Embodiment
For the purpose of this utility model, technical scheme and advantage is more clearly understood, below in conjunction with drawings and Examples,
The utility model is described in further detail.It should be appreciated that embodiment described herein is only to solve
The utility model is released, protection domain of the present utility model is not limited.
It is shown in Figure 1, it is the structural representation of digital filtering equipment of the present utility model.Numeral filter in the embodiment
Wave device, including CIC draw-out devices 100, FIR filter 200 and CIC interpolating apparatus 300;CIC refers to that cascade integral dredges shape filter
Ripple, FIR filter refers to there is limit for length's unit impact response wave filter;
The P of CIC draw-out devices 100i1Individual input receives signal data to be filtered, wherein, Pi1For the integer more than 0;
The P of CIC draw-out devices 100c1The P of individual output end and FIR filter 200c1Individual input correspondence is connected, wherein, Pc1
For the integer more than 0;
The P of FIR filter 200c2Individual output end and the P of CIC interpolating apparatus 300c2Individual input correspondence is connected, wherein, Pc2
For the integer more than 0;
The P of CIC interpolating apparatus 300i2Individual output end exports filtering data, wherein, Pi2For the integer more than 0.
In the present embodiment, digital filtering equipment, it includes CIC draw-out devices, FIR filter and the CIC being sequentially connected
Interpolating apparatus, CIC draw-out devices are received after signal data to be filtered, can be obtained multidiameter delay data, be passed through CIC draw-out devices
Extraction down coversion is carried out to multidiameter delay data, FIR filter is filtered to the signal after down coversion, inserted finally by CIC
Value device enters row interpolation up-conversion to filtered signal, by its frequency retrieval to the frequency of original signal data to be filtered, obtains
To final filter result;This programme is by primary frequency signal down coversion post filtering, then by filtered signal up-conversion to former frequency
Rate, coordinates FIR filter coefficient can configure, it is possible to achieve to be filtered processing to the signal of optional frequency in high bandwidth;Numeral
Prime carries out down coversion using CIC draw-out devices in filter apparatus can be reduced because the frequency spectrum that sample rate is not enough and causes is mixed
Folded, rear class carries out up-conversion using CIC interpolating apparatus can realize that anti-mirror image is filtered, and up-conversion and down coversion can be reduced with this
The influence filtered to high-frequency signal;Meanwhile, using multidiameter delay structure, real-time to input data it can handle (without first storing
Post processing), the Real-Time Filtering processing of high-bandwidth signals is realized under less system clock.
It should be noted that the input input of CIC draw-out devices is data signal.Due to the utility model multiple parallel
Structure can realize that the high speed of any bandwidth signal is real in the case of and resource limited in system operation clock permission in theory
When filtering process.Assuming that simultaneously line number is P to input datain, the utility model system operation clock frequency is fsys, this practicality in addition
Each module can realize processing input data in real time using pipeline organization in new, you can be inputted once in each system clock
In data, each system clock cycle can input data amount be Pin, therefore the utility model can process signal maximum sample rate
fsFor:
fs=fsys×Pin
As long as increasing PinThe maximum sample rate of energy process signal can just be increased, you can to increase the band of energy process signal
It is wide.Certainly P is increasedinCan cause resource using increase, as long as so resource allow, no matter fsysMore small (of course greater than 0) is as long as increase
Big PinThe filtering process of any frequency signal can be met.The use of common single parallel organization wave filter is that can not realize this characteristic
's.
It is preferred that, Pi1With Pi2With identical the data volume to be filtered of input can be made identical with the filtered data amount exported,
The rear class processing of signal, P are not affected thereby using the utility model or bypass the utility modelc1With Pc2It is identical, filter FIR
The input data amount of ripple device is identical with output data quantity, and the complete of signal is ensured after FIR filter is filtered processing to signal
Whole property.
Optionally, Pi1、Pi2、Pc1With Pc2Can be 1, it is now actual for single parallel organization;Pi1、Pi2、Pc1With Pc2Can be with
For the integer more than 1, now actual is multiple parallel structure.
In one of the embodiments, as shown in Fig. 2 (a), CIC draw-out devices 100 include output end, input and connected successively
The N connectcic1Individual single-stage parallel C IC draw-out devices the 110, the 1st to Ncic1- 1 single-stage parallel C IC draw-out device 110 is provided with Pi1
Individual input and Pi1Individual output end, Ncic1Individual single-stage parallel C IC draw-out devices 110 possess Pi1Individual input and Pc1Individual output
End, the P of the 1st single-stage parallel C IC draw-out device 110i1Individual input as CIC draw-out devices 100 Pi1Individual input, the
Ncic1The P of individual single-stage parallel C IC draw-out devices 110c1Individual output end as CIC draw-out devices 100 Pc1Individual output end;
As shown in Fig. 2 (b), the N that CIC interpolating apparatus 300 includes output end, input is sequentially connectedcic2Individual single-stage is parallel
CIC interpolating apparatus the 310, the 1st to Ncic2- 1 single-stage parallel C IC interpolating apparatus 310 is provided with Pc2Individual input and Pc2Individual output
End, Ncic2Individual single-stage parallel C IC interpolating apparatus 310 possesses Pc2Individual input and Pi2Individual output end;1st single-stage parallel C IC
The P of interpolating apparatus 310c2Individual input as CIC interpolating apparatus 300 Pc2Individual input, Ncic1Individual single-stage parallel C IC interpolation
The P of device 310i2Individual output end as CIC interpolating apparatus 300 Pi2Individual output end;
Ncic1、Ncic2It is the integer more than 0.
