CN1365207A - A digital baseband filter - Google Patents

Abstract

The digital base band filter based on linear interpolation includes sampling module, interpolation module, linear value calculating module and FIR low-pass module. The sampling module samples the input signal and outputs the sampled value to the interpolation module, the linear value calculating module calculates the linear value between two adjacent sampled values obtained from the sample module and outputs the calculated value to the interpolation module for interpolation, and the output of the interpolation module is output via the FIR low-pass module. The present invention can obtain high S/N ratio and suppress jamming between codes and is halpful to inhibiting the multipath effect of channel in the receiving end.

Description

A kind of digital baseband filter

The present invention relates to be used for the digital filtering technique of code division multiple address communication system, relate in particular to the baseband filtering device in a kind of code division multiple address communication system.

In present code division multiple address communication system, digital baseband filter generally adopts over-sampling (Over-sample) technology to make sample frequency much larger than the clock frequency of baseband digital signal, helps the interference of filtering sample frequency to baseband signal like this.Over-sampling is to make sample rate improve k doubly by insert k-1 new value between two sampled points.At present digital baseband filter generally adopts zero insertion value (the 285th page of " digital signal processing theory, algorithm and realization " book of publishing referring to publishing house of Tsing-Hua University in 1997) as the method for over-sampling, the baseband filter in the code division multiple access system also mostly adopts the zero insertion value as oversampler method.

Behind the baseband digital signal process zero insertion value over-sampling, can on high-end frequency domain, equally spaced produce the mapping spectrum component identical with main base-band signal spectrum structure.These new frequency component range values are identical, and number equals over-sampling multiple k and subtracts 1.Just because of the generation of these useless spectrum components, the range value of winner's base-band signal spectrum is descended, promptly the unwanted signal spectrum component of Sheng Chenging has disperseed the power of main base-band signal spectrum component, thereby has caused the generation of disturbing, and signal to noise ratio descends to some extent.

In code division multiple address communication system, by observing the output signal eye pattern of zero insertion value digital baseband filter, find that other sample points have bigger dispersing except the sample point everywhere convergent better of its best single times of spreading rate, the slope of eye pattern is bigger.Therefore, when in a single day sample point had the optimal sample point of departing from, the intersymbol interference that will be easy to bring was unfavorable for that simultaneously receiving terminal suppresses the multipath effect in the channel effectively.

The objective of the invention is to propose a kind of baseband filter, use this filter can obtain higher signal to noise ratio, suppress intersymbol interference effectively, and help receiving terminal to suppress the multipath effect of channel based on linear interpolation.

The object of the present invention is achieved like this: a kind of digital baseband filter comprises sampling module, interpolating module and FIR low-pass module; Sampling module is sampled to input signal and sampled value is outputed to interpolating module and carries out interpolation; The output of interpolating module is exported after the filtering of FIR low-pass module; This digital baseband filter also comprises a linear value computing module, and the linear value computing module obtains adjacent twice sampled value from sampling module, calculates linear value between the two and the linear value that calculates is outputed to interpolating module and carry out interpolation.

FIR low-pass module in the described digital baseband filter comprises that bus splits module 310, ROM pack module 320, the time-delay addition chain module 330 and the output module 340 that adds up; Bus fractionation module 310 will be carried out bus fractionation back timesharing through the digital signal after the linear interpolation and be outputed to the address of ROM pack module 320 as the ROM group; ROM pack module 320 is finished table lookup function the result is outputed to time-delay addition chain module 330 add operation of delaying time, and at last operation result is delivered to the output module 340 that the adds up output that adds up.

The data width of each the ROM piece in the ROM pack module 320 recited above is to determine according to the actual required figure place of the data of wherein being stored.

The present invention is described in further detail below in conjunction with accompanying drawing.

