Window type peak clipping method
Technical field
The present invention relates to a kind of window type peak clipping method, be fit to be applied in multicarrier system with higher peak-to-average power ratio or the communication system that adopts high-order modulating.
Background technology
A subject matter of multi-carrier modulation and high order modulation is that it is peculiar than higher peak-to-average power ratio (PAPR) to send signal.In order to transmit the signal of these high peak-to-average power ratios undistortedly, linearity to some parts (as power amplifier) requires very high, if the excursion of signal exceeds the range of linearity of device, then can produce tangible inband distortion and out-of-band radiation, thereby cause the rising of the error rate.So the size of the peak-to-average power ratio of signal is to multicarrier system and adopt the communication system performance of high-order modulating that direct influence is arranged.
By top argumentation as can be seen, how to reduce multicarrier system effectively and adopt the peak-to-average power ratio of high-order modulating communication system just to seem very necessary, this will be directly connected to the performance of its system and realize cost.
Existing a kind of despicking method can effectively reduce the peak-to-average power ratio of system, its cardinal principle generates the peak clipping pulse for the filter impulse responses that utilization and input signal have same frequency spectrum characteristic (Spectrum Mask), and the data that surpass the amplitude thresholding in the original signal are carried out peak clipping.Its principle as depicted in figs. 1 and 2.
Last figure is the time-domain signal amplitude curve among Fig. 1, and Ath is the peak amplitude thresholding, and A is detected peak amplitude, and A-Ath is the amplitude difference that exceeds the amplitude thresholding; Figure below is the good peak clipping pulse of structure among Fig. 1.Its operation principle is:
Detect signal peak earlier, with the filter impulse responses displacement, make its peak value corresponding with detected signal peak position, establishing filter impulse response is H (n), utilizes following formula construction to go out the peak clipping pulse:
CP_val(n)=(A-Ath)*e
jθ*H(n),1≤n≤L
Wherein e is the truth of a matter of natural logrithm, and j is the imaginary part of symbol,
θ is the phase place of signal peak, e
J θ=cos θ+jsin θ, n are integer, and 1≤n≤L deducts the peak clipping pulse with primary signal, the visible Fig. 2 of result after structure is good.
By signal waveform after the peak clipping of fine rule representative among Fig. 2, as seen, the peak value that surpasses the amplitude threshold value has been cut below the amplitude threshold value, and different with the clipping algorithm is, peak value under cutting is slick and sly, rather than directly cut away as the clipping algorithm, this is because the spike of filter impulse responses is slick and sly, so the spectral performance of this despicking method is better than the clipping algorithm.
This despicking method advantage is:
1, can reduce the peak-to-average power ratio of input signal effectively;
2, owing to use the filter impulse responses that has with input signal same frequency spectrum characteristic (Spectrum Mask) to carry out peak clipping, therefore the out-of-band noise of introducing is very little;
3, compare with other despicking method, this method is little to EVM (error vector magnitude) performance impact of signal.
But, because this despicking method is to come peak clipping with the filter pulse, shown in figure below among Fig. 1, the pulse of filter impulse responses is not desirable impulse, therefore can produce adjacent point in the peak clipping process and disturb, as exist consecutive points all to surpass the amplitude threshold value, and the situation that phase place is identical, after peak clipping is handled, the peak clipping superimposed pulses, can produce very big distortion to signal, in implementation procedure, will address this problem; In implementation procedure, require in addition to improve treatment effeciency, reduce memory space to save resource.
Summary of the invention
The objective of the invention is at the deficiencies in the prior art, a kind of window type peak clipping method that digit chip programming realizes of being convenient to is provided, signal is carried out effective peak clipping handle, to avoid the situation of the same phase neighbor point distortion that traditional despicking method caused.