In the present embodiment, effect is preferably extracted in order to reach, parallel C IC draw-out devices usually require larger extraction
Multiple, decimation factor and the extracting multiple positive correlation of parallel C IC draw-out devices, but when the extraction of a parallel C IC draw-out device
When the factor is larger, the input data bit wide of parallel C IC draw-out devices is larger, resource occupation of the parallel C IC draw-out devices to system
It is more, and the structure for using multiple single-stage parallel C IC draw-out devices to connect, the extracting multiple of parallel C IC draw-out devices is each
The product of the extracting multiple of single-stage parallel C IC draw-out devices, can obtain larger extracting multiple with less decimation factor,
It is simultaneously less to the resource occupation of system;After upper level single-stage parallel C IC draw-out devices are extracted, data are reduced, but in order to be able to
Same extraction is performed in next stage single-stage parallel C IC draw-out devices, it is necessary to which the data after extraction are carried out into delayed allocation, is made
Parallel output and the parallel input of next stage single-stage parallel C IC draw-out devices and line number is identical;
In order to reach preferable interpolation, parallel C IC interpolating apparatus usually requires larger interpolation multiple, parallel C IC
The interpolation factor of interpolating apparatus and interpolation multiple positive correlation, but when the interpolation factor of a parallel C IC interpolating apparatus is larger, and
Row CIC interpolating apparatus is more to the resource occupation of system, and the structure for using multiple single-stage parallel C IC interpolating apparatus to connect, and
The interpolation multiple of row CIC interpolating apparatus is the product of the interpolation multiple of each single-stage parallel C IC interpolating apparatus, can be with less
Interpolation factor obtains larger interpolation multiple, while less to the resource occupation of system;Upper level single-stage parallel C IC interpolation is filled
Put after interpolation, data increase, but in order to be able to perform same interpolation in next stage single-stage parallel C IC interpolating apparatus, it is necessary to will
The later data of interpolation carry out delayed allocation, make parallel output and the parallel input of next stage single-stage parallel C IC interpolating apparatus
And line number is identical.
Optionally, the number N of single-stage parallel C IC draw-out devicescic1With the number N of single-stage parallel C IC interpolating apparatuscic2Can
, can also be different with identical.
In one of the embodiments, as shown in Fig. 3 (a) and Fig. 3 (b), single-stage parallel C IC draw-out devices 110 include defeated
Go out end, the N that is sequentially connected of inputi1Individual parallel integration module 112, paralleling abstracting module 114, Nc1Individual shape module 116 thin parallel
And first gain regulation module 118, Ni1And Nc1It is the integer more than 0;
Each parallel integration module 112 possesses Pi1Individual input and Pi1Individual output end, paralleling abstracting module 114 possesses Pi1
Individual input and Pc1Individual output end, each shape module 116 thin parallel possesses Pc1Individual input and Pc1Individual output end;1st to
Ncic1Each first gain regulation module 118 in -1 single-stage parallel C IC draw-out device 110 possesses Pc1Individual input and Pi1It is individual defeated
Go out end, Ncic1The first gain regulation module in individual single-stage parallel C IC draw-out devices 110 possesses Pc1Individual input and Pc1It is individual defeated
Go out end;
As shown in Fig. 3 (c) and Fig. 3 (d), single-stage parallel C IC interpolating apparatus 310 includes output end, input and is sequentially connected
Nc2Individual shape module 312 thin parallel, parallel interpolation module 314, Ni2The individual Gain tuning mould of parallel integration module 316 and second
Block 318;
Each shape module 312 thin parallel possesses Pc2Individual input and Pc2Individual output end, parallel interpolation module 314 possesses Pc2
Individual input and Pi2Individual output end, each parallel integration module 316 possesses Pi2Individual input and Pi2Individual output end;1st to
Ncic2Each second gain regulation module 318 in -1 single-stage parallel C IC interpolating apparatus 310 possesses Pi2Individual input and Pc2It is individual defeated
Go out end, Ncic2The second gain regulation module in individual single-stage parallel C IC interpolating apparatus 310 possesses Pi2Individual input and Pi2It is individual defeated
Go out end.