Fig. 1 is the zero insertion value digital baseband filter structural representation of using always;

Fig. 2 is the digital baseband filter structural representation based on linear interpolation that the present invention proposes;

Fig. 3 is the structural representation of FIR low-pass module among the present invention;

Fig. 4 is the particular circuit configurations schematic diagram of FIR low-pass module in one embodiment of the present of invention;

Fig. 5 is the oscillogram of input cosine wave;

Fig. 6 is that the input cosine wave is through the oscillogram after 4 times of over-samplings of zero insertion value;

Fig. 7 is that the input cosine wave is through the oscillogram after 4 times of over-samplings of linear interpolation;

Fig. 8 is the normalization spectrogram of input cosine wave signal;

Fig. 9 is the normalization spectrogram of the output signal behind 4 times of over-samplings of input cosine wave signal process zero insertion value;

Figure 10 is the normalization spectrogram of the output signal behind 4 times of over-samplings of input cosine wave signal process linear interpolation;

Figure 11 is that 4 times of over-samplings of input cosine wave signal process zero insertion value are after the normalization spectrogram of the output signal after the FIR low pass filtered module;

Figure 12 is that 4 times of over-samplings of input cosine wave signal process linear interpolation are after the normalization spectrogram of the output signal after the FIR low pass filtered module;

Figure 13 is the oscillogram of the bpsk signal at random of input;

Figure 14 is the binary eye pattern of the bpsk signal at random of input;

Figure 15 is the binary eye pattern of the bpsk signal at random of input through the output signal after the zero insertion value digital baseband filter;

Figure 16 is the binary eye pattern of the bpsk signal at random of input through the output signal after the linear slotting digital baseband value filtering device.

As shown in Figure 1, Chang Yong zero insertion value digital baseband filter comprises sampling module, interpolating module and FIR low-pass module; Sampling module is sampled to input signal and sampled value is outputed to interpolating module and carries out interpolation; Interpolating module carries out zero insertion between the sampled value of sampling module output, interpolation result is exported after the filtering of FIR low-pass module.

Fig. 2 is the digital baseband filter structural representation based on linear interpolation that the present invention proposes.As shown in Figure 2, the present invention includes sampling module, interpolating module and FIR low-pass module; Sampling module is sampled to input signal and sampled value is outputed to interpolating module and carries out interpolation; The output of interpolating module is exported after the filtering of FIR low-pass module; The present invention also comprises a linear value computing module, and the linear value computing module obtains adjacent twice sampled value from sampling module, calculates linear value between the two and the linear value that calculates is outputed to interpolating module and carry out interpolation.Can find that through comparing the contribution that the present invention did is, replaces the zero insertion value with linear interpolation, thereby improve the overall performance index of digital baseband filter.

One embodiment of the present of invention are the digital baseband filters that are used for the CDMA2000 communication system.As shown in Figure 2, before CDMA2000 base band composite signal x (t) is input to the FIR low-pass module, at first it is sampled through sampling module, the spreading rate (chip rate) that sampling rate equals 1 times, this sampled value x (m) and next sampled value x (m+1) are saved in the linear value computing module, in the linear value computing module, calculate each linear value I (i) that needs insertion then, realize interpolation after again these calculated values I (i) being sent into interpolating module, send the FIR low-pass module to carry out filtering output at last.

The computational methods of linear interpolation are as follows:

If over-sampling speed is k spreading rate doubly, then j linear value that is inserted into is I (i), suc as formula (1). $I\left(j\right)=x\left(m\right)+\frac{j}{k}\·[x(m+1)-x\left(m\right)]---\left(1\right)$

J=1 wherein, 2,3 ..., k-1; M=0, ± 1, ± 2 ...

