The present disclosure relates to communication devices and systems.
In communication networks, data is formatted for transmission over a link (wired or wireless) from a source device to a destination device. One technique to format data employs orthogonal frequency division multiplexing (OFDM) techniques wherein a frequency band of a channel (wired or wireless) comprises a relatively large number of closely-spaced orthogonal subcarriers to carry data. The data is divided into several parallel data streams or channels, one for each subcarrier. Each subcarrier is modulated, such as with quadrature amplitude modulation or phase-shift keying modulation techniques.
One of the drawbacks of OFDM signals is their relatively high peak-to-average ratio (PAPR) caused by the summation of Fast Fourier Transform (FFT) sinusoids over signals with high efficiency modulations. In a wireless communication device, the OFDM symbols are converted to analog signals, upconverted to a desired transmission frequency and amplified by a power amplifier before coupling to a transmit antenna. The high PAPR of an OFDM symbol reduces the linear dynamic range of the power amplifier. In order to avoid operating the power amplifier in its non-linear range, the power amplification can be reduced to thereby minimize non-linear effects such as degradation in error vector magnitude of the signal and out-of-band emissions. Another solution is to use a higher power amplifier, but that increases the cost of the wireless communication device.
Still another method to reduce PAPR is to clip the symbol in the time domain to a certain level to remove high peaks in the signal. Clipping increases error vector magnitude and increases out-of-band emissions. There is room for improving clipping techniques in order simplify the implementation complexity and achieve improved reduction of PAPR.
Techniques are provided for crest factor reduction of a symbol to be transmitted by a communication device. The symbol may be an orthogonal frequency division multiplexed (OFDM) formatted symbol. In a communication device, samples of the symbol are clipped with a clipping level. A signal quality of the symbol is computed after it is clipped. A determination is made as to whether the signal quality satisfies a predetermined criterion. When the signal quality does not satisfy the predetermined criterion, the clipping level is adjusted. The clipping, computing, determining and adjusting operations are repeated until the signal quality satisfies the predetermined criterion. The symbol clipped by the clipping level determined to result in satisfying the predetermined criterion is output for supply to a transmitter in the communication device. The techniques described herein are applicable to OFDM symbols as well as other types of non-OFDM symbols. Moreover, the techniques described herein are applicable to wireless and wired communication devices and networks. Techniques for computing for error vector magnitude that are faster and less computationally intensive are provided, as well as a computation for distortion that can be used as a measure of error vector magnitude.
Referring first to
The modem 12 comprises various blocks to perform modulation of signals to be transmitted and demodulation of signals that are received. For simplicity, the blocks shown inside the modem 12 are only those blocks that pertain to the adaptively clipping techniques. To this end, there is a zero-padding block 20, an Inverse Fast Fourier Transform (IFFT) block 22, a clipping block 24 and a filter 26. The zero-padding block 20 is configured to receive samples of a symbol to be transmitted. The output of the zero-padding block 20 is coupled to an input of the IFFT block 22. The output of the IFFT block is coupled to the clipping block 24, and the output of the clipping block 24 is coupled to an input of the filter 26. The filter outputs clipped and filtered time-domain samples of a symbol to an input of the DAC 14. The output of the DAC 14 is coupled to an input of the transmitter 16. The modem 12 may be implemented by digital logic gates in an application specific integrated circuit (ASIC), in one example.
The transmitter 16 comprises an RF upconverter 30, e.g., mixer, and a power amplifier (PA) 20. The output of the power amplifier 32 is coupled to the transmit antenna 34.
The clipping block 24 is controlled by outputs from a controller 40. The controller 40 comprises a processor 42, e.g., microprocessor, microcontroller, digital signal processor, etc., and memory 44. Stored within memory 44 are instructions for adaptive clipping level computation process logic 100. The process logic 100 may be embodied as computer executable instructions stored or encoded in one or more computer readable storage media (i.e., a memory device) and when executed by a processor or computer, operable to cause the processor or computer to perform the operations described herein in connection with process logic 100. The memory 44 may comprise read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, acoustical or other physical/tangible memory storage devices.
While the process logic 100 is shown in
The operations of the blocks 20-26 are now generally described. The zero-padding block 20 pads the symbol with zero's. The IFFT block 22 converts the symbol to a time-domain signal comprising time-domain samples. The clipping block 24 clips the levels of the time-domain samples of the symbol to a clipping level supplied by the adaptive clipping level computation process logic 100. The result of the clipping block 24 is that some samples of the symbols whose values (peaks) exceed the clipping level are effectively “clipped” to the clipping level. Those sample peaks that are less than the clipping level are not affected by the clipping operation. The filter 26 filters the time-domain samples of the symbol (after it is clipped) to reduce out-of-band emissions within a mask. Any filtering technique may be used. For real-time (e.g., voice or video) data applications, filtering techniques with shorter delay profiles are desirable, such as finite impulse response or infinite impulse response filters.
