Method and device for detecting a peak amplitude for a frequency band

Abstract

A method for detecting a peak amplitude for a frequency band is provided. The method is performed in peak detection unit and comprises obtaining an amplitude sample in addition to two existing amplitude samples, comparing the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and determining, based on the comparison, one as the peak amplitude. A corresponding peak detection unit, a radio device, computer program and computer program products are also disclosed.

Description

[0001] METHOD AND DEVICE FOR DETECTING A PEAK AMPLITUDE FOR A FREQUENCY BAND

[0002] TECHNICAL FIELD

[0003] The technology disclosed herein relates generally to the field of wireless communication and in particular to a method and device for detecting a peak amplitude for a frequency band.

[0004] BACKGROUND

[0005] In wireless communication a high peak-to-average ratio (PAR) is undesirable, but is encountered in some communication methods such as, for instance, in Orthogonal Frequency Division Multiplexing (OFDM). One way to reduce the PAR of a signal that is supplied to a power amplifier in e.g., a user device, is to use Crest Factor Reduction (CFR). The CFR unit limits the peaks in the received signal and thereby improves the efficiency of a power amplifier (PA). The performance of the CFR is very important for the user device since it enables a reduction of the power dissipation that occurs in the PA. Even small improvements of a Peak-to-Average Ratio (PAR) (ratio of the highest value of a quantity to its average value) may have significant impact on the PA.

[0006] The reduction of PAR will by default create distortion of the transmitted signal. A key metric for evaluating Radio Frequency (RF) performance is an Error Vector Magnitude (EVM). The EVM is a measure of signal quality, which in turn is a function of noise, interfering signals, nonlinear distortion, and load of a device from which the signal is transmitted.

[0007] In multi-band configurations it is sometimes efficient to use a separate CFR for each frequency band. In such multi-band configurations, it is important to use the CFRs so that they not only minimize the PAR on each band but also minimize the combined signal’s PAR from all frequency bands, as this is the signal that will be transmitted. The combined signal has traditionally been estimated by adding the absolute value of the signals in each band. This will result in an overestimation of the peaks’ magnitude, without taking the phase information of each signal into account. This is known as a non-coherent mode.

[0008] There are also methods wherein the phase information is taken into account, known as a coherent mode. A main disadvantage of such method is that it will result in insufficient oversampling when the frequency bands are far apart. This lack of oversampling results in an increased risk of missing the maximum peak amplitude as it might fall between samples. An obvious solution for this problem is to increase the sample rate in the entire CFR. However, this leads to heavily increased processing and higher memory requirements in Application Specific Integrated Circuits (ASICs), which in turn results in higher cost and power consumption.

[0009] In view of the above noted difficulties, it is clear that there is a need for improvements in various aspects of multi-band configurations.

[0010] SUMMARY

[0011] A general objective of embodiments herein is to address and improve various aspects of wireless communication means, and the processing of wireless signalling. A particular objective is to improve detection of peaks in multi-band configurations. Another objective is to enable improved cancelling of peak amplitudes, without increasing sampling rates. Still another objective is to fulfil these objectives while also keeping down the number of added hardware pieces. These objectives and others are achieved by the methods, devices, computer programs and computer program products according to the appended independent claims, and by the embodiments according to the dependent claims.

[0012] These objective and others are accomplished by a method and devices for detecting a peak amplitude for a frequency band.

[0013] According to a first aspect there is presented a method for detecting a peak amplitude for a frequency band. The method is performed in a peak detection unit and comprises obtaining an amplitude sample in addition to two existing amplitude samples. Then the amplitude of the obtained amplitude sample is compared to one or both of the two existing amplitude samples, after which one of the amplitude samples is determined, based on the comparison, as being the peak amplitude.

[0014] The method provides an improved way to detect peaks in multi-band configurations, and in particular the peak amplitudes. The method also cancels the peaks without a need for oversampling of a complete Crest Factor Reduction (CFR). Multi-band configurations are becoming increasingly important as users would like to support several bands without installing new radio units for each band. The method provides an improved performance in such multi-band configurations without the typically high costs related to silicon devices. Further, the method reduces the power dissipation that an increased oversampling would require. Besides supporting several bands, the disclosed method may be used, for instance, to double the Instantaneous Bandwidth (IBW) of the CFR by placing two CFRs side-by-side in frequency. In addition to a lower power consumption, the design of the PA and cooling thereof can be made at a lower cost.

[0015] According to a second aspect there is presented a computer program for detecting a peak amplitude for a frequency band. The computer program comprises computer code which, when run on processing circuitry of a device, causes the device to perform a method according to the first aspect.

[0016] According to a third aspect there is presented a computer program product comprising a computer program as above, and a computer readable storage medium on which the computer program is stored.

