detect a signal, narrowband channelization is required for a weak input signal to improve the signal-to-noise ratio (SNR). On the other hand, it is desirable to minimize the delay of the receiver, e.g. for strong signals, the same narrowband channelization is not required, which increases the delay of the receiver. A fixed channelizer will typically balance the overall requirements of the receiver and cannot be personalized to meet the needs of different signals.

The multi-phase channelizer has a low complexity and an efficient implementation architecture as an important frequency domain channelizer, and the conventional multi-phase channelizer cannot dynamically adjust the channel size. Many studies have improved multiphase channelization structures to achieve better performance. In a low power consumption low complexity reconfigurable non-maximum decimation filter bank designed for high resolution wideband channels, it is proposed to increase the frequency resolution and sensitivity by increasing the number of channels and then to reduce the complexity using frequency response masking or multiplier-less techniques. The reconstruction of the channelized sub-signals to achieve variable bandwidth is attempted in the efficient wideband channelizer of a software radio system using modulated perfect reconstruction filter banks, which increases the delay, decreases the time resolution or increases the data volume. The design of low complexity reconfigurable digital intermediate frequency filters attempts to design variable bandwidth filters for wideband channelizers to achieve multi-scale reception, but they are more complex than multi-phase channelisation.

There have been many studies on narrow transition bandwidth filters, the unified approach to interpolation and frequency response mask FIR filter design, which suggests and extends the improved frequency response mask (Frequency response masking, FRM) approach to a broader filter design. The FRM method is applied to the filter bank design in the low-power consumption 16-band non-uniform filter bank design for the hearing aid so as to reduce the implementation complexity and realize the configurable uniform or non-uniform filter bank. FRM filter banks can reduce complexity but when FRM methods are used for prototype filter designs of polyphase Discrete Fourier Transform (DFT) channelizers, polyphase decomposition cannot be performed during implementation.

Disclosure of Invention

The invention aims to provide a low-complexity implementation method of a multi-scale configurable narrow transition band channelizer, which can solve the low-complexity implementation problem of a multi-phase Discrete Fourier Transform (DFT) filter bank.

The purpose of the invention is realized in the following way:

The invention discloses a low-complexity implementation method of a multi-scale configurable narrow transition band channelizer, which is characterized by comprising the following steps of:

(1) Delay and decimation:

Converting a signal with a sampling frequency of f _s to a signal with a sampling frequency of Is realized by clock synchronization and delay of data, and the frequency of the synchronous clock is set as

Extraction factor isFirst oneSynchronization processing, the rest of the sub-signals are processed by the firstSub-signal acquisition:

And Two groups of multiphase sub-signals when the synchronous clocks are CLK 1 and CLK 2 respectively;

(2) Polyphase filtering:

The prototype low-pass filter is decimated by D _h and embodied as discard in a multi-scale polyphase filter bank Prototype polyphase sub-filters, the polyphase sub-filters of the multiscale were obtained by prototype polyphase sub-filters:

Where h _dp (m) is the impulse response of the multi-scale polyphase sub-filter and h _p (m) is the impulse response of the prototype polyphase sub-filter;

The complexity of the filter is reduced by the FRM method, and the structure of the filter is optimized by the CEM-FRM method;

h _pr (n) is a half-band filter, h _r (n) is reduced to The multiphase component is expressed as

Wherein h _mr,p(n)＝h_mr(nK+p),h_mi,p(n)＝h_mi (nk+p);

representing with CEM-FRM filter bank Obtaining

Obtaining a multiphase component of a prototype filter by a CEM-FRM and coefficient extraction method, and obtaining a low-complexity comprehensive channelizer by replacing a multiphase sub-filter with a multiphase filter group component based on the CEM-FRM;

(3)FFT：

n-point FFT based on time extraction is decomposed step by step until it is divided into 2-point FFT, and the number of channels in the multi-scale channelizer is Wherein M _pr＝2^a,D_h＝2^c, a, c εN, and D _h＜M_pr, find M-point FFT from M _pro -point FFT by adjusting the data stream;

The multi-phase sub-signals with different scales are obtained through different clocks, 0-M-1 phase sub-signals are output to a multi-phase filtering module, the M-phase sub-signals after multi-phase filtering are input from pins 0 and D _h,2D_h,…,(M-1)D_h of an FFT with M _pro points, meanwhile, the other pins are input to 0, and multi-scale channelized output is obtained from output pins 0-M of a channelizer.