For realizing this purpose, the present invention takes following technical scheme:
Window type peak clipping method of the present invention is used for that signal is higher than default amplitude thresholding and partly carries out the peak-to-average power ratio (PAPR) of peak clipping to reduce this signal, comprises the steps:
1) sets the filter of odd-order (L) according to the spectral characteristic of original signal;
2) set have specific odd length (N) memory address as window logic to be used to store the sampled data of original signal;
3) default amplitude thresholding (Ath) as the peak clipping benchmark is stored in each memory address of this window logic with each sampled data serial with original signal of the mode of queue-type data flow;
4) calculating in real time judges that whether the data that are positioned at window logic meta (P) are local peaking's points of window logic, and whether its amplitude is greater than amplitude thresholding (Ath);
5) if above-mentioned steps 4) judgement set up, the data that promptly are positioned at window logic meta (P) are that local peaking's point and its amplitude are greater than amplitude thresholding (Ath), then the peak-peak sampled data of being stored corresponding to the window logic meta with the peak value of the impulse of filter is configured to each sampled data in the window logic is carried out the peak clipping pulse of peak clipping;
6) deduct the peak clipping pulse of having constructed to finish this time peak clipping process with original signal, treat to return step 4) after peak clipping sampled data all flows out, till original signal all flows out this window logic.
Step 6) is unit with the sampling time in the waiting time of returning between the step 4), on length more than or equal to the length of window logic.
Particularly, in the described step 4), adopt following formula construction peak clipping pulse:
CP_val(n)=(A-Ath)*e
jθ*H(n),1≤n≤L
Wherein, A is detected peak amplitude, and Ath is the amplitude thresholding of peak value, and e is the truth of a matter of natural logrithm, and j is the imaginary part of symbol,
θ is the phase place of signal peak, e
J θ=cos θ+jsin θ, the filter impulse pulse that H (n) sets for the spectral characteristic according to original signal, n is an integer, 1≤n≤L.
Disclose according to one embodiment of present invention, above-mentioned steps 3) in, search peak amplitude and be specially greater than the process of the sampled data of this amplitude thresholding: whether the amplitude that is positioned window logic meta (P) and detects the sampled data (data_P) in this memory address is greater than described amplitude thresholding (Ath); If the peak amplitude of current sampled data is less than or equal to this amplitude thresholding and does not just do data processing, otherwise, continue relatively to judge whether this amplitude is maximum in window logic, if not maximum is not then done data processing, if maximum then just is defined as described peak-peak sampled data with the sampled data of this meta.
If one time iteration can not be eliminated the data peaks that all surpass amplitude thresholding (Ath), even window logic by any one group of sampled data of peak clipping in, when existing at least two to have the sampled data of the peak amplitude that is higher than this power threshold, this method can be carried out repeatedly iteration, and iterative process is a repeating step 2) to step 6).The sampled data that the default number of iterations of carrying out of this method, the concrete iterations signal after with last peak clipping does not exist peak amplitude to be higher than this amplitude thresholding is exceeded.
Among the present invention, on the principle, the length of window logic is as the criterion with the exponent number that approaches filter peak clipping pulse, and comparatively suitable scheme is: the peak clipping pulse exponent number L of filter is an odd number, and the length N of described window logic equals L ± 2m, and m is an integer, 0≤m<L/4.
One embodiment of the present of invention further disclose a preferred implementation, that is: the peak clipping pulse exponent number L of this filter is an odd number, and the length N of described window logic equals L.
Compared with prior art, the present invention is suitable for programming realization in the digit chip such as FPGA, by in this type of digit chip, setting the storage area of a finite length as window logic, employing is similar to the linear data structure of fifo queue, makes it have following comparatively outstanding advantage:
At first, sampled data to original signal is carried out the flowing water type linear access, operate with more rational queue-type data structure, make full use of the characteristic of its first in first out, can find peak-peak sampled data apace greater than default amplitude thresholding, thereby can save time, improve treatment effeciency;
Secondly, because in this method, the big I of window logic is set flexibly, value that is adjacent to the progression length of peak clipping pulse of operated by rotary motion, like this, neither can make that an iterative process peak clipping number is little, thereby cause many times iteration to bring extra complexity to system because window logic length is long; Can be not too short because of window logic length yet, thus the stack owing to the peak clipping pulse in an iterative process of the contiguous sampled data with same phase of feasible and peak-peak sampled data causes distorted signals;
Moreover, by rational iteration repeatedly, make that being present in window logic length limits the crest that the interior a plurality of amplitudes of distance are higher than amplitude thresholding (Ath), though can not be by peak clipping in a same iteration, but can be, thereby guarantee that original signal can be by abundant peak clipping to obtain desirable peak-to-average power ratio by peak clipping in follow-up iterative process.