[0017] According to a fourth aspect there is presented a device, in particular a peak detection unit, for detecting a peak amplitude for a frequency band. The peak detection unit is configured to obtain an amplitude sample in addition to two existing amplitude samples, and to compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples. The peak detection unit is configured to determine, based on the comparison, one of the amplitudes as the peak amplitude. The peak detection unit provides advantages corresponding to the ones mentioned for the method.

[0018] Other objectives, features and advantages of the enclosed embodiments will be apparent from the following detailed disclosure, from the attached dependent claims as well as from the drawings.

[0019] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a / an / the element, apparatus, component, means, module, action, etc." are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, module, action, etc., unless explicitly stated otherwise. The actions of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated.

[0020] BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The inventive concept is now described, by way of example, with reference to the accompanying drawings, in which:

[0022] Fig. 1 illustrates a peak detection unit according to various embodiments.

[0023] Fig. 2 illustrates a crest factor unit in non-coherent peak detection mode.

[0024] Fig. 3 illustrates a coherent mode, with coherent peak detection.

[0025] Figs. 4a and 4b illustrate a comparison between a prior art estimate and an estimate according to the present teachings.

[0026] Fig. 5 illustrates a filter device for estimating the amplitude of a true peak.

[0027] Fig. 6 illustrates an exemplary circuit for even samples stream KPE processing.

[0028] Fig. 7 illustrates an exemplary circuit for odd samples stream KPE processing.

[0029] Fig. 8 illustrates a polyphase implementation.

[0030] Fig. 9 illustrates an intermodulation distortion in an input signal.

[0031] Figs. 10a and 10b illustrate an effect of taking the absolute value on the Feature metric.

[0032] Fig. 11 shows a linear regression with generative machine learning for a 100 MHz carrier.

[0033] Fig. 12 is a flowchart of various embodiments of a method.

[0034] Fig. 13 illustrates a radio unit according to an embodiment.

[0035] Fig. 14 is a schematic diagram showing functional units of a device according to an embodiment. Fig. 15 is a schematic diagram showing functional modules of a device according to an embodiment.

[0036] Fig. 16 shows one example of a computer program product comprising computer readable means according to an embodiment.

[0037] DETAILED DESCRIPTION

[0038] The inventive concept will now be described more fully hereinafter with reference to the accompanying drawings, in which certain embodiments of the inventive concept are shown. This inventive concept may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete, and will fully convey the scope of the inventive concept to those skilled in the art. Like numbers refer to like elements throughout the description. Any action or feature illustrated by dashed lines should be regarded as optional.

[0039] Briefly, a method is provided in various embodiments, in which maximum amplitudes of peaks are detected in a multi-band configuration. The detected and measured peaks are then applied to a peak cancellation method. By using this method, the peaks can be cancelled without the currently required oversampling of a complete CFR. Crest factor is the peak amplitude of the waveform divided by the RMS value of the waveform.

[0040] Fig. 1 illustrates a peak detection unit 10 according to various embodiments. In a dual-band mode, as in in the illustrated case, two frequency bands are handled with a respective (separate) Crest Factor Reduction (CFR) unit 2a, 2b, one for each band, in a respective Multi Carrier Branch (MCB) la, lb. In the non-coherent mode, the sum of the absolute values from the two bands are used for estimating the absolute value (ABS) of the final combined bands sent to a power amplifier. The signal used for the peak detection may be performed on a four times oversampled signal. A Differential Phase Detection (DPD) 3 performs predistortion on the signals from the CFRs 2a, 2b.

[0041] Fig. 2 illustrates non-coherent peak detection and an exemplary internal implementation of the CFR. The CFR 2a, 2b reduces the PAR of the signal supplied to a power amplifier. In the illustrated CFR 2a, 2b, a signal that is used for the peak detection is performed on a four times up-sampled signal. As the detection of peaks is typically performed based on magnitude of the input samples, a conversion from rectangular form to polar form is made.

[0042] Fig. 3 illustrates the coherent peak detection and an exemplary internal implementation of the CFR. In Coherent Mode each signal, from the two bands, is moved in frequency to their final positions in a carrier frequency. This is, in this exemplary case, accomplished as the signals are four times oversampled. This oversampling is needed in both single band and dual band configurations to find peaks between the original samples. However, in coherent mode the carriers are moved away from the middle of the band coming closer to the sampling frequency. This frequency shift will increase the risk of missing peaks between the samples due to insufficient oversampling necessary to achieve desired PAR reduction in such signals. To avoid this reduced performance, a higher oversampling rate would be required of the input signal.

[0043] An oversampling, preferably by a factor two, in the CFR would require twice as many samples in the used pulses, increasing the size of the pulse memory a factor of two and potentially doubling the clock rate. This higher rate proportionally increases the power consumption. The delay buffer would also be twice as large. The number of operations would also be double resulting in even more increased hardware. This is, in some implementations, not a feasible way to solve the problem due to the sampling rate increase and the limitations in the Application Specific Integration Circuit (ASIC) not allowing for further increases in the sampling rate.