The invention has the advantages that: the channel size of the low-complexity multi-scale channelizer can be configured according to input signals, performance indexes can be set more flexibly, better time and frequency resolution can be achieved, sensitivity is improved, delay is reduced as much as possible, meanwhile, a filter is optimized through a CEM-FRM method, and engineering implementation is easy.

Drawings

FIG. 1 is a diagram of a conventional multi-phase channelization structure;

FIG. 2 is a schematic diagram of a spectrum of a multi-scale channel division;

FIG. 3 is a frequency domain schematic of the CEM-FRM method;

FIG. 4 is a schematic diagram of latency and execution time of decimation;

FIG. 5 is a block diagram of a polyphase filter bank;

FIG. 6 is a schematic diagram of an 8-point FFT based on time domain decimation;

FIG. 7 is a block diagram of a multi-scale multi-phase channelizer;

Fig. 8 is an amplitude-frequency response of a multi-scale low-pass filter.

Detailed Description

The invention is described in more detail below, by way of example, with reference to the accompanying drawings:

in connection with fig. 1-8, the present invention proposes a low-complexity multi-scale multi-phase digital channelizer that changes the sub-channel bandwidth by means of coefficient extraction, only requires to design a prototype low-pass filter, and maintains the simplicity and low complexity of the multi-phase channelizing structure.

In a conventional M-channelizer, the input signal x (n) is first mixed with the center frequency of the sub-channel, then low-pass filtered, and finally downsampled to obtain the output sub-signal y _k (M). The computation of the output sub-signals can be expressed as

Where M is the channel number, k is the channel number, k=0, 1, …, M-1, d is the downsampling factor,H (n) is the impulse response of the low pass filter,Is a convolution.

The downsampling factor D is typically an integer no greater than M, and is typically selected as m·2 ^-b, b e N in a polyphase channelization structure. Let n=nm+p, we can obtain the sub-signals of the polyphase channelizer output as

Where x _p(m)＝x(mD-p),h_p (m) =h (mm+p).

From equation (2), a conventional polyphase channelizer can be obtained, as shown in figure one.

The multi-scale channelizer aims at realizing the channelizing of different sizes, and a frequency domain diagram of multi-scale channel division is shown in a second diagram. The relationship between the sub-channel bandwidth B and the number of channels M satisfies b·m=2pi. The number of polyphase sub-filters and the size of the Discrete Fourier Transform (DFT) are both M, the bandwidth of the low-pass filter is the same as the bandwidth of the sub-channels. Since DFT is often replaced by a Fast Fourier Transform (FFT) in the implementation process, the number of channels in the multi-scale channelizer satisfies m=2 ^a, and the bandwidth satisfiesa∈N。

In order to obtain the multi-scale channelizer, a coefficient extraction method is adopted to obtain a variable bandwidth low-pass filter. When the low-pass filter is downsampled by a factor D _h, the discrete-time fourier transform (DTFT) of the decimated low-pass filter can be represented by the prototype low-pass filter DTFT

Where H _d(e^jω) is the DTFT of the decimated low pass filter and H (e ^jω) is the prototype low pass filter DTFT.

As can be seen from (3), H _d(e^jω) is a copy of the amplitude and frequency of the period DTFT H (e ^jω) scaled by D _h times, shifted by an integer multiple of 2 pi. Thus, the integer multiple decimation of the low pass filter can meet the bandwidth requirements of a multi-scale, multi-phase channelizer.

In order to ensure that the decimated low pass filter maintains a linear phase during decimation of the prototype low pass filter, the prototype low pass filter should be of even order so that the integer multiple of decimated filter coefficients are even (odd) symmetric. The relation of the decimating low pass filter and the prototype low pass filter can be expressed as

h_d(n)＝h(nD_h+mod₁) (4)

Wherein h _d (n) is the impulse response of the decimated low-pass filter, h (n) is the impulse response of the prototype low-pass filter, h (n) isN _pro is the order of the prototype low-pass filter divided by the remainder of D _h.

According to the channel division shown in figure two, the decimation factor is chosen to beD and M of the multi-scale channelizer can be represented by D _pro and M _pro of the prototype channelizer with the narrowest bandwidth and the most channels, e.gAndBy bringing (4) into (2), we can obtain a multi-scale output sub-signal

Wherein the method comprises the steps ofh_dp(m)＝h(mM_pro+pD_h+mod₁)。

According to (2) and (5), the multi-scale multi-phase channel signal processing architecture has the following variations compared to a conventional multi-phase channelizer:

1) Downsampling factor becomes

2) Due to decimation, some polyphase sub-filters are discarded,

3) Channel and FFT size becomes

The decimation factor D _h is variable, different D _h means different scales, and the multi-scale channelizer can be configured by selecting different decimation factors D _h.