The adaptive clipping level computation process logic 100 is configured to adaptively compute the clipping level used to clip a symbol in order to determine the best clipping level that guarantees that a signal quality of the resulting clipped symbol will satisfy a predetermined criterion. Examples of the signal quality are error vector magnitude and distortion, and techniques for computing these signal quality metrics are described hereinafter. The clipping level that achieves the predetermined criterion is computed individually for each symbol and thus may vary from one symbol to another symbol as explained further hereinafter. In this way, each symbol is guaranteed to satisfy the predetermined signal quality criterion before it is supplied for transmission.
Reference is now made to
At 130, a control signal indicating an initial clipping level is supplied to the clipping block 24 for a new symbol to be transmitted. The initial clipping level is set to a most aggressive clipping level. The clipping block 24 clips the time-domain samples of the symbol using the initial clipping level.
After the symbol is initial clipped at 130, then at 140, a signal quality of the symbol after it is clipped is computed. Examples of signal quality metrics and their computations are described in more detail hereinafter. At 150, the signal quality is evaluated to determine whether it satisfies a predetermined criterion, e.g., meets or exceeds (greater than or equal to) a threshold. As explained above, the predetermined criterion may be set based on the highest modulation/coding scheme for the symbol. For example, if the highest modulation/coding scheme for a given symbol is 64 quadrature amplitude modulation, the predetermined criterion may be more rigorous (e.g., a higher EVM or rigorous distortion measure) than if the highest modulation/coding scheme is 16 quadrature amplitude modulation. Consequently, since the highest modulation/coding scheme can vary from one symbol to another, the predetermined criterion may also vary for each symbol.
When the signal quality is determined to satisfy the predetermined criterion at 150, then at 160, the samples of the clipped symbol (with the most recent clipping level) are supplied to the filter 26 and then ultimately (after RF upconversion) to the power amplifier 32 for transmission. The next symbol can thereafter be used to begin the process again at 110. It is unusual that the signal quality after the initial clipping level will satisfy the predetermined criterion. More likely, after the initial clipping level operation at 130, the evaluation at 150 will yield a negative outcome and the process proceeds to operation 170 for another iteration.
At 170, the clipping level is increased from the previous level (which after the first time through would be the initial clipping level) and a control signal is supplied to the clipping block 24 with the new clipping level so that the clipping block 24 clips the samples of the symbol with the new clipping level.
After the time-domain samples of the symbol are clipped with the new clipping level, the process goes back to operation 140 where the signal quality for the clipped symbol (with the new clipping level) is computed. The signal quality is then evaluated at operation 150 to determine whether the clipped symbol at the new clipping level satisfies the predetermined criterion. If not, then the process continues such that the clipping level is adjusted (increased) again at operation 170.
The loop defined by operations 140, 150 and 170 in
Thus, the operation flow depicted by
The signal quality metric computed at operation 150 in
The following description sets forth a computation technique useful for computing error vector magnitude, described thereafter. Let X(i), i=1, 2, . . . , N denote an input symbol vector of N samples of an OFDM symbol. The vector is interpolated such that Xint(i)=[0(L-1)N/2×0(L-1)N/2]T, where L is the interpolation ratio. The time-domain signal is represented by the vector x=F−1Xint, where F and F−1 are FFT and IFFT operation matrices, respectively.
The clipping block 24 performs polar clipping on the time-domain symbol samples and the clipping operation is represented as represented as
where Amax is the clipping level. The set C contains the indices of samples of the OFDM symbol that have been clipped. In polar clipping, the phase of the signal remains the unchanged.
The time error in the OFDM symbol after clipping is denoted Δt=x−xc, where xc is the clipped symbol vector. The frequency error can be derived from the time error as Δf=F·Δt. Since not all of the samples of a symbol are clipped, only time errors for indices of C that are non-zero are computed, C={i1i2 . . . iN
Error Vector Magnitude as the Signal Quality Metric
Error vector magnitude (EVM) is defined as
where Mmax is the maximum magnitude of the modulation symbol and is set for the highest modulation of the OFDM symbol.
Thus, in one example, the signal quality computation at operation 140 involves computing EVM in the time-domain using the computation:
where the matrix B is Nc×Nc with bij elements of
and where B is a Hermitian matrix bij=b*ij, Nc is the number of clipped time-domain samples of a symbol and H is the Hermitian operation.