[0044] Figs. 4a and 4b illustrate a comparison between a prior art estimate and an estimate according to the present teachings. Fig. 4a illustrates the prior art method, in which two consecutive samples, shown as dots Si, S2, are used for estimating the amplitude of the true peak Al. The sample having the highest amplitude is selected, i.e., sample S2 in the illustrated case. This estimated maximum sample is then used for pulse selection, and then also the amplitude and phase of the selected pulse.

[0045] Fig. 4b illustrates an estimate according to the present teachings. A2 is the amplitude of the true peak, and Si and S2 are two samples which are used for estimating the amplitude of the true peak A2. The samples Si, S2 can be used in different ways for estimating the true peak A2. One way is to perform an interpolation between the old samples Si, S2, obtaining an interpolation sample S3, and then selecting the maximum of the old sample Si and the interpolated sample S3. In the illustrated case, the interpolated sample S3 is estimated as the maximum sample and hence used for pulse selection, and then amplitude and phase of the selected pulse.

[0046] In contrast to the prior art solution, the present teachings take advantage of the fact that the performance of the CFR is less sensitive to a small timing error than to an amplitude error. This means that the amplitude of the peak is estimated more correctly, and a pulse with the correct amplitude can be applied, but at the reduced sampling rate. This is made at the cost of introducing a small timing error relative to the true position of the peak. If the correct amplitude of the true peak can be estimated, the amplitude of one of the neighbor samples can be replaced. Thereby a similar performance as twice, or higher, oversampling, can be obtained by the present teachings.

[0047] Several ways to estimate the amplitude of the true peak may be used. A first method is to use a low complexity local up-sampling filter. A second method is to use a low complexity interpolation filter, as exemplified next with reference to figure 5.

[0048] Fig. 5 illustrates an embodiment according to the present teachings, and in particular a filter device 40 for estimating the amplitude of a true peak. As this filter device 40 is used only for the peak estimation, the filter device 40 requires only a small number of filter coefficients. In the illustrated case, the particular exemplary filter device 40 has two unique taps: Int3 = [- 12 o 75 12875 o -12] / 128. The filter may produce one, or several, new samples in between the existing samples. The new at least one sample amplitude is compared with the amplitude of the existing neighbor samples and the one with the highest amplitude is chosen for further peak detection. In some embodiments, the largest one of the existing neighboring samples is replaced, in other embodiments the first or second existing samples are always replaced.

[0049] The peak amplitude is estimated by the filter device 40, exemplified here by a low complexity Half Band filter with only two unique filter taps. One new intermediate sample may be estimated between every existing sample pair. The intermediate new sample is compared, in a comparator 44, to the preceding sample and if the amplitude of the new sample is larger, the amplitude of, for instance, the previous sample is replaced by the new amplitude. Although the sample rate is technically increased by the interpolating filter device 40, the implementation according to figure 5 may have the same rate as currently existing methods, by means of handling new and old samples in parallel.

[0050] A particular, exemplary embodiment of the filter device 40 is shown in box 40a. The half band interpolation, having an interpolation factor equal to two, comprises a number of multipliers and adders and provide the required samples as output. Further, a particular, exemplary embodiment of the comparator 44 is also illustrated for sake of completeness. Specifically, Box 44a exemplifies how to implement the comparing of samples received from the half-band interpolation filter 40, after which the result, i.e., the maximum sample amplitude, is provided as output to peak detector.

[0051] Fig. 6 illustrates another embodiment according to the present teachings, and in particular another embodiment for estimating peak amplitude. The exemplary circuit is for even samples stream KPE processing. Briefly, this embodiment contrasts in functionality with the above-described interpolation filter device 40 in that there is no signal filtering, instead a prediction is made based on two peak-neighboring samples; rather than calculating an interpolated value, the detected peak is scaled by a gain; and finally, in this embodiment three samples are used to detect the event of a peak and predict a peak value from two of these three samples.

[0052] This embodiment, termed herein as Kernel Peak Enhancement (KPE), reacts to three signal samples at the input, and in the event that the middle sample is a peak, a gain is then applied to the sample in order to effectuate a “peak enhancement”. This is a straightforward operation and may be implemented aided by off-line Machine Learning strategies to determine the gain as a function of a feature calculated from two of the three samples and trained to the configuration of varying bandwidths signals located at known carrier frequency for a known total (multi-)band bandwidth.

[0053] Another advantage of the various KPE embodiments is that it requires, for example in a preferred hardware embodiment, less than a quarter the area and power in comparison to the interpolation filter 40 described with reference to figure 5. Further distinctions to the interpolation filter 40 are that KPE embodiments may be trained on any given Carrier Configuration offline, for example with Generative Machine Learning strategies, to populate a look-up table (LUT) of values that map to the gain that is applied to the detected peak. This results in a facsimile of the interpolated value in the first embodiment. Since the training may be calculated offline, the training signal can be represented at a much higher sampling rate to improve the prediction accuracy, while a single interpolation filter only recreates an up- sampling by two.