To reduce complexity, an improved method of FRM was introduced. The FRM-based filter h (n) can be obtained by (6), h _a (n) and h _c (n) are the impulse responses of two spectrally complementary modal filters, L is the decimation factor of the FRM method, and h _ma (n) and h _mc (n) are the impulse responses of two masking filters.

Can obtain

Will beAnd h _ma (n) into K subsequencesAnd h _ma,p (n), where p=0, 1..k-1

When L is an integer multiple of K, we can further obtain

The polyphase component of h (n) can be expressed as

Since the conventional FRM method cannot easily satisfy the condition that L is an integer multiple of K, we propose a CEM-FRM method in order to obtain a polyphase component of a FRM-based prototype filter.

A frequency domain representation of the CEM-FRM method is shown in figure three. H _m (z) is the z-transform of the masking filter H _m (n) having pass and stop band cut-off frequencies ω _mp and ω _ms, H _pr (z) is the z-transform of the half-band filter H _pr (n) having pass and stop band cut-off frequencies ω _ap and ω _as, and H (z) is the z-transform of the synthesis filter H (n) having pass and stop band cut-off frequencies ω _p and ω _s. H _a(z),H_c(z),H_ma (z) and H _mc (z) are the z-transforms of filters H _a(n),h_c(n),h_ma (n) and H _mc (n). The order of the filter h _m (N) is N _m and the order of the filter h _pr (N) is N _pr.

H _a (n) and h _c (n) are represented by h _pr (n)

H _ma (n) and h _mc (n) are represented by h _m (n)

Order the

We can obtain CEM-FRM method

The key to the implementation of a multi-scale channelizer is how to effectively implement different scales. The implementation of a multi-scale multiphase channelizer can be divided into three main modules: delay and decimate, polyphase filtering, and also FFT. The implementation is described below from three sub-modules to the whole channelizer.

1. Delay and decimation

The delay and decimate module converts a signal with a sampling frequency of f _s to a sampling frequency ofThis can be achieved by synchronizing the delay data using different clocks as shown in figure four.

The frequency of the synchronous clock can be set to

In the fourth view of the drawing, there is shown,AndTwo sets of multiphase sub-signals with synchronous clocks CLK 1 and CLK 2, respectively. Due to the extraction factor ofThus only the first one is neededThe rest of the sub-signals can be processed synchronously by the first oneSub-signal acquisition:

2. polyphase filtering

To achieve different scale bandwidths, the prototype low-pass filter is decimated by D _h and embodied as a discard in a multi-scale polyphase filter bankPrototype polyphase sub-filters. The multi-scale polyphase sub-filter can be obtained by prototype polyphase sub-filters:

Where h _dp (m) is the impulse response of the multi-scale polyphase sub-filter and h _p (m) is the impulse response of the prototype polyphase sub-filter.

The complexity of the filter can be reduced by the FRM method, but as mentioned before, conventional FRM methods do not necessarily enable to obtain the polyphase components of the FRM-based prototype filter. The structure of the filter is optimized by the CEM-FRM method as follows.

Since h _pr (n) in (12) is a half-band filter, h _r (n) can be reduced to(12) The multiphase component of (2) can be expressed as

Wherein h _mr,p(n)＝h_mr(nK+p),h_mi,p(n)＝h_mi (nK+p).

Using CEM-FRM filter bank representation (15), one can obtain

According to (16), the polyphase component of the prototype filter can be obtained by CEM-FRM and coefficient extraction methods. By replacing the polyphase sub-filters with the CEM-FRM based polyphase filter component amounts, a low complexity integrated channelizer is obtained.

The structure of the polyphase filter bank is shown in fig. five, the polyphase sub-filters in the dashed box are discarded, and the remaining sub-filters in the solid box form a new polyphase filter bank. To guarantee the linear phase of the channelizer, the inclusion of impulse responses cannot be discardedIs included in the filter.

3、FFT

As shown in fig. six, the time-decimated N-point FFT may be decomposed step-wise until it is divided into several 2-point FFTs. The number of channels in the multi-scale channelizer isWhere M _pr＝2^a,D_h＝2^c, a, c εN, and D _h＜M_pr. Thus, by adjusting the data stream, an M-point FFT can always be found from the M _pro -point FFT.