In the present embodiment, single-stage parallel C IC draw-out devices include Ni1Individual parallel integration module, paralleling abstracting module, Nc1
Individual shape module thin parallel and the first gain regulation module, by such a annexation, can be carried out parallel to the signal of input
Extract, and reduce due to the spectral aliasing that the possible caused sample rate of extraction is not enough and causes, parallel C IC abstraction modules are to letter
Number there is gain amplification, therefore, need before output by the first gain regulation module, gain tune is carried out to signal
It is whole;Moreover, in order to be able to make the processing that same and line number is performed in next stage single-stage parallel C IC draw-out devices, the first Gain tuning
Module needs that later data progress delayed allocation will be extracted, and makes parallel output and next stage single-stage parallel C IC draw-out devices
Parallel input correspondence;
Single-stage parallel C IC interpolating apparatus includes Nc2Individual shape module thin parallel, parallel interpolation module, Ni2Individual parallel integration mould
Block and the second gain regulation module, by such a annexation, can carry out parallel interpolation, and realize to the signal of input
The filtering of anti-mirror image, because parallel C IC interpolating modules play the role of gain amplification to signal, therefore, need before output by
Second gain regulation module, Gain tuning is carried out to signal;It is same in order to be able to be performed in next stage single-stage parallel C IC interpolating apparatus
Interpolation, the second gain regulation module needs the later data of interpolation carrying out delayed allocation, makes parallel output and next stage list
The parallel input correspondence of level parallel C IC interpolating apparatus.
In one of the embodiments, as shown in figure 4, parallel integration module includes matrix of device [Ai,j], wherein, 1≤i
≤ p, 1≤j≤p, p are the parallel channel number of parallel integration module, and i, j, p are integer;
The device A of i-th row i-th -1 rowi,i-1The device A arranged with pthi,pIt is adder, matrix of device [Ai,j] in it is remaining
Device be storage delay register;
Matrix of device [Ai,j] in be sequentially connected per the device of a line, the adder A of the i-th row i-th -1 rowi,i-1It is sequentially connected;
Storage delay register A1,1Input and adder A2,1First input end connection, adder Ak,k-1Output
End and adder Ak+1,kFirst input end connection, while adder Ak,k-1Output end and storage delay register Ak,kIt is defeated
Enter end connection, wherein, 2≤k≤p-1, k is integer;
Storage delay register Ah,h-2Output end and adder Ah,h-1The second input connection, wherein, 3≤h≤p,
H is integer;
Storage delay register Ag,p-1Output end and adder Ag,pThe second input connection, wherein, 1≤g≤p-1,
G is integer, adder Ap,p-1Output end and adder Ap,pThe second input connection, adder Ap,pOutput end and addition
Device Ai,pFirst input end connection;
The input of parallel integration module includes storage delay register A1,1Input, adder A2,1Second input
End and storage delay register Ah,1Input, the output end of parallel integration module includes adder Ai,pOutput end;
Parallel integration module is parallel integration module 112 or parallel integration module 316.
In the present embodiment, parallel integration module is mainly realized by adder, is calculating the integration of a certain clock cycle
As a result it is preceding, it is necessary to by before plus with result calculate and complete, integration structure to add up before data, this programme passes through device square
Battle array mode is handled signal data, and storage delay register can export one clock cycle of data delay of storage,
Adder can also postpone a clock cycle output after two input values are added, and first be added up in the preceding p-1 row of matrix of device each
The value of level, present clock period result, last clock cycle are obtained in last row with upper clock cycle result phase Calais again
Last data is the previous data that present clock period inputs first data in p data in p data of input,
Need to wait last clock cycle input data to carry out the calculating of present clock period after the completion of calculating in theory, and it is our
Structure used in case, it is first that present clock period input data is at different levels cumulative, when present clock period inputs at different levels added up
The result of a upper clock cycle just calculates completion, and the accumulated value of present clock period input just can be with a upper clock
Last result adds up in computation of Period result, obtains the integral result value of present clock period.In each clock cycle
It is subjected to new data simultaneously to be handled, i.e., data can input processing in real time.
Optionally, the parallel channel number in parallel integration module 212 can with the parallel channel number in parallel integration module 416
, can also be different with identical;When parallel integration module is parallel integration module 212, parallel channel number p is Pi1;When parallel product
When sub-module is parallel integration module 416, parallel channel number p is Pi2。
In one of the embodiments, as shown in figure 5, thin shape module includes q subtracter and a storage delay parallel
Register, wherein, q is the parallel parallel channel number for dredging shape module;
The input of parallel thin shape module includes the first input end of all subtracters, the output end bag of parallel thin shape module
Include the output end of all subtracters;
The first input end of r-th of subtracter is connected with second input of (r+1) individual subtracter;Wherein, 1≤r≤
Q-1, r, q are integer;
The first input end of q-th of subtracter is connected with the input of storage delay register, storage delay register
Output end is connected with the second input of the 1st subtracter;
Thin shape module dredges shape module 116 or parallel thin shape module 312 to be parallel parallel.