Calculate to such an extent that linear value I (i) is inserted current sampled value x (m) successively in interpolating module.Corresponding this sampling obtains linear interpolation sequence v after finishing insertion _{J, k}(n), speed equals k spreading rate doubly.

v _{J, k}(n) expression formula is suc as formula (2):

v _{J, k}(n) filtering promptly gets output signal X through the FIR low-pass module again _L(n).The FIR low-pass module is typical finite impulse response filter, and its transfer function is formula (3), $H\left(z\right)=\underset{k=0}{\overset{N-1}{\mathrm{\Σ}}}h\left(k\right)z^{-k}---\left(3\right)$

H (k) is pulse response, it be with the specific window function to approaching that the endless pulse of perfect low pass responds, have limited exponent number.H (0), h (1) ..., h (N-1) forms the coefficient of FIR filter, they multiply each other with each delay cell of input signal respectively, output at last stacks up.

According to the characteristics of digital baseband filter employing linear interpolation, the present invention has also done corresponding improvement to the FIR low-pass module in the filter.Fig. 3 is the structural representation of FIR low-pass module among the present invention.As shown in Figure 3, the FIR low-pass module in the digital baseband filter that the present invention proposes comprises that bus splits module 310, ROM pack module 320, the time-delay addition chain module 330 and the output module 340 that adds up; Bus fractionation module 310 will be carried out bus fractionation back timesharing through the digital signal after the linear interpolation and be outputed to the address of ROM pack module 320 as the ROM group; ROM pack module 320 is finished table lookup function the result is outputed to time-delay addition chain module 330 add operation of delaying time, and at last operation result is delivered to the output module 340 that the adds up output that adds up.

Fig. 4 is the particular circuit configurations schematic diagram of FIR low-pass module in the embodiment of the invention.As shown in Figure 4, the data of input are that width is 16bit through the digital signal after the 1:4 linear interpolation, and speed is 4 times of 1.2288Mcps.Bus fractionation module 310 is split as 4 the tunnel from low to high with the bus of 16bit, every road 4bit closes 1 multiplexer (MUX) 311 by 4 then and outputs to ROM pack module 320, as the address wire 312 of ROM, therefore address wire 312 is 4bit, and speed is 16 times of 1.2288Mcps.The purpose of doing like this is to allow subsequent module timesharing computing, is cost to improve working clock frequency, and hardware resource is saved greatly.

ROM pack module 320 is that 16 ROM piece is formed by 24 addressing degree of depth, and the output figure place of 24 ROM pieces is respectively W0, W1 ..., Wi ... W23.What 16 address locations of each ROM piece were deposited is 0 to 15 times value of the FIR coefficient of corresponding CDMA2000 recommendation.Therefore export figure place W0, W1 ..., Wi ... W23 is determined by the actual required figure place of 15 times of values of the FIR coefficient of corresponding CDMA2000 recommendation.If do not consider actual required figure place, in order not overflow, the output figure place of 24 ROM all need unify to be increased to 20bit so, then ROM pack module 320 and follow-up time-delay addition chain module 330, output module 340 adds up, because data bits increases, hardware resource also can correspondingly increase greatly.In addition, utilize the symmetry of the FIR coefficient of CDMA2000 recommendation, can only get half, promptly therefore 24 coefficients are used only with 24 ROM pieces.

Time-delay addition chain module 330 is made up of 47 adders and 48 delay register groups.The delay register group is made up of 4 delay registers, finishes the time-delay (work clock is 16 times of 1.2288MHz) in 4 work clock cycles.The actual required figure place W0 that the figure place of adder and delay register group is exported by ROM pack module 320, W1 ..., Wi ... W23 calculates and determines, so also helps saving hardware resource.

Output after addition chain module 330 dateouts of will delaying time the output module 340 that adds up add up.In cumulative process, the output 342 of register 341 is divided into two-way 343 and 344, wherein one the tunnel 344 enters delay register for minimum 4bit, adds up for remaining high-order feedback enters adder in another road 343.Bus data 345 is formed by the merging of 4 circuit-switched data, through output register 346 outputs.The work clock of register 341 and register 348 is 16 times of 1.2288MHz, and register 346 work clocks are 4 times of 1.2288MHz.The speed of dateout 347 is 4 times of 1.2288Mcps.