[0054] Additional power savings comes from the application of the 3-samples peak search and peak prediction reacting to a single signal, in this case the magnitude of the complex signal, rather than a real and imaginary individual processing. Figure 6 shows these operations in circuit form for the polyphase condition, which requires more detailed consideration than for a single stream signal.

[0055] The CFR is likely to operate at, or near, a maximum clock rate on the signal that serves as input to the CFR. To effectively mitigate peaks above a target Peak-to- Average Ratio, the CFR should preferably operate at 8x (or higher) sampling rate relative to the input sampling rate. When an up-sampling by the CFR exceeds the maximum clock rate, then a signal processing approach selected is that of a polyphase implementation. For example, a single signal stream is processed at the maximum clock rate, and to effectively process at double that clock rate while operating at the maximum clock rate, those skilled in the art are enabled to create two signal streams at the maximum clock rate with one stream processing even-indexed samples, and the other the odd-indexed samples.

[0056] The polyphase implementation of the KPE is functionally the same as with a non- polyphase input signal. However, because the input samples are presented as simultaneous pairs, and the processing delay is one sample, then care must be given to maintain the polyphase property at the output. Therefore, this method is described as a preferred implementation.

[0057] The even and odd samples are split into two parallel data streams, noted by y2in figure 6 as an input arriving from a parallel circuit, thus the circuit reacts to a sequence of four consecutive samples that are presented at two sequential clock instances. The comparator (COMP) elements in Fig. 6 thus monitor the input magnitudes (which are real and non-negative values) for the presence of a peak, when the sample present in y3is greater than y2and y4. In one exemplary calculation, this peak detection event triggers the enabling of a “x / y” divider and Look Up Table (LUT) to make the prediction of the peak between two samples, where y3, being the peak, would divide the greater value between y2and y4. Herein this division’s result is termed a Feature metric, which is the observation that will be trained offline to fill the values of the LUT, which outputs the gain (g) to multiply y3when it is a peak.

[0058] Thus, it can be observed that the circuit in Fig. 6 acts as a pass-through circuit on the input samples, until a peak is detected in sample y3(since the “odd” stream circuit is shown in Fig. 6), and it is acted upon by the divider and LUT to effectuate the prediction of a (possible) higher value at the peak. As described for the filter approach, the peak replaces the highest value sample by the higher interpolated value, and in the KPE innovation, the sample is scaled directly into the predicted peak value.

[0059] Fig. 7 illustrates an exemplary circuit for odd samples stream KPE processing. Simultaneously with the processing of odd samples stream shown in Fig. 6, there is a parallel circuit effecting the same operations but on the even samples. For sake of completeness this even-stream processing is illustrated in fig. 7, and a description of the even samples stream corresponds to the description in relation to fig. 6, i.e. to the even samples stream.

[0060] Fig. 8 illustrates a top-level strategy utilizing the elements shown in Fig. 6 and Fig. 7. If the clock rate allows for a non-polyphase implementation, an adjustment to the circuits can be made in order to process a single signal stream. Furthermore, the swapping of even- and odd-indexed samples is a consequence of inducing a delay of one sample by the KPE processing.

[0061] In the CFR, the complex signal is transformed from a typical time series of complex (Cartesian) samples to their magnitude and phase (Polar) representation. This calculation, while simplifying the KPE to operate on one real (non-negative) valued signal, introduces Intermodulation Distortions. An example of this phenomenon is shown in Fig. 9. Fig. 9 illustrates an intermodulation distortion in the KPE’s input signal. In the figure, the lower waveform denotes the real and imaginary components’ spectrum of a two-carrier signal with interpolation distortion harmonics present. This complexvalued signal is interpolated individually for the real and imaginary components by the interpolating filter method described earlier (figure 6). The other waveform, with additional peaks, and higher signal levels, is the spectrum of the absolute value of the waveform to be processed by KPE embodiments described with reference to figures 6 and 7.

[0062] The KPE relies on a Feature metric that is the ratio of the peak sample to the next- higher neighboring value that either precedes or follows it. That is, the signal appears at the input to the KPE one sample at a time, and the KPE then observes the new sample with the preceding two samples. If the previous sample, to the current input sample, is a peak, then the Feature metric is calculated as a ratio of the maximum between the current input sample and the second sample to precede it to the peak. In terms of a simple indexing, if y3is the currently presented input sample to the KPE, then if y2is the peak, the Feature metric is calculated as y3 / y2, if y3> yi, or yi / y2if yi > y3. Thus, y3is the current sample and y2is the preceding sample, with yi the sample that precedes y2.

