US20200252032A1 - Compensating for transmitter nonlinearities - Google Patents
Compensating for transmitter nonlinearities Download PDFInfo
- Publication number
- US20200252032A1 US20200252032A1 US16/779,610 US202016779610A US2020252032A1 US 20200252032 A1 US20200252032 A1 US 20200252032A1 US 202016779610 A US202016779610 A US 202016779610A US 2020252032 A1 US2020252032 A1 US 2020252032A1
- Authority
- US
- United States
- Prior art keywords
- coefficients
- transmitter
- nonlinear
- error signal
- memory depth
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 claims abstract description 31
- 230000004044 response Effects 0.000 claims abstract description 15
- 238000004891 communication Methods 0.000 claims abstract description 13
- 238000004458 analytical method Methods 0.000 claims description 7
- 230000005693 optoelectronics Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 description 13
- 230000003287 optical effect Effects 0.000 description 8
- 238000005457 optimization Methods 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000000116 mitigating effect Effects 0.000 description 3
- 230000009022 nonlinear effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000010363 phase shift Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F1/00—Details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements
- H03F1/32—Modifications of amplifiers to reduce non-linear distortion
- H03F1/3241—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits
- H03F1/3247—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits using feedback acting on predistortion circuits
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F1/00—Details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements
- H03F1/32—Modifications of amplifiers to reduce non-linear distortion
- H03F1/3241—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits
- H03F1/3258—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits based on polynomial terms
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F1/00—Details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements
- H03F1/32—Modifications of amplifiers to reduce non-linear distortion
- H03F1/3241—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits
- H03F1/3276—Modifications of amplifiers to reduce non-linear distortion using predistortion circuits using the nonlinearity inherent to components, e.g. a diode
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F3/00—Amplifiers with only discharge tubes or only semiconductor devices as amplifying elements
- H03F3/189—High frequency amplifiers, e.g. radio frequency amplifiers
- H03F3/19—High frequency amplifiers, e.g. radio frequency amplifiers with semiconductor devices only
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F3/00—Amplifiers with only discharge tubes or only semiconductor devices as amplifying elements
- H03F3/20—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers
- H03F3/24—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers of transmitter output stages
- H03F3/245—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers of transmitter output stages with semiconductor devices only
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F2200/00—Indexing scheme relating to amplifiers
- H03F2200/207—A hybrid coupler being used as power measuring circuit at the output of an amplifier circuit
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F2201/00—Indexing scheme relating to details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements covered by H03F1/00
- H03F2201/32—Indexing scheme relating to modifications of amplifiers to reduce non-linear distortion
- H03F2201/3224—Predistortion being done for compensating memory effects
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F2201/00—Indexing scheme relating to details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements covered by H03F1/00
- H03F2201/32—Indexing scheme relating to modifications of amplifiers to reduce non-linear distortion
- H03F2201/3233—Adaptive predistortion using lookup table, e.g. memory, RAM, ROM, LUT, to generate the predistortion
Definitions
- the present invention relates generally to communication systems, and particularly to methods and devices for mitigating distortion in such systems.
- optical interconnects use low-cost, power-efficient optoelectronic components, such as vertical-cavity surface-emitting lasers (VCSELs) and integrated Mach-Zender modulators (MZMs).
- VCSELs vertical-cavity surface-emitting lasers
- MZMs integrated Mach-Zender modulators
- SNR signal/noise ratio
- nonlinear filters grow at polynomial rates or even faster, as both the number of non-linear orders and the length of each order increase.
- the “length” of a filter refers to the number on taps in a discrete time-domain implementation of the filter, i.e., the number of input samples that are concurrently multiplied by respective tap coefficients and added to produce each filtered output sample.
- the filter length is also referred to as the memory depth or memory order.
- Embodiments of the present invention that are described hereinbelow provide improved methods for mitigating distortion, as well as devices and systems implementing such methods.
- a method for communication which includes producing an error signal by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal.
- the error signal is decomposed into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth.
- First coefficients, up to the first memory depth, of a linear predistortion kernel are computed for application to the driving signal so as to compensate for the linear component of the error signal.
- Second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel are computed for application to the driving signal so as to compensate for the nonlinear component of the error signal. Operation of the transmitter is optimized using the first and second coefficients.
- decomposing the error signal includes computing a Volterra series representation of the error signal, whereby the linear and nonlinear components of the error signal are represented respectively by linear and nonlinear parts of the Volterra series.