Below in conjunction with Fig. 5～Figure 16 and embodiments of the invention beneficial effect of the present invention is described.In view of the requirement of CDMA2000 system, every index request of digital baseband filter is as follows: 0～590kHz is the 3dB free transmission range, and 590～740kHz is the transition band scope, greater than 740kHz is stopband range, stopband attenuation 40dB, passband ripple 1.5dB, FIR exponent number are 48 rank.In an embodiment of the present invention, all be used as reference with every index of CDMA2000 system requirements.

Fig. 5 is the oscillogram of input cosine wave.Be example now, come the signal output waveform and the spectrum structure of comparison linear interpolation and zero insertion value filtering device with the cosine wave input signal.The single-frequency value of this cosine signal equals 34.133kHz, and spreading rate is 1228800Hz, and a cosine cycle is 29.2969us, comprises 36 chips altogether.Cosine signal is respectively through after 4 times of linear interpolation over-samplings and the 4 times of zero insertion value over-samplings, exported the result separately: as shown in Figure 6 through the later signal output waveform of 4 times of zero insertion value over-samplings, through the later signal output waveform of 4 times of linear interpolation over-samplings as shown in Figure 7, can see from figure that output waveform and original waveform that linear interpolation is later quite approach, and the formation of the output waveform of zero insertion value over-sampling is a cosine envelope.

Respectively the signal among Fig. 5, Fig. 6, Fig. 7 is carried out the FFT conversion, can obtain their normalization spectrum structure figure, be respectively Fig. 8, Fig. 9 and Figure 10.

As shown in Figure 8, the normalization frequency spectrum of former input cosine signal presents two spectral lines, respectively corresponding 34.133kHz and 1194.666kHz.Fig. 9 is the normalization spectrogram of the output signal behind 4 times of over-samplings of input cosine wave signal process zero insertion value.As shown in Figure 9, the normalization frequency spectrum of 4 times of zero insertion value over-sampling output signals has comprised 4 pairs of spectral lines, a pair of spectral line on 34.133kHz and 4881.0667kHz point frequency, and the equally spaced three pairs of spectral lines that increase in the middle of also having, and their amplitude is identical.The a pair of single-frequency spectral power that former cosine wave signal is described is divided equally on four pairs of frequency components.

Figure 10 is the normalization spectrogram of the output signal behind 4 times of over-samplings of input cosine wave signal process linear interpolation.As shown in figure 10, the normalization frequency spectrum of 4 times of linear interpolation over-sampling output signals has also comprised 4 pairs of spectral lines, but except 34.133kHz and the 4881.0667kHz point a pair of main spectral line amplitude on frequently is higher, obviously little a lot of of the three pairs of spectral line amplitudes in its centre that increases newly.From figure, the amplitude of three newly-increased spectral lines is than the about little 30dB of main spectral line amplitude.The basic frequency of signal spectral power that hence one can see that after linear interpolation is very concentrated, illustrates that linear interpolation is than the higher signal to noise ratio of the easy acquisition of zero insertion value.

Make the output signal of 4 times of linear interpolation over-samplings and 4 times of zero insertion value over-samplings all pass through the FIR low-pass module at last and carry out low-pass filtering, obtain the normalization spectrogram of filtering output signal respectively, as Figure 11 and shown in Figure 12.

Figure 11 is that 4 times of over-samplings of input cosine wave signal process zero insertion value are after the normalization spectrogram of the output signal after the FIR low pass filtered module.As shown in figure 11, three pairs of high fdrequency components of the centre of 4 times of zero insertion value over-sampling generations are attenuated, and the decay back is than the low approximately 40dB of main spectral line.Figure 12 is that 4 times of over-samplings of input cosine wave signal process linear interpolation are after the normalization spectrogram of the output signal after the FIR low pass filtered module.Three pairs of high fdrequency components of the centre of 4 times of linear interpolation over-sampling generations also are attenuated as shown in figure 12, almost disappear among the noise, and the stopband attenuation at the 740kHz place reaches 59dB.