[0063] Fig. 10a and Fig. lob illustrate the effect of taking the absolute value on the Feature metric. The smearing in the absolute value signal necessitates the application of Generative Machine Learning strategies to extract the relevant relationship between the Feature metric and the gain to be applied to the peak and remove biases from the correlated relationship between the Feature metric and the mapping to a gain value to provide the necessary accuracy.

[0064] There are two values to be learned from the training via Generative Machine Learning strategies, which inform the parameterization of two Kernels. A Kernel is, typically, a function valued between ±1, though there are exceptions, such as the Gaussian Kernel.

[0065] The various embodiments of the KPE method preferably uses two kernels: the first is to identify the spacing between samples (when a peak sample is detected), and this first kernel is termed the Sampling Kernel (Fig. toa). The second kernel is to calibrate according to the gain values range for the particular carriers configuration in the input signal, and it is termed the Gain Kernel (Fig. lob).

[0066] In Generative Machine Learning, Kernels are used in many applications, such as nonparametric regression. In a preferred embodiment, Kernels are used to model the peak waveform and to obtain a persistent correlation / prediction between samples. The sampling kernel can predict the amplitude of the next (previous) sample, if one sample is given and the distance between samples is known. Consequently, there is a need to train the Sampling Kernel such as to best estimate this distance from a synthetic model (in simulation) of the real signal’s carrier configuration and placements in their respective band(s).

[0067] Once the Sampling Kernel is specified by the distance between samples, then a mapping can be made to the location in the Gain Kernel such that the proper gain can be applied to the peak sample. This is necessary since the Sampling Kernel’s distance between sample may not achieve the desired gain.

[0068] The mapping between the two Kernels is preferably calculated offline, thus providing the entries to the LUT with the Feature metric ratio previously described, and LUT output comprising the gain value from the second Kernel.

[0069] In order train the two parameters, as in the distance between samples in the Sampling Kernel and the required (maximum) gain in the Gain Kernel, use can be made of Generative Machine Learning Linear Regression. These values will depend at least on the combination of carrier bandwidths present in the input signal. Given the distortion, or smearing, then the Generative Machine Learning approach will preferably identify outliers that can wrongly bias the desired parameter estimation.

[0070] Fig. 11 shows a linear regression with generative machine learning for a too MHz carrier. All the points in Fig. 11 are measurements of the Feature metric when a peak occurs above the target Peak-to-Average Ratio. The preferred embodiment for KPE uses a linear regression of these points in order to determine the two parameters corresponding to the Sampling Kernel and Gain Kernel. Those skilled in the art can select myriad of schemes to select the points that are significant to the parameter estimation and remove those that are detrimental to the KPE performance: outlier points. In the selected method shown, the linear regression line is calculated in an iteration scheme and the points outside one standard deviation from the line are removed. The iteration process continues until a tolerance in convergence is achieved for a given selection criterion. In Fig. 11, the smaller dots denote the points used for Linear Regression calculation, while the bigger points are the outliers.

[0071] The two parameters are then determined from the linear regression line: 1) the sampling distance is calculated with the two samples that are most unequal, and the Feature metric value indicates yields a gain of 1 (from the figure, this can be discerned at a ratio of 0.73 for the Feature metric); 2) the required maximum gain can be determined from the Feature metric having a value of 1. At this value for the Feature metric, the two samples are at the same amplitude, and the detected peak’s gain is the maximum, which can be determined from the Y-Axis value at the Feature metric value of 1.

[0072] It will be apparent to those skilled in the art that other strategies may be used for selecting outliers and optimize the KPE performance accordingly.

[0073] Fig. 12 is a flowchart of various embodiments of a method according to the herein presented teachings. The method 20 may be used for detecting peaks amplitude for a frequency band, although it is noted that the method 20 is suitable also for multiple frequency bands. The method 20 is performed in peak detection unit 10.

[0074] The method 20 comprises obtaining 21 an amplitude sample in addition to two existing amplitude samples.

[0075] The method 20 comprises comparing 22 the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples.

[0076] The method 20 comprises determining 23, based on the comparison 22, one as the peak amplitude.

[0077] The method 20 provides a number of advantages. For example, by means of the method a significantly improved performance of CFR in multiband scenarios is obtained and it may also improve CFR performance in single band configurations. Further, the method 20 is enabled with very small additions of hardware, thereby keeping down the costs thereof as well as size. Multiband configurations are becoming increasingly important as there is a request and need for supporting an increased number of bands without installing new radio units for each band.

[0078] The disclosed peak detection unit io can be provided at low costs in view of both silicon area devices and power dissipation, which the increased oversampling would else require. The method supports and is suitable for several bands, and may, for instance, be used in order to double the Instantaneous Bandwidth (IBW) of the CFR by placing two CFRs side-by-side in frequency.

[0079] In an embodiment, the comparing 22 comprises comparing the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and determining 23, based on the comparison, the highest amplitude among the compared amplitude samples as the peak amplitude.

