Method for reducing the peak factor of a multichannel emission by adaptive and intelligent clipping/filtering

Abstract

A method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels, the signal using a plurality of channels in the band comprises: a step of clipping the signal, a step of subtracting the clipped signal from the signal, so as to obtain a peak signal, a step of filtering the peak signal with the aid of a multichannel filter configured to comply with a predetermined spectral mask for each of the channels used by the signal, and a step of subtracting the filtered peak signal from the signal. A device for emitting a multichannel signal implementing the method for reducing the peak factor is also provided.

Claims

1. A method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels, the signal using a plurality of channels in the band, comprising: a step of clipping the said signal, a step of subtracting the said clipped signal from the said signal, so as to obtain a peak signal, a step of filtering the said peak signal with the aid of a multichannel filter configured to comply with a predetermined spectral mask for each of the channels used by the signal, and a step of subtracting the filtered peak signal from the signal.

2. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 1, wherein the spectral mask of the said multichannel filter is predetermined for each of the channels as a function of a maximum admissible power level in sub-bands of this channel and of its adjacent channels.

3. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 1, wherein at least one channel of the said frequency band is not used by the signal.

4. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 2, wherein the said multichannel filter is furthermore adapted so as not to reject at least one channel, not used by the said signal, of the frequency band.

5. The method for reducing the peak factor of a signal in a frequency band comprising several channels according to claim 4, wherein the at least one channel not rejected by the multichannel filter is selected from among the channels of the frequency band, not used by the said signal, having the lowest signal-to-noise ratio.

6. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 1, wherein the step of clipping the signal is a deep clipping.

7. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 1, wherein the said steps of clipping the signal, of obtaining a peak signal, of filtering the peak signal, and of subtracting the filtered peak signal from the signal are carried out iteratively.

8. The method for reducing the peak factor of a signal transmitted in a frequency band comprising several channels according to claim 1, wherein the said step of filtering the signal of peaks uses a multichannel filter embodied on the basis of a prototype filter associated with a spectral mask, the said multichannel filter being obtained by the summation of the said prototype filter shifted in frequency for each of the channels.

9. The method for reducing the peak factor of a signal in a frequency band comprising several channels according to claim 1, wherein the steps are carried out on a baseband digital signal, at modem output.

10. A device for emitting a signal in a frequency band comprising several channels, the said signal using a plurality of channels in the band, comprising: a module for clipping the signal, a module for calculating a peak signal by subtracting a clipped signal from the said signal, a module for filtering a peak signal with the aid of a multichannel filter configured to comply with a predetermined spectral mask for each of the channels used by the said signal, a module for subtracting a filtered peak signal from the said signal.

11. The device for emitting a signal in a frequency band comprising several channels according to claim 10, wherein the said signal is transmitted in the High-Frequency band.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be better understood and other characteristics and advantages will be better apparent on reading the nonlimiting description which follows, and by virtue of the appended figures among which:

(2) FIG. 1 presents the steps implemented by the method according to an embodiment of the invention,

(3) FIG. 2 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention, in which a single carrier is represented,

(4) FIG. 3 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention, in which the multichannel signal is transmitted in five channels,

(5) FIG. 4a presents an example of the impulse response of a multichannel filter used in the method according to the invention,

(6) FIG. 4b presents an example of the frequency response of a multichannel filter used in the method according to the invention,

(7) FIG. 5 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention when it is applied to a multichannel signal comprising five channels for the transmission of the useful signal and two channels dedicated to reducing the peak factor,

(8) FIG. 6 illustrates the difference between the deep clipping mechanism and a traditional clipping,

(9) FIG. 7 gives the performance of the method for reducing the peak factor according to the invention in one of its embodiments,

(10) FIG. 8 presents the method according to an iterative embodiment of the invention.

DETAILED DESCRIPTION

(11) The invention therefore applies to a signal composed of several channels, each of them having a bandwidth that may be identical or different, the various channels being able to be allotted to various transmissions (such as for example in the case of a signal emitted from a GSM base station), or to one and the same transmission.

(12) The signal is emitted by an emission item of equipment comprising a modem intended to shape the signal, and of a radio chain comprising elements making it possible to transpose the signal on carrier frequency and at least one power amplifier.