In the present embodiment, thin shape module is mainly realized by subtracter parallel, the signal number of upper clock periodical input
According to last data for present clock period input first data previous data, present clock period is inputted
Each data and upper clock periodical input one clock cycle of last data delay after carry out making difference processing, to obtain
The differentiation result value of current period.New data are subjected in each clock cycle and are handled, i.e., data can be real
When input processing.
Optionally, the parallel channel number in parallel thin shape module 216 can with the parallel channel number in parallel thin shape module 412
, can also be different with identical;When parallel thin shape module is thin shape module 216 parallel, parallel channel number q is Pc1;Dredged when parallel
When shape module is thin shape module 412 parallel, parallel channel number q is Pc2。
In one of the embodiments, as shown in Fig. 6 (a) and Fig. 6 (b), FIR filter 200 postpones including data distribution
Chain module 210 and Pc2The individual parallel FIR filter 220 of single-stage;
Data distribution delay chain module 210 includes (N+Pc2- 1) the individual storage delay register being arranged in order, wherein, n-th
The output end of individual storage delay register and (n+Pc1) individual storage delay register input connection, 1≤n≤N-1, n, N
It is integer, N is the exponent number of the parallel FIR filter 220 of single-stage;
M is to the output end of (N+m-1) individual storage delay register and the input of the parallel FIR filter 220 of m-th of single-stage
End correspondence connection, wherein, 1≤m≤Pc2, m is integer;
The input of FIR filter 200 includes the 1st to Pc1The input of individual storage delay register, FIR filter 200
Output end include the output end of the parallel FIR filter 220 of all single-stages.
In the present embodiment, a length of (N+P of the delay chain of data distribution delay chain modulec2- 1) individual data, each clock week
Phase is from the 1st to Pc1Individual storage delay register inputs Pc1Individual data, in following clock cycle, the 1st to Pc1Individual storage delay deposit
The data of device are moved to Pc1+ 1 to 2Pc1In individual storage delay register, m is exported in each clock cycle to (N+m-1)
The data of individual storage delay register are to the parallel FIR filter of m-th of single-stage, and the mode similar to sliding window is exported, each
The parallel FIR filter of single-stage can export a filter result, P in a clock cyclec2The individual parallel FIR filter of single-stage is one
The individual clock cycle can export P simultaneouslyc2Individual filter result;Data are distributed in sliding window mode, it is possible to use less storage
Delay time register realizes parallel filtering.
It is preferred that, the fan-in P of FIR filter 200c1With the fan-out P of FIR filter 200c2It is identical.
In one of the embodiments, it is single as shown in fig. 7, when the exponent number of the parallel FIR filter 220 of single-stage is even number
The parallel FIR filter 220 of level includes N/2 adder, N/2 multiplier and an accumulator;
In the N number of storage delay register being arranged in order being connected with the parallel FIR filter of single-stage 220, s-th of storage is prolonged
The output end of storage that delays in the dispatch of and the output end of (N+1-s) individual storage delay register are connected respectively to a corresponding adder
Two inputs, wherein, 1≤s≤N, the input of the output end of each adder and a corresponding multiplier is connected, often
The output end of individual multiplier is connected with the corresponding input of accumulator, and the output end of accumulator is the parallel FIR filter of single-stage
220 output end.
In the present embodiment, when the exponent number of the parallel FIR filter 220 of single-stage is even number, N/2 adder, N/2 are utilized
Individual multiplier and an accumulator are that the filtering to one group of data can be achieved.
Optionally, accumulator can be adder tree or meet each clock cycle can multiple parallel input, output
The parallel organization of one accumulation result.
In one of the embodiments, it is single as shown in figure 8, when the exponent number of the parallel FIR filter 220 of single-stage is odd number
The parallel FIR filter 320 of level includes (N-1)/2 adder, delayer, (N+1)/2 multiplier and an accumulator;
In the N number of storage delay register being arranged in order being connected with the parallel FIR filter of single-stage 220, s-th of storage is prolonged
The output end of storage that delays in the dispatch of and the output end of (N+1-s) individual storage delay register are connected respectively to a corresponding adder
Two inputs, wherein, 1≤s≤N, the input of the output end of each adder and a corresponding multiplier is connected, the
(N+1)/2 input connection of the output end of a storage delay register and delayer, the output end of delayer and corresponding one
The input connection of individual multiplier, the corresponding input connection of the output end of each multiplier with accumulator, accumulator it is defeated
Go out output end of the end for the parallel FIR filter 220 of single-stage.
In the present embodiment, when the exponent number of the parallel FIR filter 220 of single-stage is odd number, (N-1)/2 addition is utilized
Device, delayer, (N+1)/2 multiplier and an accumulator are that the filtering to one group of data can be achieved, because exponent number is
Odd number, one of storage delay register does not have corresponding adder to be attached thereto, and adder is when carrying out additional calculation
Delay is had, accordingly, it would be desirable to the storage delay register is connected with delayer, while to ensure that each multiplier receives data
Property.