[0080] In a variation of the above embodiment, the method comprises replacing the selected highest amplitude with one of the two existing amplitude samples; repeating the obtaining 21, comparing 22, determining 23; and then outputting 24 the highest amplitude as a peak amplitude.

[0081] In various embodiments, the method 20 comprises performing the steps obtaining 21, comparing 22, and determining 23 separately for two frequency bands; summing the absolute values of the determined respective peak amplitudes of the two frequency bands; and using the summed absolute values as peak amplitude of the two frequency bands combined.

[0082] In a variation of the above embodiment, the method 20 comprises sending the peak amplitude of the combined frequency bands to a power amplifier.

[0083] In various embodiments, the peak detection unit is a filter unit.

[0084] In one set of embodiments, the method 20 in its broadest embodiment, (i.e., comprising the obtaining 21 an amplitude sample in addition to two existing amplitude samples, the comparing 22 the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and the determining 23, based on the comparison 22, one as the peak amplitude), the two existing amplitude samples are two peak-neighboring samples, and the 23 comprises determining the obtained amplitude sample to be the peak amplitude for the case that it is greater than the two peak-neighboring samples.

[0085] In a variation of the above embodiment, the method 20 comprises, for the case that the amplitude sample is a peak amplitude, adding a gain to the amplitude sample, thus providing a peak enhancement.

[0086] In a variation of the above two embodiments, the two peak-neighboring samples comprises samples of a respective signal stream.

[0087] In various embodiments, the method 20 comprises providing the resulting peak amplitude to a Crest Factor Reduction, CFR, unit 2a, 2b for cancelling of the peak amplitude. In some embodiments, the providing comprises applying, in the CFR unit, the crest factor reduction algorithm separately for two or more frequency bands.

[0088] In various embodiments, the method 20 comprises using a look-up table of values mapping gain to the detected peak amplitude, wherein the table may be based on machine learning data.

[0089] The method 20 entails a number of advantages. For example, by means of the method a significantly improved performance of CFR in multi band scenarios is obtained and it may also improve performance in single band configurations. Further, the method 20 is enabled with very small additions of hardware, thereby keeping down the costs thereof as well as size. Multi Band configurations are becoming increasingly important as customer wants to support an increased number of bands without installing new radio units for each band.

[0090] The main advantage from the invention is improved performance in the multi band configurations without the very large cost in silicon area and power dissipation that the increased oversampling would require. Besides supporting several bands, the proposed methods can be used, for instance, to double the Instantaneous Bandwidth (IBW) of the CFR by placing two CFRs side-by-side in frequency.

[0091] A peak detection unit 10 is also disclosed, the peak detection unit 10 being configured to perform any of the embodiments of the method 20, as has been described. The peak detection unit io can be used for detecting a peak amplitude for a frequency band. The peak detection unit io is configured to: obtain an amplitude sample in addition to two existing amplitude samples, compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and determine, based on the comparison 22, one as the peak amplitude.

[0092] In an embodiment, the peak detection unit 10 is configured to compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and to determine, based on the comparison, the highest amplitude among the compared amplitude samples as the peak amplitude.

[0093] In a variation of the above embodiment, the peak detection unit 10 is configured to: replace the selected highest amplitude with one of the two existing amplitude samples, repeat the obtaining, comparing, and determining, and output the highest amplitude as a peak amplitude.

[0094] In variations of the above two embodiments, the peak detection unit 10 is configured to perform the obtaining, comparing and determining separately for two frequency bands, to sum the absolute values of the determined respective peak amplitudes of the two frequency bands, and to use the summed absolute values as peak amplitude of the two frequency bands combined.

[0095] In a variation of the above embodiment, the peak detection unit 10 is configured to send the peak amplitude of the combined frequency bands to a power amplifier.

[0096] In various embodiments, the peak detection unit 10 is a filter unit.

[0097] In one set of embodiments, in the peak detection unit 10 in its broadest embodiment, the two existing amplitude samples are two peak-neighboring samples, and the peak detection unit is configured to determine the obtained amplitude sample to be the peak amplitude for the case that it is greater than the two peak-neighboring samples.

[0098] In a variation of the above embodiment, the peak detection unit 10 is configured to, for the case that the amplitude sample is a peak amplitude, add a gain to the amplitude sample, thus providing a peak enhancement. In variations of the above two embodiments, the two peak-neighboring samples comprise samples of a respective signal stream.

[0099] In variations of the above three embodiments, the peak detection unit io is configured to provide the resulting peak amplitude to a Crest Factor Reduction, CFR, unit 2a, 2b for cancelling of the peak amplitude.

[0100] In a variation of the above embodiments, the peak detection unit io is configured to provide by applying, in the CFR unit 10, the crest factor reduction algorithm separately for two or more frequency bands.