(13) The signal emitted has a bandwidth corresponding to the width of the emission radio filter situated downstream of the power amplifier, or, if this filter does not exist, to the distance between the channel at the highest frequency and the channel at the lowest frequency of the transmission.

(14) In the case of HF transmissions, the quality of the propagation channels varies in the course of a day. Depending on the time, certain channels are propitious to transmissions, while others exhibit a very degraded signal-to-noise ratio, rendering their use difficult, or indeed impossible.

(15) The allocations of channels to transmissions are therefore carried out on a case by case basis, dynamically, and may evolve over time. Such is the case for example in European patent application EP 2458 770 A1.

(16) The various channels making up the signal to be transmitted are not systematically adjacent. Therefore, the implementation of an efficacious solution for filtering the signal may not be carried out on the basis of a single filter whose passband corresponds to the total useful band of the signal, but on the basis of a filter adapted to the channel allocations.

(17) FIG. 1 presents the steps implemented by the method for reducing the PAPR of a multichannel signal according to an embodiment of the invention.

(18) The method comprises a step 102 of clipping the multichannel signal 101.

(19) During this clipping step, the power of the signal is compared with a threshold. When it is greater than this threshold, the value of the signal is adjusted in such a way that the power of the signal is less than or equal to the value of the threshold. When the power is less than or equal to the threshold, the signal is not modified.

(20) The value of the threshold is positioned as a function of the target PAPR sought by the method.

(21) A second step 103 consists in generating a peak signal by computing the difference between the signal to be transmitted and the clipped signal produced during step 102. This peak signal corresponds to the signal part that was attenuated during the clipping step. Outside of the temporal intervals associated with the peaks, the peaks signal is zero, thereby making it possible to reduce the complexity of calculation of the filtering step.

(22) Working on a peak signal rather than on the clipped signal during the filtering step, also makes it possible to ensure better integrity of the useful signal in the useful channels.

(23) A third step 104 consists in filtering the peak signal, with the aid of a multichannel filter adapted to the various channels used by the signal. The frequency response of this multichannel filter is calculated so as to attenuate the spectrum of the peak signal, in a manner differentiated according to frequency, so that its spectral occupancy, outside of the band of the useful channels, is controlled, not necessarily zero but below a spectral emission mask.

(24) Finally, a step 105 of subtracting the filtered peak signal from the input signal makes it possible to obtain an output signal 106 whose peak factor is less than the initial signal 101, and whose mask is adapted to an emission mask.

(25) FIG. 2 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention. In FIG. 2, observation is limited to a single channel.

(26) In the example illustrated by FIG. 2, the assumption is made that the spectral emission mask is hardly constraining on the first adjacent channel, but that it becomes very demanding from the second adjacent channel onwards. Let us consider for example that the level required in the first adjacent channel must be situated at at least 40 dB with respect to the power of the channel considered, while it must be 70 dB in the second adjacent channel. Curve 201, alternating dots and dashes, represents the spectrum of the initial signal on the channel. The solid curve 202 represents the spectrum of a signal which has been clipped to a value close to the target PAPR. The spectrum of this signal requires hardly any filtering on the first adjacent channel (only a few dB), but requires a strong attenuation in order to fit the mask on the second adjacent channel (about 30 dB). The frequency response 203, shown dashed, of the filter proposed in this example is therefore rather more gentle on the first adjacent channel and strong from the second adjacent channel onwards, in such a way that the spectrum of the signal 204, shown dotted, after applying the method according to the invention, complies with the emission mask, while exhibiting a reduced peak factor with respect to the initial signal.

(27) The filtering step being applied solely to the peak signal, the constraints of embodiment of the filter pertain mainly to the out-of-useful-band rejection, thereby making it possible to use a filter whose frequency response is not necessarily flat in the useful band of the signal, without impairing the integrity of the useful signal. Thus, the complexity of the filter (equivalent to the length of its impulse response in the case of a finite impulse response filter) is reduced.

(28) FIG. 3 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention, in which the multichannel signal is transmitted on five channels.

(29) In FIG. 3, the solid curve 301 represents the spectrum of the initial signal, which uses five channels 311, 312, 313, 314 and 315. The channels may or may not be adjacent.

(30) The dotted curve 302 represents the signal after the clipping step. The dashed curve 303 represents the signal after applying the method according to the invention for reducing the peak factor.