Optionally, accumulator can be adder tree or meet each clock cycle can multiple parallel input, output
The parallel organization of one accumulation result.
In one of the embodiments, as shown in figure 9, digital filtering equipment also includes withdrawal device 400 and interpolater 500,
Withdrawal device 400 is connected between the input of the output end of CIC draw-out devices 100 and FIR filter 200, and interpolater 500 is connected
Between the delivery outlet of FIR filter 200 and the input of CIC interpolating apparatus 300.
In the present embodiment, withdrawal device can be used to carry out assisted codirectional CIC draw-out devices, general withdrawal device is directly to take out
Mode is taken to be extracted, it is less to take resource, can be with taking out when causing inadequate resource using parallel C IC draw-out devices completely
Device is taken to substitute part parallel CIC draw-out devices;Interpolater can be used to carry out assisted codirectional CIC interpolating apparatus, general interpolater is
Row interpolation is entered in Direct interpolation mode, occupancy resource is less, when causing inadequate resource using parallel C IC interpolating apparatus completely,
Part parallel CIC interpolating apparatus can be substituted with interpolater, rationally to utilize system resource.
Optionally, withdrawal device 400 possesses Pi1Individual input, Pc1Individual output end;Interpolater 500 possesses Pc2Individual input, Pi2
Individual output end.
In one of the embodiments, as shown in Figure 10, digital filtering equipment also includes host computer 600 and coefficient register
Configure EBI 700;
Host computer 600 by coefficient register configure EBI 700 respectively with CIC draw-out devices 100, FIR filter
200th, CIC interpolating apparatus 300 is connected.
In the present embodiment, host computer configures EBI by coefficient register and CIC draw-out devices, FIR can be filtered
Ripple device, CIC interpolating apparatus enter row coefficient setting, extraction coefficient, the filter factor of FIR filter, the CIC of such as CIC draw-out devices
Interpolation coefficient of interpolating apparatus etc., Reasonable adjustment, balance numeral are carried out to CIC draw-out devices, FIR filter, CIC interpolating apparatus
The performance and system resource of filter apparatus so that digital filtering equipment is more flexibly practical.
In one of the embodiments, as shown in figure 11, host computer 600 configures EBI 700 by coefficient register
Also it is connected respectively with withdrawal device 400, interpolater 500.
In the present embodiment, host computer configures EBI by coefficient register to enter to withdrawal device, interpolater
Row coefficient is set, extraction coefficient, the interpolation coefficient of interpolater of such as withdrawal device, and Reasonable adjustment is carried out to withdrawal device, interpolater,
Balance the performance and system resource of digital filtering equipment so that digital filtering equipment is more flexibly practical.
In a specific embodiment, digital filtering equipment can be realized in FPGA, or be designed specifically for filtering
Asic chip, available in the application for needing the adjustable high speed Real-Time Filtering of any cut-off frequency, in such as oscillograph instrument.
The position of digital filtering equipment in systems as shown in figure 12, simulation process of the signal by prime, ADC (moduluses
Converter) etc. processing after be input in digital filtering equipment, because the structure of digital filtering equipment can make input and output data
Amount is constant, so prime ADC can bypass digital filter apparatus and be directly connected to rear class processing, without influenceing the treated of rear class
Journey.It can use digital filtering equipment to be handled again after filtering, can also directly handle ADC output datas, without influenceing
Rear class processing.The use of digital filtering equipment and bypass circuit can be selected by selector, the selection work(of selector
It can be controlled by the enable signal of Enable Pin.
Digital filtering equipment chief component is " Digital Down Convert " device, " multiple parallel FIR " wave filters, " numerically
Frequency conversion " device and " coefficient register configuration " EBI.Wherein " Digital Down Convert " device is filled by " multistage parallel CIC extractions "
Put and " withdrawal device " device composition;" Digital Up Convert " device is by " multistage parallel CIC interpolation " device and " interpolater " device group
Into;" coefficient register configuration " EBI can be the bus on chip interfaces such as AXI-Lite, Wishbone or Avalon-MM,
With coefficient register configure the cut-off frequency real time modifying that the host computer that be connected of EBI can be as needed extract, interpolation again
Number and FIR filter coefficient." multistage parallel CIC extractions " device can be only included in " Digital Down Convert " device, " is numerically become
Frequently " multistage parallel CIC extractions " device can be only included in device ", " withdrawal device " device and " interpolater " device can not be set
Put.