The elements bij are a function of (i-j) in mod of LN. Therefore, the maximum possible of bij in mod of LN is LN elements. The LN elements of bij can be computed once and stored in memory. Using the indices of set C, the matrix B can be retrieved from memory. The FASTEVM computation described above has a reduced complexity compared with an EVM computation that requires FFT computations and involves the complexity of α(LN)log(LN) in addition to N multiplications for computing the frequency error powers, where α is greater than one and depends on the particular FFT implementation made. In other words, standard EVM computations require multiple LN log(LN) computations. By contrast, the complexity of computing the FASTEVM metric is Nc(Nc+1)/2 (Nc<<N). Nc is a function of the clipping level but is on the order of less than 3 percent of LN. For example, for a 6.2 dB clipping level (which may be the most aggressive or initial clipping level used), max(Nc) was found to be 104 over 10000 OFDM symbols. Thus, the FASTEVM computation achieves a 90% reduction in complexity (assuming α=1), it does not require FFT computations and it makes use of the time-domain clipped samples of a symbol.
An example of a sequence through the process logic 100 is as follows. An initial aggressive level of PAPR level for clipping is selected, for example 6 dB. In some symbols 6 dB clipping may clip 5% of the samples of the symbol because only 5% of the samples, while in other OFDM symbols, the 6 dB clipping level may clip a greater or lesser percentage of samples.
After the initial clipping, the EVM is computed as −26 dB. The goal is to reach an EVM of −35 dB, for example. Thus, the predetermined criterion in this example is a floor threshold of −35 dB. The clipping level is thus increased to 6.4 dB and the EVM for that level is computed as −32 dB. The clipping level is increased again to 6.7 dB and the EVM is computed as −35 dB. The process can stop and the symbol clipped at 6.7 dB is sent to the filter 26 and then ultimately to the power amplifier in the transmitter 16 for transmission. There may be some cases that increasing the clipping level might not reduce the number of clipped samples of a symbol but increasing the clipping level will result in a decrease in the EVM (or distortion).
Distortion as the Signal Quality Metric
Another signal quality metric that may be computed at operation 140 is a distortion measure. A distortion computation is even less complex than the FASTEVM computation. Distortion is computed to guarantee a given EVM. Distortion is defined as the power of the error of the clipped signal.
Frequency error be re-written as
where Fi is column i of the FFT matrix F.
Distortion is computed as
where ∥·∥2 is the norm operation, N is the number of time-domain samples of the symbol, Nc is the number time-domain samples that are clipped, L is an interpolation ratio, Δt=x−xc represents a time error of the clipped symbol, x is a vector that represents the symbol in the time-domain and xc is the clipped symbol vector and is the index of samples in a set C that are clipped.
The complexity of computing distortion is significantly low, i.e., Nc multiplications. For example, if a symbol of 4×1024 has only 63 samples that have been clipped, then the distortion computation is only the power of the error of those 63 samples.
Reference is now made to
The adaptive clipping techniques described herein provide improved crest factor or PAPR reduction than other techniques heretofore known. Each symbol is treated individually and the amount of clipping is determined for each symbol in order to achieve a desired or target EVM (or distortion). Using this adaptive or “soft” clipping technique, an EVM of a desired level can be guaranteed and as a result there is no need for in-band post-processing of the symbol. Moreover, since each symbol is guaranteed to achieve the desired EVM, various post-clipping filtering operation may be used. The filter techniques that are faster in real-time may be used instead of FFT-based filtering methods. Consequently, the overall crest factor reduction solution is less expensive, yet still achieves a desired EVM performance.
In sum, in one form, a method is provided comprising clipping time-domain samples of a symbol with a clipping level; computing a signal quality of the symbol after it is clipped; determining whether the signal quality satisfies a predetermined criterion; adjusting the clipping level when the signal quality does not satisfy the predetermined criterion; repeating the clipping, computing, determining and adjusting until the signal quality satisfies the predetermined criterion; and outputting for supply to a transmitter in the wireless communication device the symbol clipped by the clipping level determined to satisfy the predetermined criterion.
In another form, an apparatus is provided comprising a modem configured to generate symbols for transmission, the modem configured to clip time-domain samples of a symbol prior to supplying the symbol for transmission; and a controller configured to compute a signal quality of the symbol after it is clipped, determine whether the signal quality satisfies a predetermined criterion, adjust the clipping level when the signal quality does not satisfy the predetermined criterion and control the modem to repeat the clipping operation for a new clipping level, and repeat the compute, determine, adjust and control operations until the signal quality satisfies the predetermined criterion.
In still another form, one or more computer readable storage media are provided encoded with software comprising computer executable instructions and when the software is executed operable to: compute a signal quality of the a symbol to be transmitted after is clipped with a clipping level; determine whether the signal quality satisfies a predetermined criterion; adjust the clipping level when the signal quality does not satisfy the predetermined criterion and control the modem to repeat the clipping operation for a new clipping level; and repeat the compute, determine, adjust and control operations until the signal quality satisfies the predetermined criterion.
The techniques described herein are is useful in connection with any communication technology using OFDM techniques, such as WiFi™, WiMAX™, Long Term Evolution (LTE), as well as other technologies that using non-OFDM techniques.
The above description is intended by way of example only. Various modifications and structural changes may be made therein without departing from the scope of the concepts described herein and within the scope and range of equivalents of the claims.