(31) In the method according to the invention, the filter adapted to the signal is not a simple low-pass filter whose passband comprises the whole of the useful band of the signal, but a multichannel filter precisely adjusted to the channels used by the signal. When the channels used by the signal change, the multichannel filter must be modified to adapt to the new channels used.

(32) In contradistinction to the use of one or more low-pass filters to filter the peak signal, the objective of the filter or filters being to be passing in the band of the useful signal and blocking outside of this band so as to reject the whole of the clipping noise lying outside the useful band, the multichannel filter used in the invention follows a predetermined mask on each of the channels used.

(33) Such a filter makes it possible to:

(34) comply with a spectral emission mask for each of the channels used, and not a spectral mask for the whole of the multichannel signal,

(35) control the intermodulation products corresponding to the peak signal,

(36) tolerate spectral regrowth, in a controlled proportion, in the channels used,

(37) tolerate spectral regrowth, in a controlled proportion, in the unused channels positioned in the band of the multichannel signal.

(38) In FIG. 3, the multichannel filter therefore follows a spectral mask predetermined by the emission constraints on the adjacent channels inside the band of the multicarrier signal, and is configured so as not to tolerate noise regrowths in the channels not used by the signal (for example the channels situated between the channel 311 and the channel 312). This mask may be identical or different for each of the channels.

(39) The multichannel filter used by the method according to the invention to filter the peak signal can be generated simply by dimensioning firstly a prototype filter with respect to a single baseband channel, as a function of a maximum admissible power level in sub-bands of the channel and adjacent channels.

(40) The type of impulse response used to generate the filter influences the performance of the reduction in the peak factor, but not the method in itself. All types of impulse responses can be envisaged, provided that the filter complies with the rejection constraints given by the desired mask of the signal.

(41) On the basis of this prototype filter whose impulse response is denoted h(n), a multichannel filter h.sub.Total(n) is generated in an agile manner, so as to cover the set of channels of the signal, according to the formula:

(42) $\begin{array}{cc}{h}_{\mathrm{Total}}\left(n\right)=\underset{k=0}{\overset{K-1}{.Math.}}h\left(n\right).Math.\exp \left[j.Math.2.Math.F_k.Math.\frac{n}{F_{\mathrm{smp}}}\right],& \left(1\right)\end{array}$

(43) with K the number of channels used by the signal, F.sub.k the frequency of the K channels, and F.sub.smp the sampling frequency.

(44) FIG. 4a presents an example of the impulse response of a multichannel filter used in the method according to the invention. The dashed curve 401 is the impulse response of the prototype filter used to generate the multichannel filter. The prototype filter of the example uses an impulse response of Blackman-Harris type. The coefficients of this filter are real.

(45) The solid curve 402 and the dotted curve 403 represent the real part and the imaginary part of the impulse response of the multichannel filter used in the method according to the invention.

(46) The multichannel filter thus embodied comprises an identical number of coefficients to the number of coefficients of the prototype filter, independently of the number of channels. Its impulse response is generally complex.

(47) FIG. 4b presents an example of the frequency response of the same multichannel filter, where the solid curve 411 designates the frequency response of the prototype filter, and the dotted curve 412 the frequency response of the multichannel filter.

(48) In the case where the channels do not all have the same width, the multichannel filter can be embodied in an identical way by using several prototype filters of different widths.

(49) FIG. 5 is a frequency representation of the signal at various steps of the method according to an embodiment of the invention, when it is applied to a multichannel signal comprising five channels for the useful signal and two channels dedicated to reducing the peak factor.

(50) The invention proposes to use the channels that are not used by the transmission since they exhibit a poor signal-to-noise ratio, to transmit data intended to reduce the peak factor of the signal. These channels are not used to transmit data, and are therefore ignored by the reception item of equipment. Such a principle can be taken into account in the protocols for adapting the waveform to the propagation environment.

(51) The only variation to the embodiment of the invention presented in FIG. 3 resides in the embodiment of the multichannel filter, the latter then being calculated so as not to attenuate the frequencies of the additional channels dedicated to reducing the peak factor in addition to the channels used by the signal.

(52) In FIG. 5, the solid curve 501 represents the initial signal, which uses five channels 511, 512, 513, 514 and 515.