Description of drawings
Fig. 1 is the waveform schematic diagram of signal peak detection with the peak clipping pulse of traditional despicking method;
Fig. 2 is the signal power curve synoptic diagram of despicking method before and after peak clipping of Fig. 1;
Fig. 3 when using window despicking method of the present invention, the logical construction schematic diagram of window logic in the digit chip;
Fig. 4 is the iteration theorem schematic diagram of window despicking method of the present invention.
Embodiment
Followingly describe in conjunction with the accompanying drawings and embodiments:
Window despicking method of the present invention is fit to be integrated in FPGA (FieldProgrammable Gate Array with the form of program; field programmable gate array) in and so on the digit chip; this chip is widely used in the communication system; thereby have representative widely; be used to auxiliary the elaboration in the present embodiment; but; known other of those skilled in the art has the digit chip of programmability, and is developed out the new digital chip with equal capability in the future and should be excluded outside protection scope of the present invention.
In this enforcement, be optimized based on traditional despicking method, thereby its cardinal principle remains utilizes the filter impulse responses that has with input signal same frequency spectrum characteristic (Spectrum Mask) to generate the peak clipping pulse, and the data that surpass amplitude thresholding (Ath) in the original signal are carried out peak clipping.So can consult the method for the known structure peak clipping pulse that Fig. 1 and Fig. 2 disclose, the equation of structure peak clipping pulse is same as the formula that background technology is quoted:
CP_val(n)=(A-Ath)*e
jθ*H(n),1≤n≤L
Wherein, A is detected peak amplitude, and Ath is the amplitude thresholding of peak value, and e is the truth of a matter of natural logrithm, and j is the imaginary part of symbol,
θ is the phase place of signal peak, e
J θ=cos θ+jsin θ, the filter impulse pulse that H (n) sets for the spectral characteristic according to original signal, n is an integer, 1≤n≤L.
Be suitable for said method, the method for setting corresponding filter according to the spectral characteristic of original signal has been well known in the art, thereby not all right giving unnecessary details.And filter is once setting, and its exponent number L that makes up the peak clipping pulse also is set.Exponent number L will be incorporated by reference so that set the length of window logic in step of the present invention.
Use for reference known analog-to-digital sampling principle, the so-called point that surpasses the amplitude thresholding, and each point in the original signal, all the form with the discrete sampling data is present in the physical address of memory in digit chip, the sampled data that each physical address unit corresponding stored one disperses, by order match, just constitute an analog signal figure, as depicted in figs. 1 and 2 to many discrete sampling data.Repeat this principles well-known and help the auxiliary the present invention that understands.
The window logic that defines among the present invention, it is linear storage area of definition in memory, this linearity storage area both can be continuous physically zone, also can be logically continuous zone such as pointer alignment, no matter which kind of mode of employing, it all forms the queue-type data store organisation with first in first out characteristic, the rule visit of the memory address by physically, but the just sampled data of being stored in a certain memory address of access, thereby obtain numerical value more specifically such as its corresponding amplitude, phase place.In addition, the present invention also defines one group of sampled data that disperses in order on the time domain of original signal when storing, its storing process for from an end of this window logic according to the order of sequence serial enter, and flow out one by one according to the order of sequence, to form the flowing water operation at the other end of window logic.
Based on the set content of above-mentioned announcement, the concrete steps of window type peak clipping method of the present invention are as follows:
Step 1: according to the filter of the spectral characteristic of signal design odd-order (L), the impulse response of record filter.
Step 2: set the length N of window logic, the N value conference make the peak value number that reams in the single iteration few, therefore need more iterationses, can increase system complexity; The too small meeting of N value makes and peak clipping superimposed pulses in the peak clipping process causes distorted signals; Therefore the N value is unsuitable excessive also unsuitable too small, can select close with peak clipping pulse progression length L.In the present embodiment, N=L.
And relax special case as one, and the length N of window logic can value N=L ± 2m, and wherein, m is an integer, and 0≤m<L/4.Like this, under the prerequisite of N appropriateness, window logic not only not too greatly but also not too little, can satisfy aforesaid condition.
Step 3: set the value Ath of amplitude thresholding, the variable of establishing concrete sampled data is data, and the sampled data data that all amplitudes are higher than amplitude thresholding Ath all needs by peak clipping.