Digital filtering equipment by " Digital Down Convert " device by input data down coversion, i.e., by the sampling frequency of input data
Rate fiAdjusting " (can be by the way that " coefficient register configures bus and connect in the filter range that multiple parallel FIR " wave filters satisfaction is required
Mouthful " configuration " filter coefficient " changing " multiple parallel FIR " sample frequency);Again by the way that " coefficient register configures bus and connect
Mouthful " the specific filter cutoff frequency (cut-off frequencies of data after down coversion) of configuration " filter coefficient " regulation;Finally by " number
Word up-conversion " device will pass through down coversion and filtered data are restored to original sample frequency fi;So by this three
Individual device coordinates the filter effect that any cut-off frequency can be achieved.Cut-off frequency if desired for filtering is fbd, and by configuring FIR
The cut-off frequency that filter coefficient obtains FIR filter can be fbf, so by the extraction times of whole " Digital Down Convert " device
Number RlIt is configured to that (to make output identical with input data amount, the interpolation multiple of " Digital Up Convert " part will also be configured to identical
Value):
Signal frequency is down-converted to " in the sample frequency that multiple parallel FIR " wave filters are supported, thus using cut-off
Frequency is fbfWave filter realize cut-off frequency for fbdFiltering, certain " Digital Up Convert " device needs correspondence configuration will
Again up-conversion returns original frequency to signal.
Prime " multistage parallel CIC extractions " device can reduce general serial CIC and extract what is caused sample rate not enough and occur
Spectral aliasing, makes parallelism wave filter reach better performance.Because CIC transmission functions can be expressed as:
Wherein N is cascade coefficient, and M is delay factor, and R is extraction or interpolation factor, and z represents Z-shaped conversion signal.By upper public affairs
Formula is understood:
That is it is (RM) that CIC, which extracts gain,N, then it is y to export maximumout:
Wherein BinFor CIC input datas bit wide (binary system), CIC is calculated data and not spill over, then be used in CIC devices
Data bit width BoutFor:
Bout=Nlog2(RM)+Bin
Delay factor M generally takes fixed value 1, and cascading the coefficient that coefficient N is CIC draw-out devices, (not parallel C IC extracts dress
Put the number of middle single-stage parallel C IC draw-out devices), 1~3 is generally selected as needed, and decimation factor R usually requires reality
Existing larger extracting multiple, the data bit width B when decimation factor R is largeroutIt is larger, because CIC is realized using parallel pipeline structure
(streamline memory cell is more), so resource is taken when data bit width increases in CIC increases more, therefore used here as multistage
Parallel C IC is extracted, and the extracting multiple of " multistage parallel CIC extractions " is the product of extracting multiples at different levels, can be extracted with less
The factor realizes larger extracting multiple, and the resource taken is less.
, will the bypass of " withdrawal device " device if can be without using " withdrawal device " in the case where resource is enough;Certain property worked as
It can reach requirement and can use during less resource and take resource less " withdrawal device " and (can directly extract mode to be simple
Realize) part " multistage parallel CIC extractions " device is replaced, because CIC has the characteristic of low pass, which reduce rear class " withdrawal device " and lead
The spectral aliasing of cause.If ADC is 2G sample rates, and prime " simulation process " is by signal transacting to 100M bandwidth, due to here
Sample rate is much larger than signal bandwidth, and sample rate is directly reduced so may be used herein and take resource less " withdrawal device ", and
Filtering performance will not be reduced completely.
By " needing to return the data of down coversion again up-conversion into original frequency after multiple parallel FIR " devices, here
Interpolation also coordinates regulation resource and performance, " multistage parallel CIC interpolation " dress with " multistage parallel CIC interpolation " device and " interpolater "
Put as the anti-mirror image filtering after interpolation, also using the structure of plural serial stage, to save resource, (reason is with " CIC extractions " portion
Point).Interpolating portion is divided into slotting 0 value mode interpolation in " multistage parallel CIC interpolation " device;" interpolater " can be used to close on to replicate and inserted
Other interpolation methods for meeting design requirement such as value, linear interpolation.
" multistage parallel CIC extractions " device, " multistage parallel CIC interpolation " device and withdrawal device, interpolater coordinate adjustment to need
The performance and the balance of resource wanted, can make digital filtering equipment more practicality.
" multistage parallel CIC extractions " apparatus structure is as shown in Fig. 2 by Ncic1Individual " single-stage parallel C IC extractions " apparatus structure
Series connection is realized.
Wherein " single-stage parallel C IC extractions " apparatus structure is as shown in figure 3, by Ni1Individual " parallel integration " module, " take out parallel
Take " module, Nc1Individual " dredging shape parallel " module and " Gain tuning " module are constituted.Increasing due to parallel C IC abstraction modules to data
Benefit is (RM)N, so each " single-stage parallel C IC extraction " device can carry out Gain tuning (divided by gain (RM) before exportingN) defeated afterwards
Go out.