(53) In this example, two additional channels, 521 and 522, are used to reduce the PAPR of the signal. The number of additional channels can vary, and is limited only by the frequency resource. The more the number of additional channels increases, the better the performance of the method for reducing the peak factor.

(54) The dotted curve 502 represents the signal subsequent to the clipping step. The dashed curve 503 represents the signal after applying the method according to the invention for reducing the peak factor.

(55) The use of a multichannel filter, designed so as not to attenuate the frequencies of the channels 521 and 522 during the filtering of the peak signal 502, causes a spectrum regrowth in the channels 521 and 522, the regrowth being compatible with the spectral mask constraints.

(56) The generation of a multichannel filter such as this can be done in an identical manner to the procedure exhibited in equation (1), taking into account the additional channels 521 and 522.

(57) The method according to the invention then uses channels that are not allocated to the transmission of the useful signal, such as for example the carriers whose signal-to-noise ratio hinders the use for the transmission of data. The multichannel filter adapted to each of the channels that are used makes it possible to select the channels in or outside of the useful band of the signal. The use of a multichannel filter generated on the basis of a predetermined spectral mask calculated for each channel, whether these be channels used by the signal, channels not used by the signal but used to favour the reduction in the peak factor, or unused channels, makes it possible to guarantee compliance with regulatory signal emission constraints, be it globally or channel by channel.

(58) As regards the hardware architecture of the emitter, the use of a single multichannel filter rather than of N filters each associated with a channel offers great flexibility of adaptation to the number of channels used, to their position and to the use of channels dedicated to reducing the peak factor. Such adaptability in an architecture where the filtering is carried out for each of the channels makes it necessary to provide a processing chain per channel in the band of the emitter, thus increasing its complexity, its cost and its consumption.

(59) So as to improve the performance of the method, it is possible to use, during the signal clipping step, a deep clipping mechanism.

(60) In contradistinction to traditional clipping, which limits the output power and fixes it at the value of a threshold for any power greater than this threshold, deep clipping consists in decreasing the output power for any signal exceeding a threshold. Deep clipping therefore compresses the peak signal to a lower value than the PAPR sought.

(61) FIG. 6 describes the deep clipping mechanism, comparing it with the traditional clipping.

(62) Curve 601 represents the evolution over time of the power of the useful signal to be transmitted. This power is compared with a threshold, the role of the clipping mechanism being to limit the power of the signal to this threshold.

(63) Curve 611 represents the power of the signal, after a traditional clipping step 610. The signal whose power exceeds the threshold has been modified in such a way that the power of the clipped signal is now limited to the value of the threshold.

(64) Curve 612 represents the power of the signal, after a step 630 of filtering the signal clipped by the traditional clipping procedure. On account of its nature, the filtering causes a upswing in the peaks of the power of the clipped and filtered signal to values greater than the threshold.

(65) The deep clipping mechanism implements the following relation:

(66) $\begin{array}{cc}{\mathrm{out}}_{\mathrm{DeepClipping}}\left(n\right)=p\left(n\right).Math.\exp \left[j.Math.\left(n\right)\right],\mathrm{with}\text{:}& \left(2\right)\\ p\left(n\right)=\{\begin{array}{ccc}.Math.{\mathrm{in}}_{\mathrm{DeepClipping}}\left(n\right).Math.& \mathrm{if}& .Math.{\mathrm{in}}_{\mathrm{DeepClipping}}\left(n\right).Math.\mathrm{Thresh}\\ \mathrm{Thresh}+a_0.Math.\left(.Math.{\mathrm{in}}_{\mathrm{DeepClipping}}\left(n\right).Math.-\mathrm{Thresh}\right)& \mathrm{if}& .Math.{\mathrm{in}}_{\mathrm{DeepClipping}}\left(n\right).Math.>\mathrm{Thresh}\end{array},\mathrm{and}& \left(3\right)\\ \left(n\right)=\mathrm{Arg}\left[{\mathrm{in}}_{\mathrm{DeepClipping}}\left(n\right)\right],\mathrm{where}\text{:}& \left(4\right)\end{array}$ in.sub.DeepClipping (n) is the input signal of the deep clipping algorithm, out.sub.DeepClipping (n) is the output signal of the deep clipping algorithm, a.sub.0 is the slope of the deep clipping algorithm, Thresh is the threshold of the deep clipping algorithm, Arg[x] is the function giving the phase of the signal x.