[0101] In various embodiments, the peak detection unit io is configured to use a look-up table of values mapping gain to the detected peak amplitude, the table being based on machine learning data.

[0102] Fig. 13 illustrates a radio unit. A radio unit 50 comprising a peak detection unit 10 as described herein is also provided and illustrated highly schematically in figure 16. The peak detection unit 10, as has been described, may, for instance, be implemented as an Application Specific Integrated Circuit (ASIC).

[0103] Fig. 14 schematically illustrates, in terms of a number of functional units, the components of a peak detection unit 10, according to an embodiment. Processing circuitry 110 is provided using any combination of one or more of a suitable central processing unit (CPU), multiprocessor, microcontroller, digital signal processor (DSP), etc., capable of executing software instructions stored in a computer program product 330 (as shown in Fig. 16), e.g., in the form of a storage medium 130. The processing circuitry 110 may further be provided as at least one application specific integrated circuit (ASIC), or field programmable gate array (FPGA).

[0104] Particularly, the processing circuitry 110 is configured to cause the peak detection unit 10 to perform a set of operations, or actions, as disclosed herein. For example, the storage medium 130 may store the set of operations, and the processing circuitry 110 may be configured to retrieve the set of operations from the storage medium 130 to cause the peak detection unit 10 to perform the set of operations. The set of operations may be provided as a set of executable instructions. The processing circuitry 110 is thereby arranged to execute methods as herein disclosed. The storage medium 130 may also comprise persistent storage, which, for example, can be any single one or combination of magnetic memory, optical memory, solid state memory or even remotely mounted memory.

[0105] The peak detection unit 10 may further comprise a communications interface 120 for communications with other entities, functions, nodes, and devices, over suitable interfaces. As such the communications interface 120 may comprise one or more transmitters and receivers, comprising analogue and digital components.

[0106] The processing circuitry 110 controls the general operation of the peak detection unit 10 e.g., by sending data and control signals to the communications interface 120 and the storage medium 130, by receiving data and reports from the communications interface 120, and by retrieving data and instructions from the storage medium 130. Other components, as well as the related functionality, of the peak detection unit 10 are omitted in order not to obscure the concepts presented herein.

[0107] Fig. 15 schematically illustrates, in terms of a number of functional modules, the components of a peak detection unit 10 according to an embodiment. The peak detection unit 10 of Fig. 15 comprises a number of functional modules; an obtain module 210 configured to obtain an amplitude sample in addition to two existing amplitude samples; and a compare module 220 configured to compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples; and a determine module 230 configured to determine, based on the comparison, a peak amplitude. The peak detection unit 10 of Fig. 15 may further comprise a number of optional functional modules, such as an output module 240, configured output the highest amplitude as a peak amplitude. In general terms, each functional module 210 - 240 maybe implemented in hardware or in software. Preferably, one or more or all functional modules 210 - 240 may be implemented by the processing circuitry 110, possibly in cooperation with the communications interface 120 and the storage medium 130. The processing circuitry 110 may thus be arranged to from the storage medium 130 fetch instructions as provided by a functional module 210 - 240 and to execute these instructions, thereby performing any actions of the peak detection unit 10 as disclosed herein.

[0108] Fig. 16 shows one example of a computer program product 330 comprising computer readable means 340. On this computer readable means 340, a computer program 320 can be stored, which computer program 320 can cause the processing circuitry 110 and thereto operatively coupled entities and devices, such as the communications interface 120 and the storage medium 130, to execute methods according to embodiments described herein. The computer program 320 and / or computer program product 330 may thus provide means for performing any actions of the peak detection unit 10 as disclosed herein.

[0109] In the example of Fig. 16, the computer program product 330 is illustrated as an optical disc, such as a CD (compact disc) or a DVD (digital versatile disc) or a Blu-Ray disc. The computer program product 330 could also be embodied as a memory, such as a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or an electrically erasable programmable read-only memory (EEPROM) and more particularly as a non-volatile storage medium of a device in an external memory such as a USB (Universal Serial Bus) memory or a Flash memory, such as a compact Flash memory. Thus, while the computer program 320 is here schematically shown as a track on the depicted optical disk, the computer program 320 can be stored in any way which is suitable for the computer program product 330.

[0110] The inventive concept has mainly been described above with reference to a few embodiments. However, as is readily appreciated by a person skilled in the art, other embodiments than the ones disclosed above are equally possible within the scope of the inventive concept, as defined by the appended patent claims.

Claims

CLAIMS1. A method (20) for detecting a peak amplitude for a frequency band, the method (20) being performed in peak detection unit (10) and comprising:- obtaining (21) an amplitude sample in addition to two existing amplitude samples,- comparing (22) the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and- determining (23), based on the comparison (22), one as the peak amplitude.