Wherein (square frame that " D " is marked in figure is storage delay register cell to " integration parallel " module, originally as shown in Figure 4
Adder can postpone a clock cycle output after two values are added in figure), include the matrix of device of a p rows p row
[Ai,j], " parallel integration " modular concurrent inputs p data, due to p data inputting simultaneously between and each input data all
There is precedence relationship, result of calculation needs to complete after the completion of result before and (integrate the data before structure will add up), institute
With not simple multiple single stage integration structures it is in parallel it is achieved that but handled by pipeline system, first with streamline
Mode is added up values at different levels, and present clock is obtained with upper clock cycle result phase Calais again before the output of afterbody streamline
Cycle result.Last data (data p-1) is present clock period input p in p data of upper clock periodical input
The previous data of first data (data 0) in individual data, need to wait p data of upper clock periodical input in theory
It could carry out present clock period streamline after the completion of calculating, and structure used in the utility model, first will with pipeline organization
Present clock period input data is at different levels cumulative, when the row of pth -1 present clock period inputs at different levels added up, upper clock week
Phase result is just completed in pth column count, and the accumulated value of present clock period input just can be with upper clock computation of Period
As a result the result of last in adds up, and obtains present clock period result integrated value.
" parallel dredge shape " module as shown in Figure 5 (when delay factor M takes 1), upper clock periodical input data last
Individual data (data q-1) are the previous data for first data (data 0) that present clock period is inputted, so needs ought
Each data of preceding clock cycle input and one clock cycle of last data delay of upper clock periodical input are laggard
Row makees difference processing, obtains the result of present clock period.
Similar " multistage parallel CIC extractions " apparatus structure, " multistage parallel CIC interpolation " apparatus structure as shown in fig. 6, by
Ncic2Individual " single-stage parallel C IC interpolation " device series connection is realized.
Wherein " single-stage parallel C IC interpolation " apparatus structure is as shown in fig. 7, by Nc2It is individual " parallel dredge shape " module, " parallel to insert
Value " module, Ni2Individual " parallel integration " module and " Gain tuning " module are constituted.Increasing due to parallel C IC interpolating modules to data
Benefit is M (RM)N-1(because what interpolation was inserted is worth for 0, so gain is small R times compared with paralleling abstracting module gain), so each " single-stage
Gain tuning can be carried out before the output of parallel C IC interpolation " device (i.e. divided by gain M (RM)N-1) export afterwards.
Here " dredging shape parallel " module and " parallel integration " module is similar in " parallel C IC extractions ";It is " parallel to insert
Value " module is worth mode to insert 0.
" structures of multiple parallel FIR " wave filters is as shown in figure 8, main by " data distribution delay chain " module and " single-stage is simultaneously
Row FIR " wave filter groups into.
The a length of N+P of shift delay chain of " data distribution delay chain " modulec2(N is the parallel FIR filter of single-stage to -1 data
Exponent number), each clock cycle from data distribution delay chain the 1st to Pc1Individual storage delay register inputs M data, simultaneously
P is exported in sliding window modec2Individual data group, is separately input to Pc2It is individual " in the parallel FIR " wave filters of single-stage by filter based on
Calculate.
Series connection series N wherein in parallel C IC draw-out devicescic1With the series connection series N in parallel C IC interpolating apparatuscic2
For the integer more than 0, and Ncic1And Ncic2Value can be with identical, can also be different;Parallel integration module and parallel thin shape module and
Line number p and q, serial number NiAnd NcIt is the integer more than 0, and in parallel C IC draw-out devices and parallel C IC interpolating apparatus two
The corresponding value of module can be with identical, can also be different;Exponent number N in the parallel FIR filter of single-stage can be the integer more than 0.
" the parallel each system clock cycle of FIR " wave filters of single-stage can export a filter result, Pc2It is individual that " single-stage is parallel
FIR " wave filters can export P in each clock cyclec2Individual filter result (P in figurec2Represent " multiple parallel FIR " wave filters and
Line number)." the parallel FIR " wave filters of single-stage can be the FIR filter of any parallel organization, such as conventional even-order, coefficient pair
" the parallel FIR " filter constructions of single-stage can be called structure as shown in Figure 9 (but being not limited to this structure).The FIR filter is passed
Delivery function is:
The parallel FIR filter exponent number of single-stage is N in Fig. 9 (N is even number);H (0)~h (N/2-1) is that filter coefficient (can
Configured by " coefficient register configuration " EBI);Square frame below figure is multi-accumulator structure, is device tree side with additive here
Formula adds up, and is realized using pipeline system;It can be multistage that " adder ", " multiplier " in requirement, figure can be met for sequential
Streamline is realized;The square frame for owning " D " mark in figure is storage delay register cell, is " data distribution in structure is realized
Storage delay register cell in delay chain " module, i.e., the structure of upper broken line inframe is " data distribution delay chain " in figure
In one " data group " structure, Pc2It is individual that " the parallel FIR " wave filters of single-stage use the P in " data distribution delay chain " successivelyc2
Individual " data group ", every group of N number of storage delay register cell due to being reused using sliding window mode, then uses N altogether
+Pc2- 1 storage delay register cell, i.e. " each data storage delay deposit in top in the parallel FIR " filter graph architectures of single-stage
The data group that the output correspondence of device is exported from " data distribution delay chain ".