(67) Curve 621 represents the power of the signal, after a deep clipping step 620. The signal whose power exceeds the threshold has been modified, in such a way that the power of the clipped signal is less than the value of the threshold by a level inversely proportional to the power level by which the initial signal is exceeded.

(68) Curve 622 represents the power of the signal, after a step 630 of filtering the signal clipped by the deep clipping procedure. The regrowths observed in the peaks of the power of the signal are much lower than the regrowths observed when using the traditional clipping procedure.

(69) FIG. 7 gives the performance of the method for reducing the peak factor according to the invention in one of its embodiments.

(70) In FIG. 7, the solid curve 701 gives the distribution function for the peak factor of a signal transmitted on 5 channels.

(71) The dotted curve 702 gives the distribution function for the peak factor of the same signal, after applying the method according to an embodiment of the invention. In this specific case and with respect to constraints of the spectral mask, the gain in peak factor is about 1.5 dB.

(72) The dashed curve 703 gives the distribution function for the peak factor of the signal, after applying the method according to an embodiment of the invention. In this case, two additional channels have been allocated to the transmission, so as to further reduce the peak factor, affording a gain in the PAPR, of the order of an additional 0.5 dB.

(73) In another embodiment of the invention, described in FIG. 8, the steps of clipping 102, of determining a peak signal 103, of filtering the peak signal 104, and of subtracting the filtered peak signal from the initial signal 105, which are described in the method, are applied to the signal 801 originating from the modem. The resulting signal 802 is looped back, and the various steps of the method according to the invention are applied thereto again. Each iteration makes it possible to improve the peak factor gain.

(74) Indeed, as illustrated at 622 in FIG. 2, although the principle of deep clipping, associated with the filtering of the peaks, is effective and limits the regrowths in the peaks, the latter remain unavoidable. It may then be advantageous to repeat these operations one or more times so as to converge towards an output signal 803 reaching a target PAPR.

(75) The iterations make it possible to find an optimal solution of the relation xf(x)=0 in the form x(i+1)=f(x(i)), each iteration making it possible to improve the gain in terms of peak factor reduction, doing so as long as a maximum number of iterations is not reached.

(76) The use of a deep clipping mechanism makes it possible to decrease the number of iterations required for obtaining a given target PAPR.

(77) The method according to the invention applies to a digital signal at modem output and upstream of the emission radio chain, preferably but not limitingly in baseband. The various modules executing the various steps of the method can be code portions recorded in a non-volatile memory, and intended to be executed on a calculation machine such as for example a reprogrammable calculation machine (a processor or a microcontroller for example) or a dedicated calculation machine (for example a set of logic gates such as an FPGA or an ASIC), or any other hardware module.

(78) It can also apply to an analogue signal, since it calls only upon power limiters (such as for example a diode based limiter), analogue filter banks, delay lines and differential summators or amplifiers.

(79) The method is intended to be implemented in a radiocommunications device emitting a signal using several channels. This device can be configured to emit in the HF band, but can also emit in the other frequency bands, and in a propagation environment other than a wireless radio environment, such as for example a wired network or a fibre optic network.

(80) The advantages of the method according to the invention are as follows:

(81) the method is totally transparent to the type of modulation transmitted on each of the channels, it is positioned at the output of the modem and requires only the knowledge of the channels used and of the spectral emission mask associated with each of the transmission channels,

(82) the method offers a good PAPR reduction capacity: the use of deep clipping makes it possible to limit the regeneration of the peaks caused by the filtering step, and can if necessary be associated with an iterative mechanism,

(83) the method makes it possible to effectively reduce the PAPR as a function of the spectral emission mask: the proposed solution makes it possible to adapt exactly to an spectral emission mask and to the channels available so as to optimize the PAPR/out-of-band emission level compromise,

(84) the method makes it possible to occupy the whole set or a subset of channels available for transmission, such as the transmission channels which are not occupied for the transmission of useful data but which are relevant in reducing the PAPR,

(85) its complexity of implementation is limited, this being due in particular to the use of a fairly non-complex multiband filter, whose constraints in the band of each useful channel are weak,

(86) the filtering step is carried out on a peak signal, thereby simplifying the filtering work, and making it possible to avoid degrading the useful signal,

(87) the method can be implemented digitally, on a baseband signal at modem output, or by using analogue hardware components.