2. The method (20) as claimed in claim 1, wherein the comparing (22) comprises comparing the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and determining (23), based on the comparison (22), the highest amplitude among the compared amplitude samples as the peak amplitude.

3. The method (20) as claimed in claim 2, comprising:- replacing the selected highest amplitude with one of the two existing amplitude samples,- repeating the obtaining (21), comparing (22), determining (23), and- outputting (24) the highest amplitude as a peak amplitude.

4. The method (20) as claimed in claim 1 or 2, comprising:- performing the steps (21, 22, 23) separately for two frequency bands,- summing the absolute values of the determined respective peak amplitudes of the two frequency bands, and- using the summed absolute values as peak amplitude of the two frequency bands combined.

5. The method (20) as claimed in claim 4, comprising sending the peak amplitude of the combined frequency bands to a power amplifier.

6. The method (20) as claimed in any of the preceding claims, wherein the peak detection unit (10) comprises a filter unit.

7. The method (20) as claimed in claim 1, wherein the two existing amplitude samples are two peak-neighboring samples, and the determining (23) comprises determining the obtained amplitude sample to be the peak amplitude for the case that it is greater than the two peak-neighboring samples.

8. The method (20) as claimed in claim 7, comprising, for the case that the amplitude sample is a peak amplitude, adding a gain to the amplitude sample, thus providing a peak enhancement.

9. The method (20) as claimed in claim 7 or 8, wherein the two peak-neighboring samples comprises samples of a respective signal stream.

10. The method (20) as claimed in any of the preceding claims, comprising providing the resulting peak amplitude to a Crest Factor Reduction, CFR, unit (2a, 2b) for cancelling of the peak amplitude.

11. The method (20) as claimed in claim 10, wherein the providing comprises applying, in the CFR unit (2a, 2b), the crest factor reduction algorithm separately for two or more frequency bands.

12. The method (20) as claimed in any of claims 7 - 11, comprising using a look-up table of values mapping gain to the detected peak amplitude, the table being based on machine learning data.

13. A computer program (320) for detecting a peak amplitude for a frequency band, the computer program comprising computer code which, when run on processing circuitry (110) of a peak detection unit (10), causes the peak detection unit (10) to:- obtain an amplitude sample in addition to two existing amplitude samples,- compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and- determine, based on the comparison, one as the peak amplitude.

14. A computer program product (330) comprising a computer program (320) according to claim 13, and a computer readable storage medium (340) on which the computer program (320) is stored.

15. A peak detection unit (10) for detecting a peak amplitude for a frequency band, the peak detection unit (10) being configured to:- obtain an amplitude sample in addition to two existing amplitude samples,- compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and- determine, based on the comparison (22), one as the peak amplitude.

16. The peak detection unit (10) as claimed in claim 15, configured to compare the amplitude of the obtained amplitude sample to one or both of the two existing amplitude samples, and to determine, based on the comparison, the highest amplitude among the compared amplitude samples as the peak amplitude.17- The peak detection unit (io) as claimed in claim 16, configured to:- replace the selected highest amplitude with one of the two existing amplitude samples,- repeat the obtaining, comparing, and determining, and- output the highest amplitude as a peak amplitude.

18. The peak detection unit (io) as claimed in claim 15 or 16, configured to:- perform the obtaining, comparing and determining separately for two frequency bands,- sum the absolute values of the determined respective peak amplitudes of the two frequency bands, and- use the summed absolute values as peak amplitude of the two frequency bands combined.

19. The peak detection unit (10) as claimed in claim 18, configured to send the peak amplitude of the combined frequency bands to a power amplifier.

20. The peak detection unit (10) as claimed in any of claims 15 - 19, wherein the peak detection unit (10) is a filter unit.

21. The peak detection unit (10) as claimed in claim 15, wherein the two existing amplitude samples are two peak-neighboring samples, and the peak detection unit being configured to determine the obtained amplitude sample to be the peak amplitude for the case that it is greater than the two peak-neighboring samples.

22. The peak detection unit (10) as claimed in claim 21, configured to, for the case that the amplitude sample is a peak amplitude, add a gain to the amplitude sample, thus providing a peak enhancement.

23. The peak detection unit (10) as claimed in claim 21 or 22, wherein the two peakneighboring samples comprises samples of a respective signal stream.

24. The peak detection unit (10) as claimed in any of claims 15 - 23, configured to provide the resulting peak amplitude to a Crest Factor Reduction, CFR, unit (2a, 2b) for cancelling of the peak amplitude.

25. The peak detection unit (10) as claimed in claim 24, configured to provide by applying, in the CFR unit (10), the crest factor reduction algorithm separately for two or more frequency bands.

26. The peak detection unit (10) as claimed in any of claims 21 - 25, configured to use a look-up table of values mapping gain to the detected peak amplitude, the table being based on machine learning data.

27. A radio device (50) comprising a peak detection unit (10) as claimed in any of claims 15 - 26.

Images