For example, in order to implement a 4-point FFT by an 8-point FFT, signals may be input from input pins 0, 2, 4, and 6, and then output from output pins 0,1, 2, and 3, with inputs of other pins being 0, or signals are input from input pins 1,3, 5, and 7, with inputs of other pins being 0, and output from pins 4, 5, 6, and 7. Blue and green in fig. six are schematic diagrams of implementing 4-point and 2-point FFTs using 8-point FFTs, respectively.

4. Multi-scale multiphase channelizer

The implementation structure of the multi-scale multi-phase channelizer is shown in fig. seven. The delay and extraction module obtains multiphase sub-signals with different scales through different clocks, and outputs 0-M-1 phase sub-signals to the multiphase filtering module. The multi-phase filtered M-phase sub-signals are input from pins 0, D _h,2D_h,…,(M-1)D_h of the M _pro point FFT, and the other pins are input to be 0, and then we can obtain multi-scale channelized output from the 0-M output pins of the channelizer.

Performance of different scale filters

In the simulation, the maximum number of channels is assumed to be 64, the bandwidth of the prototype low-pass filter is 0.015625 pi, and the filter parameters of different scales are shown in the table one. When the prototype low-pass filters are decimated by factors 2,4, 8 and 16, 32, 16, 8 and 4 channel channelizers are obtained, respectively. The amplitude-frequency response of these low-pass filters varies with the decimation factor, while the stopband attenuation deteriorates with decimation, as shown in figure eight. The stopband rejection of the channelized output result is affected by the stopband attenuation of the filter and the decimation factor of the output signal. Therefore, the stopband rejection of the multi-scale channelized output meets the same criteria and is not degraded by degradation of the stopband attenuation of the multi-scale filter.

Performance index of surface two multi-scale channelizer

For an electronic warfare receiver with a sampling frequency of 1GHz, the noise figure is assumed to be 0dB and the recognition factor is assumed to be 5dB. Table two lists the delay, frequency resolution, time resolution and sensitivity for several different scales. In a polyphase channelized receiver, the delay and time resolution must deteriorate as the sensitivity and frequency resolution increase, and vice versa. The multi-scale channelizer provided by the invention can flexibly change the channelizing scale through dynamic adjustment, and adapt to the detection requirements of different signals, thereby improving the dynamic performance of a receiver.

Claims (1)

1. A low-complexity implementation method of a multi-scale configurable narrow transition band channelizer is characterized by comprising the following steps:

(1) Delay and decimation:

Converting a signal with a sampling frequency of f _s to a signal with a sampling frequency of Is realized by clock synchronization and delay of data, and the frequency of the synchronous clock is set as

Extraction factor isFirst oneSynchronization processing, the rest of the sub-signals are processed by the firstSub-signal acquisition:

And Two groups of multiphase sub-signals when the synchronous clocks are CLK1 and CLK2 respectively;

(2) Polyphase filtering:

The prototype low-pass filter is decimated by D _h and embodied as discard in a multi-scale polyphase filter bank Prototype polyphase sub-filters, the polyphase sub-filters of the multiscale were obtained by prototype polyphase sub-filters:

Where h _dp (m) is the impulse response of the multi-scale polyphase sub-filter and h _p (m) is the impulse response of the prototype polyphase sub-filter;

The complexity of the filter is reduced by the FRM method, and the structure of the filter is optimized by the CEM-FRM method;

h _pr (n) is a half-band filter, h _r (n) is reduced to The multiphase component is expressed as

Wherein h _mr,p(n)＝h_mr(nK+p),h_mi,p(n)＝h_mi (nk+p);

representing with CEM-FRM filter bank Obtaining

Obtaining a multiphase component of a prototype filter by a CEM-FRM and coefficient extraction method, and obtaining a low-complexity comprehensive channelizer by replacing a multiphase sub-filter with a multiphase filter group component based on the CEM-FRM;

(3)FFT：

n-point FFT based on time extraction is decomposed step by step until it is divided into 2-point FFT, and the number of channels in the multi-scale channelizer is Wherein M _pr＝2^a,D_h＝2^c, a, c εN, and D _h＜M_pr, find M-point FFT from M _pro -point FFT by adjusting the data stream;

The multi-phase sub-signals with different scales are obtained through different clocks, 0-M-1 phase sub-signals are output to a multi-phase filtering module, the M-phase sub-signals after multi-phase filtering are input from pins 0 and D _h,2D_h,…,(M-1)D_h of an FFT with M _pro points, meanwhile, the other pins are input to 0, and multi-scale channelized output is obtained from output pins 0-M of a channelizer.