Then, can begin all sampled datas, in the mode of queue-type data flow, from the end serial input of window logic with original signal, and in the series read-out of the window logic other end, in original signal all sampled datas all flow through according to the order of sequence window logic and be read out till.
Step 4: the operation of this step is only carried out at the local sampling data in the window logic, promptly in N the sampled data that flows in the window logic, search wherein peak amplitude and be higher than the peak-peak sampled data of amplitude thresholding Ath, as shown in Figure 3: the locator data pointer respectively has in meta P both sides in the centre position of window logic (memory address) P
(symbol
Expression rounds downwards) individual sampled data, relatively whether the peak amplitude A among the sampled data data_P that points to of P greater than amplitude thresholding Ath, if A≤Ath then data are not handled continues the relatively more next sampled data that inputs to meta P; If A>Ath, then continue relatively to judge whether amplitude maximum in the sampled data of whole window logic of data_P, if the amplitude of data_P is not maximum in window logic, then data are not handled, continue the relatively more next sampled data that is input to meta P; If data_P has maximum peak amplitude in whole window logic, then: at this moment, peak-peak sampled data with maximum peak amplitude has occupy the meta P of window logic, so can carry out the peak clipping operation to current all sampled datas in the window logic, promptly continues next step.
Step 5: made the peak-peak sampled data in current all sampled datas in the window logic occupy the meta P of window logic in the previous step, in order to realize peak clipping, corresponding to the peak-peak sampled data that window logic meta P is stored, utilize aforementioned known CP_val formula construction to be used for all sampled datas in the window logic are carried out the peak clipping pulse of peak clipping with the peak value of the impulse of filter then.
Step 6: structure good with the interior corresponding peak clipping pulse of current all sampled datas of window logic after, deduct the peak clipping pulse with primary signal, promptly finished based on of the peak clipping operation of data_P point all current sampled datas in the window logic, even after the peak clipping, still have other amplitude to be higher than the sampled data of amplitude thresholding in the window logic, also handle no longer at this moment, but timing N is after the sampling time, treat in the window logic that sampled data all flows after (reading) go out window logic, after follow-up sampled data has also entered window logic in the original signal at this moment, getting back to step 4 just now circulates, certainly, if all sampled datas of original signal all flow out, recirculation step four not then.A described N sampling time shows the concrete waiting time of this step before entering step 4, and its unit is the sampling time, and its length is N, and N is the length of window logic, and this time span guarantees that the sampled data in the window logic can fully be read.Certainly, also be allowed to greater than the situation of N, those skilled in the art all know this accommodation.
Treat that primary signal all flows through behind the window logic, promptly finish the peak clipping first time original signal, still, peak clipping only once sometimes and the sampled data of failing amplitude in the original signal is higher than amplitude thresholding (Ath) all eliminate.For example, if in by the window logic of peak clipping, exist when adjacent or identical peak amplitude is higher than the crest of amplitude thresholding on phase place more than two, at this moment may only there be a crest to be eliminated the anticipation state fully, and other crest is operated only carry out a peak clipping with a collection of sampled data owing to window logic, so still exist.
See also Fig. 4, in order fully to eliminate a plurality of crests that exist in the primary signal, the present invention does further processing to this situation: the window logic that a plurality of length are N promptly is set, when data flow through first window logic, after promptly finishing the iteration first time, enter second window logic, promptly carry out the iteration second time, the setting of its iterations should make does not have peak value to surpass the sampled data of amplitude thresholding Ath in the dateout after the last iteration.
The detailed process of each iteration is step 2 to step 6 and constitutes, and it is pointed out that wherein repeatedly a plurality of window logics that iteration is required, both can distribute according to need in the iteration each time, also can and carry out integrated management in disposable distribution in the first time iteration.
In sum, window type peak clipping method of the present invention, by primary signal is carried out digitized processing, Utilization adds window logic to sampled data wherein and carries out the single peak clipping, and the mode carried out of iteration repeatedly, Make primary signal can and obtain desirable peak-to-average power ratio by peak clipping effectively, introduce very little EVM (mistake And this kind method meets Program Thought, is convenient in the numeral such as FPGA the difference vector amplitude), Programming realizes in the chip, can obtain the effect of high efficiency and time conservation.