- the one or more polynomial orders include a plurality of polynomial orders
- computing the second coefficients includes constraining the second memory depth to a given number over all of the polynomial orders, so that the nonlinear predistortion kernel includes the same given number of the second coefficients for all of the polynomial orders.
- computing the first coefficients includes finding an optimal number of the first coefficients in order to compensate for the linear component of the error signal
- constraining the second memory depth includes constraining the given number of the second coefficients to be less than the optimal number of the first coefficients.
- computing the second coefficients includes finding an optimal number of the second coefficients over all of the polynomial orders in order to compensate for the nonlinear component of the error signal.
- optimizing the operation of the transmitter includes applying a digital predistortion, using the first and second coefficients, to a sequence of samples of the driving signal that are input to the transmitter.
- applying the digital predistortion includes applying a look-up table (LUT) based on the second coefficients and a finite impulse response (FIR) filter including the first coefficients to the sequence of the samples in order to generate predistorted samples for input to the transmitter.
- applying the LUT includes looking up the samples in the sequence in the LUT in order to generate linearized samples, and applying the FIR includes multiplying the linearized samples by the first coefficients to generate respective tap products, and summing the tap products to generate the predistorted samples.
- optimizing the operation of the transmitter includes identifying optimal values of the first and second memory depths and setting respective numbers of the first and second coefficients to be used in the digital predistortion according to the optimal values of the first and second memory depths.
- optimizing the operation of the transmitter includes estimating a dependence of the second coefficients on an operating parameter of the transmitter, and setting the operating parameter responsively to the estimated dependence.
- setting the operating parameter includes adjusting a bias voltage applied to the transmitter.
- setting the operating parameter includes adjusting a modulation range of the transmitter.
- optimizing the operation of the transmitter includes analyzing a compatibility of the transmitter with a physical link over which the output signal is to be transmitted, and modifying a property of the physical link in order to rectify a detected incompatibility.
- the transmitter includes an optoelectronic component to which the driving signal is applied.
- a system for communication analysis including a memory configured to record an error signal produced by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal.
- a processor is configured to decompose the error signal into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth, to compute first coefficients, up to the first memory depth, of a linear predistortion kernel for application to the driving signal so as to compensate for the linear component of the error signal, to compute second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel for application to the driving signal so as to compensate for the nonlinear component of the error signal, and to optimize operation of the transmitter using the first and second coefficients.
- a computer software product including a non-transitory computer-readable medium in which program instructions are stored, which instructions, when read by a processor, cause the processor to receive an error signal produced by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal, to decompose the error signal into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth, to compute first coefficients, up to the first memory depth, of a linear predistortion kernel for application to the driving signal so as to compensate for the linear component of the error signal, to compute second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel for application to the driving signal so as to compensate for the nonlinear component of the error signal, and to optimize operation of the transmitter using the first and second coefficients.
- FIG. 1 is a block diagram that schematically illustrates a communication system with digital predistortion (DPD), in accordance with an embodiment of the invention
- FIG. 2 is a block diagram that schematically illustrates components involved in adjustment of a transmitter with DPD, in accordance with an embodiment of the invention
- FIG. 3 is a flow chart that schematically illustrates a method for optimizing DPD coefficients, in accordance with an embodiment of the invention
- FIG. 4 is a schematic plot of SNR as a function of the number of taps in a linear component of a DPD circuit, in accordance with an embodiment of the invention
- FIG. 5 is a schematic plot of SNR as a function of the numbers of taps and nonlinear orders in a nonlinear component of a DPD circuit, in accordance with an embodiment of the invention.
- FIGS. 6A and 6B are schematic plots of SNR as a function of nonlinear order in an MZM, showing the effect of bias on nonlinear distortion, in accordance with an embodiment of the invention.
- Embodiments of the present invention that are described herein address the problem of characterizing the nonlinear distortion in a communication system, in a way that enables efficient compensation for nonlinear effects in optimizing the operation of the system.
- This optimization may take the form, for example, of selecting the number of nonlinear orders to handle in designing an equalization filter, as well as selecting the memory depth (i.e., the number of taps or coefficients) of the filter that should be applied to each order. Additionally or alternatively, such optimization may be applied in setting coefficient values and other operating parameters in order to improve the performance of an existing circuit.