We can find that the linear interpolation over-sampling can more effectively suppress dispersing of useful signal spectrum power than zero insertion value over-sampling frequency spectrum by the output signal of linear interpolation and zero insertion value filtering device relatively, thereby obtain higher signal to noise ratio; By after the low pass filter, easier simultaneously by the unnecessary high fdrequency component of filtering, thus reduction is to the performance requirement of digital baseband filter.

Figure 13 is the oscillogram of the bpsk signal at random of input.We are input as example with bpsk signal at random, come the output signal eye pattern of comparison linear interpolation over-sampling filter and zero insertion value over-sampling filter.Figure 14 is the binary eye pattern of bpsk signal at random of input, this at random bpsk signal send into 4 times of zero insertion value filtering devices and 4 times of linear interpolation filters respectively, after over-sampling and low-pass filtering, obtain signal eye diagram separately respectively.

Figure 15 is the binary eye pattern of the bpsk signal at random of input through the output signal after the zero insertion value digital baseband filter; Figure 16 is the binary eye pattern of the bpsk signal at random of input through the output signal after the linear slotting digital baseband value filtering device.Relatively two eye patterns can see significantly that the output signal eye pattern degree of divergence of zero insertion value filtering device is bigger, except the sample point everywhere convergent better of its best single times of spreading rate, at other sample points bigger dispersing are arranged, and the slope of eye pattern is bigger.Therefore, when in a single day sample point has the optimal sample point of departing from, just be easy to bring intersymbol interference, be unfavorable for that simultaneously receiving terminal suppresses the multipath effect in the channel effectively.Correspondingly the degree of divergence of linear interpolation filter is less relatively, and except the sample point everywhere convergent better of its best single times of spreading rate, the convergence ratio at other sample point places adopts zero insertion value method better, and the slope of eye pattern is less.Therefore,, can not bring more intersymbol interference yet, more help simultaneously receiving terminal and suppress multipath effect in the channel effectively even sample point and optimal sample point have certain departing from.From the another one angle, adopt digital baseband filter can also reduce the requirement of system effectively to the sampling clock accuracy based on linear interpolation.

Embodiment above comprehensive and to the analysis of experimental result, we can draw, the present invention has following beneficial effect: the digital baseband filter based on linear interpolation not only can obtain higher signal to noise ratio, suppress intersymbol interference effectively, and help the code division multiple address communication system receiving terminal to suppress the multipath effect of channel.

Claims (3)

1, a kind of digital baseband filter comprises sampling module, interpolating module and FIR low-pass module; Sampling module is sampled to input signal and sampled value is outputed to interpolating module and carries out interpolation; The output of interpolating module is exported after the filtering of FIR low-pass module; It is characterized in that: also comprise a linear value computing module, the linear value computing module obtains adjacent twice sampled value from sampling module, calculates linear value between the two and the linear value that calculates is outputed to interpolating module and carry out interpolation.

2, a kind of digital baseband filter as claimed in claim 1 is characterized in that: described FIR low-pass module comprises that bus splits module (310), ROM pack module (320), time-delay addition chain module (330) and the output module that adds up (340); Bus fractionation module (310) will be carried out bus fractionation back timesharing through the digital signal after the linear interpolation and be outputed to the address of ROM pack module (320) as the ROM group; ROM pack module (320) is finished table lookup function the result is outputed to time-delay addition chain module (330) add operation of delaying time, and at last operation result is delivered to the output module that adds up (340) output that adds up.

3, a kind of digital baseband filter as claimed in claim 2 is characterized in that: the data width of each the ROM piece in the described ROM pack module (320) is to determine according to the actual required figure place of the data of wherein being stored.

Images