It is similar, odd-order " parallel FIR " the filter constructions such as Figure 10 (N is odd number) of single-stage, due to for odd-order, wherein
Between single order be directly entered multiplier (because a data need not be added, if note adder using pipeline system realize need
Notebook data was delayed after the clock cycle identical with adder streamline and is input to multiplier again) it is multiplied with wave filter coefficient of correspondence
After can input summer tree added up.
Accumulator in above structure can be used to differ in the adder tree that pipeline system is realized, certain practical application
It is set to adder tree mode, can multi input, the flowing water knot of one accumulation result of output as long as meeting each system clock cycle
Structure.
Each module is handled using pipeline system in the utility model, and each system clock cycle can be inputted, exported once
Data, the delay period number D for simply entering data to output data is:
D=D1+Dfir+D2
Wherein D1For " Digital Down Convert " module delays, D2It is delayed for " Digital Up Convert " module, DfirFor " multiple parallel FIR "
Filter delay:
D1=Ncic1×(Ni1×P1+Dr1+Nc1)+Ddr1
D2=Ncic2×(Ni2×P2+Dr2+Nc2)+Ddr2
Wherein Ncic1For the series of " single-stage parallel C IC extractions " device in " multistage parallel CIC extractions " device;Ni1For " list
The series of parallel integration module in level parallel C IC extractions " device;P1For parallel integration mould in " single-stage parallel C IC extractions " device
Block and line number;Dr1For the delay number of paralleling abstracting module in " single-stage parallel C IC extractions " device;Nc1For " single-stage parallel C IC
The series of parallel thin shape module in extraction " device;Ddr1For the delay number of withdrawal device in " Digital Down Convert ".DmultFor " multiple parallel
The pipeline series of multiplier in FIR ";NfirFor the exponent number of FIR filter.Ncic2For in " multistage parallel CIC interpolation " device
The series of " single-stage parallel C IC interpolation " device;Ni2For the series of parallel integration module in " single-stage parallel C IC interpolation " device;P2
For in " single-stage parallel C IC interpolation " device parallel integration module and line number;Dr2To be parallel in " single-stage parallel C IC interpolation " device
The delay number of interpolating module;Nc2For the series of parallel thin shape module in " single-stage parallel C IC interpolation " device;Ddr2" numerically to become
Frequently the delay number of interpolater in ".Adder (same to subtracter) in above formula is all thought of as using level production line realization, this
In the unit of delay that calculates be system clock cycle.
And the wave filter of similar structures is realized using common MCU (microcontroller), its delay for calculating a result is estimated
It is:
Because MCU modes each clock cycle can only perform an instruction (only considering monokaryon MCU) DmcuOne in delay time
It is straight to be busy with calculating, it is impossible to receive new data, and the utility model based on FPGA, even if being input to output has one section of delay D,
But due to streamline implementation, new data and processing, i.e. data can be continued to during this delay can be defeated in real time
Enter processing.
The utility model uses the structure of CIC extractions+withdrawal device+FIR filter+CIC interpolation+interpolater.By extracting
Come down coversion, FIR filter filtering after again interpolation carry out up-conversion, realize that any cut-off frequency is adjustable with this;Extracted by prime
FIR filter exponent number can be reduced by reducing progress FIR filtering after sample frequency;Input data can also be made by first extracting interpolation structure
Points it is identical with output data points, so make the utility model in the design whenever with or without (bypass) without influence
Prime or rear class processing;Multistage CIC is extracted and multistage CIC interpolation can use fewer resource to realize larger extraction and interpolation again
Number;Prime can reduce the spectral aliasing caused due to down coversion using CIC draw-out devices, and rear class uses CIC interpolating apparatus
Filtered as anti-mirror image;Extracted by CIC, interpolation and withdrawal device, interpolater coordinate the structure of adjustment, can flexible modulation resource with
The balance of performance, more practicality.
Each module is all Parallel Implementation in above structure:There is the implementation of parallel C IC structures, especially there is its integral part
Realization, the related parallel computation of data before and after cleverly being realized using pipeline organization;Multiple parallel FIR structures, use slip
Window mode distributes data, the use of less storage organization is that multiple parallel wave filter can be achieved.All modules use Parallel Implementation, can
Realize that (can just be filtered without data are first stored) filters at a high speed, in real time.
Each technical characteristic of embodiment described above can be combined arbitrarily, to make description succinct, not to above-mentioned reality
Apply all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, the scope of this specification record is all considered to be.
Embodiment described above only expresses several embodiments of the present utility model, and it describes more specific and detailed,
But therefore it can not be interpreted as the limitation to utility model patent scope.It should be pointed out that for the common skill of this area
For art personnel, without departing from the concept of the premise utility, various modifications and improvements can be made, these are belonged to
Protection domain of the present utility model.Therefore, the protection domain of the utility model patent should be determined by the appended claims.