- Volterra series is an expansion of a Taylor series to include temporal effects.
- Discrete-time, real-valued Volterra series have the following general form, which assumes a uniform memory depth for all the nonlinear orders:
- x(n) is the input to the transmitter
- y(n) is the output
- K is the number of nonlinear orders in the model
- M is the memory depth, which is assumed to be constant for all orders.
- the set of coefficients of the Volterra series ⁇ h k (t l . . . t k ) ⁇ is also referred to as the kernel, whose values characterize the nonlinear response of the transmitter (or other system components that are subjects of the model). Due to cross-terms in the series, the size of the kernel ⁇ grows rapidly with the number of nonlinear orders and the memory depth. Specifically,
- transmitters can be represented accurately by a Volterra series that is constrained by the assumption that all of the higher polynomial orders have the same memory depth, and thus the same number of coefficients M(k>1), which is less than M(1).
- M(k>1) which is less than M(1).
- these constraints reduce the number of degrees of freedom sufficiently so that the computation of the optimal kernel ⁇ h k (t 1 . . . t k ) ⁇ becomes mathematically tractable, without substantially degrading the accuracy of the model.
- this sort of analysis may be applied with fewer constraints, though at the expense of higher computational complexity.
- x ⁇ ( n ) h 0 + h 1 ⁇ y ⁇ ( n ) + h 2 ⁇ y ⁇ ( n - 1 ) + h 3 ⁇ y ⁇ ( n - 2 ) + h 4 ⁇ y ⁇ ( n - 3 ) ⁇ linear ⁇ + h 5 ⁇ y ⁇ ( n ) 2 + h 6 ⁇ y ⁇ ( n - 1 ) 2 + h 7 ⁇ y ⁇ ( n ) ⁇ y ⁇ ( n - 1 ) ⁇ nonlinear
- some embodiments of the present invention define and apply two different compensators: a simple linear compensator with longer memory, and a nonlinear compensator with shorter memory.
- Distortion compensation can be applied either by an equalizer on the receive side of the communication line or by predistortion on the transmit side.
- receive-side equalization may be preferable when distortion is linear
- the inventors have found that the most efficient and effective way to compensate for nonlinear distortion originating in the transmitter is by adding digital predistortion (DPD) to the input signal that is used to drive the transmitter.
- DPD digital predistortion
- the method of analysis described above is reversed in order to find the coefficients of a predistortion kernel for application to the input driving signal x(n), to give a predistorted digital signal ⁇ circumflex over (x) ⁇ (n) for input to the transmitter.
- the predistortion kernel is optimized, as described below, so as to minimize the error between the original input x(n) and the transmitter output signal y(n).
- Embodiments of the present invention use these principles of analysis and predistortion in optimizing the operation of a transmitter.
- an error signal is produced (either by actual measurement or simulation) by comparing a driving signal x(n) that is applied to the transmitter to an output signal y(n) that is generated by the transmitter in response to the driving signal.
- a processor decomposes the error signal into a linear component having a given memory depth and a nonlinear component having one or more polynomial orders and a memory depth that is smaller than that of the linear component.
- the processor computes coefficients, up to the given memory depth, of a linear predistortion kernel for application to the driving signal.
- the processor also computes coefficients for the one or more polynomial orders, up to the smaller memory depth, of a nonlinear predistortion kernel.
- linear and nonlinear predistortion kernels to the driving signal compensates respectively for the linear and nonlinear component of the error signal.
- the characteristics of these kernels in terms of their memory depths, number of significant nonlinear orders, and coefficient values, are used, as noted earlier, in optimizing the operation of the transmitter, for example in selecting filter characteristics the design phase of the transmitter or in setting operating parameters of an existing transmitter of system.
- FIG. 1 is a block diagram that schematically illustrates a communication system 20 with digital predistortion (DPD), in accordance with an embodiment of the invention.
- DPD digital predistortion
- FIG. 1 shows a typical operating environment in which principles of the present invention may be implemented.
- an optical transmitter 22 transmits a digital signal to an optical receiver 24 over a communication link 26 , such as a fiberoptic cable.
- Optical transmitter 22 is assumed to include an optoelectronic component, such as a VCSEL or MZM, whose modulation gives rise to nonlinear distortion at the transmitter output.
- the principles of the present invention may be applied to transmitters that operate over links of other sorts, such as RF cables and wireless links.
- a signal generator 28 in system 20 receives digital data for transmission over link 26 , digitally modulates the data to generate a corresponding sequence of symbols, and converts the sequence of symbols into a digital driving signal having the form of a stream of digital samples x(n), as is known in the art.
- the digital samples are input to a nonlinear DPD block 30 , which applies a digital predistortion to the samples, including both linear and nonlinear components, as explained above.
- the structure and operation of block 30 are described in greater detail hereinbelow.
- An analog drive circuit converts the predistorted sample stream into an analog driving signal, which it applies to drive transmitter 22 .
- a sensor 34 comprising a suitable photodiode, for example, samples the optical signals that that are output from transmitter 22 and generates a corresponding stream of output samples y(n).
- a system analyzer 36 produces an error signal by comparing the samples x(n) of the driving signal that is applied to transmitter 22 to the samples y(n) of the output signal. As explained above, system analyzer 36 decomposes the error signal into linear and nonlinear components, and based on these components, computes the coefficient values of the linear and nonlinear predistortion kernels for application by nonlinear DPD block 30 to the samples x(n).
- System analyzer 36 typically performs this kernel computation as a preliminary step, before transmitter 22 begins actual data transmission, in order to set the coefficient values in DPD block 30 and/or set other operating parameters of transmitter 22 . Additionally or alternatively, system analyzer 36 may continue to evaluate and update the coefficient values during operation of system 20 in order to compensate for changes in the distortion characteristics of the transmitter over time.
- system analyzer 36 comprises a general-purpose computer processor 38 , along with a memory 40 , which records the samples and the error signal that are used by the processor in its kernel computation.
- Processor 38 performs the functions that are described herein under the control of software, which may be downloaded to the system analyzer in electronic form, over a network, for example.
- the software may be stored in tangible, non-transitory computer-readable media, such as optical, magnetic, or electronic memory media.
- at least some of the functions of processor 38 may be performed by a digital signal processor or by digital logic circuits, which may be hard-wired or programmable.
- FIG. 2 is a block diagram that schematically shows details of the structure and adjustments of nonlinear DPD block 30 , in accordance with an embodiment of the invention.
- DPD block 30 comprises a look-up table (LUT) 46 with entries based on the nonlinear kernel coefficients, and a finite impulse response (FIR) filter 48 with taps based on the linear kernel coefficients.
- LUT look-up table
- FIR finite impulse response
- DPD block 30 first looks up each successive sample or sub-sequence of samples (with the number of samples depending on the memory depth M(k>1)) in LUT 46 to read out linearized sample values.
- the DPD block then inputs these linearized samples to FIR 48 , in which the linearized samples are multiplied by the linear kernel coefficients to generate respective tap products, which are summed to generate the predistorted samples for application to transmitter 22 .
- DPD block 30 in this embodiments are typically implemented in digital hardware logic, which may be hard-wired or programmable, along with a suitable memory to hold LUT 46 .
- digital hardware logic which may be hard-wired or programmable, along with a suitable memory to hold LUT 46 .
- DPD block may be implemented in a programmable DSP or in software or firmware running on a microcontroller or other processor.
- the value of the memory depth M(1) of the linear part of the DPD can be optimized, in the sense that the number of taps in FIR filter 48 is chosen so as to be sufficient to achieve a certain level of SNR, without exceeding the number of taps necessary for this purpose.
- the value of the memory depth M(k>1) for the nonlinear part of the DPD is likewise optimized. Because the optimal value of M(k>1) is generally small (for example, equal to 2 or 3), the number of entries in LUT 46 is also manageable, with one table entry for each sub-sequence of M(k>1) input samples. Another advantage of this approach is that because all the nonlinear orders are constrained to have the same memory depth, the size of the LUT is not affected by the number of nonlinear orders that are included in the kernel.
- FIG. 2 shows a particular serial implementation of DPD block 30
- DPD block 30 may apply other filtering architectures that are known in the art, such as frequency-domain filtering.
- system analyzer 36 computes an error signal based on the difference between the respective samples output by signal generator 28 and by sensor 34 , and decomposes the error signal into linear and nonlinear components.
- Processor 38 uses these error components in computing the linear and nonlinear parts of the predistortion kernel.
- processor 38 may apply a Wiener filter 42 in computing the linear predistortion coefficients, while applying a polynomial Volterra filter 44 , as defined above, in computing the nonlinear predistortion coefficients.
- the kernel computations may be carried out using any suitable method of optimization that is known in the art, typically using the mean squared error (MSE) as the cost function.
- MSE mean squared error
- processor 38 may apply a single-step linear least squares estimation algorithm for this purpose, or an adaptive least mean squares algorithm.
- the kernel computations may use other cost functions, such as the bit error rate (BER), as well as other sorts of estimation algorithms that are known in the art, such as recursive least squares (RLS) estimation.
- BER bit error rate
- RLS recursive least squares
- FIG. 3 is a flow chart that schematically illustrates a method for optimizing DPD coefficients, in accordance with an embodiment of the invention. For the sake of concreteness and clarity, it is assumed that this method is carried out by system analyzer 36 in system 20 . The principles of this method may similarly be applied, mutatis mutandis, in correcting for nonlinearities in other systems, as well as in optimizing other operating parameters.
- the method of FIG. 3 begins with a process of kernel estimation 50 , in which processor 38 computes coefficients of the linear and nonlinear distortion kernels, as explained above.
- Processor 38 may repeat the computation under different system conditions and using different values of the Volterra series parameters M(1), M(k>1) and K, for the purpose of extracting model properties 52 . This step enables the designer or system operator to observe and understand the nonlinearities in system 20 and how these nonlinearities are affected by the various parameters.
- Model properties 52 also enable the designer or system operator to choose the parameter values that will best mitigate nonlinear effects.
- the model properties indicate an optimal memory depth 54 for purposes of mitigating distortion in DPD block 30 .
- the model properties enable the designer or operator to find the optimal number of taps to include in FIR filter 48 , as well as the optimal number of coefficients (and hence input samples) to use in generating LUT 46 .
- the model properties can also be used to find an optimal nonlinear order 56 to use in computing the entries of LUT 46 .
- the numbers of taps, coefficients and orders are “optimal” in the sense that they are sufficient to achieve a certain level of SNR, without exceeding the number of taps necessary for this purpose, and thus represent the best tradeoff between the cost and performance of the target system.
- FIG. 4 is a schematic plot of SNR as a function of the number of taps (coefficients) in linear FIR filter 48 in DPD block 30 , in accordance with an embodiment of the invention.
- the SNR in this example represents the difference between the optimal non-distorted signal and the equalized received signal, and is formulated as:
- E[ ⁇ ] is the expectation operator.
- the plot in FIG. 4 shows that the optimal number of taps is six, since the SNR increases significantly with each tap that added up to the sixth, but only negligibly with the seventh and subsequent taps.
- FIG. 5 is a schematic plot of SNR as a function of the numbers of taps and nonlinear orders used in generating LUT 46 in DPD block 30 , in accordance with an embodiment of the invention.
- the starting point for this plot is the SNR level of about 22 dB that is achieved by application of FIR 48 , as illustrated in FIG. 4 .
- the characteristics in FIG. 5 are shown solely by way of example. The characteristics in this case were computed subject to the constraint that all orders share the number of taps (i.e., have the same memory depth). Alternatively, plots of this sort can be computed and used without this constraint, though at the cost of greater complexity.
- the optimal number of nonlinear taps is three, since SNR increases significantly up to the third tap but only negligibly thereafter. Incorporating seven nonlinear orders gives the greatest improvement in the SNR. Although this high order of nonlinearity complicates the computation of the entries in LUT 46 , the computation can be performed offline and does not increase the actual size of the LUT.
- FIGS. 6A and 6B are schematic plots of SNR as a function of nonlinear order in an MZM, showing the effect of bias on nonlinear distortion, in accordance with an embodiment of the invention.
- the MZM comprises a Mach-Zender interferometer with two parallel legs, in an electro-optical medium that applies a variable phase shift to (at least) one of the legs as a function of the applied voltage.
- the MZM is inherently nonlinear, with a sinusoidal response to variations in the applied voltage level about the default bias (V ⁇ /2 ).
- system analyzer 36 computes the nonlinear predistortion kernel of an MZM in order to analyze its voltage response and bias voltage.
- the nonlinear coefficients derived by system analyzer 36 provide an estimate of the dependence of the nonlinear coefficients on the operating parameters of the transmitter—specifically on the bias voltage in the present example. Based on this information, a system operator will adjust the bias voltage of the MZM in order to achieve the desirable characteristic shown in FIG. 6A .
- the system designer or operator may observe the effect of the range of modulation voltages on the nonlinear kernel coefficients computed by the system analyzer.
- the voltage range can be increased in order to increase the depth of modulation and thus support higher-order digital modulation schemes.
- the limit of the modulation range may be set at the point at which it is no longer possible to achieve the desired SNR due to excessive nonlinear distortion. More generally, the limit of the modulation range is chosen based on the optimal tradeoff between cost and performance, which can include additional factors besides the distortion.
- estimation of the nonlinear kernel coefficients may be used in defining physical link properties, including nonlinear reflections. The compatibility of a particular physical link with a given transmitter and receiver can be analyzed in this fashion, and the link properties can then be modified as necessary to rectify problems that are detected.
- estimation of the nonlinear kernel coefficients can be used to analyze and optimize the equalization that is required to compensate for instabilities (such as temperature-related drift) of an external optical modulator, such as a micro-ring modulator.
Abstract
Description
- This application claims the benefit of U.S. Provisional Patent Application 62/800,579, filed Feb. 4, 2019, which is incorporated herein by reference.
- The present invention relates generally to communication systems, and particularly to methods and devices for mitigating distortion in such systems.
- Demands for ever-greater communication data rates in data center networks have led to increasing use of optical interconnects, and even so have pushed transmission components to their physical limits. These optical interconnects use low-cost, power-efficient optoelectronic components, such as vertical-cavity surface-emitting lasers (VCSELs) and integrated Mach-Zender modulators (MZMs). In order to overcome bandwidth limitations, many links use high-order modulation formats, which in turn require large amplitude swings in the output signals from the transmitter. As a consequence, the transmission components are forced to operate outside their linear regimes. The resulting nonlinear distortion leads to a significant reduction in overall signal/noise ratio (SNR).
- Methods for estimating and compensating for nonlinear distortion are known in the art, but they are difficult to implement in practice. Whereas the length of a linear filter tends to grow linearly with the severity of the inter-symbol interference (ISI) for which it is required to compensate, nonlinear filters grow at polynomial rates or even faster, as both the number of non-linear orders and the length of each order increase. (The “length” of a filter, as used in the present description and in the claims, refers to the number on taps in a discrete time-domain implementation of the filter, i.e., the number of input samples that are concurrently multiplied by respective tap coefficients and added to produce each filtered output sample. The filter length is also referred to as the memory depth or memory order.) Therefore, it is often difficult for the designer to satisfy the conflicting demands of both reducing nonlinear distortion and minimizing circuit size, cost and power consumption.
- Embodiments of the present invention that are described hereinbelow provide improved methods for mitigating distortion, as well as devices and systems implementing such methods.
- There is therefore provided, in accordance with an embodiment of the invention, a method for communication, which includes producing an error signal by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal. The error signal is decomposed into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth. First coefficients, up to the first memory depth, of a linear predistortion kernel are computed for application to the driving signal so as to compensate for the linear component of the error signal. Second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel are computed for application to the driving signal so as to compensate for the nonlinear component of the error signal. Operation of the transmitter is optimized using the first and second coefficients.
- In a disclosed embodiment, decomposing the error signal includes computing a Volterra series representation of the error signal, whereby the linear and nonlinear components of the error signal are represented respectively by linear and nonlinear parts of the Volterra series.
- In some embodiments, the one or more polynomial orders include a plurality of polynomial orders, and computing the second coefficients includes constraining the second memory depth to a given number over all of the polynomial orders, so that the nonlinear predistortion kernel includes the same given number of the second coefficients for all of the polynomial orders. In one embodiment, computing the first coefficients includes finding an optimal number of the first coefficients in order to compensate for the linear component of the error signal, and constraining the second memory depth includes constraining the given number of the second coefficients to be less than the optimal number of the first coefficients. Additionally or alternatively, computing the second coefficients includes finding an optimal number of the second coefficients over all of the polynomial orders in order to compensate for the nonlinear component of the error signal.
- In some embodiments, optimizing the operation of the transmitter includes applying a digital predistortion, using the first and second coefficients, to a sequence of samples of the driving signal that are input to the transmitter. In one embodiment, applying the digital predistortion includes applying a look-up table (LUT) based on the second coefficients and a finite impulse response (FIR) filter including the first coefficients to the sequence of the samples in order to generate predistorted samples for input to the transmitter. In a disclosed embodiment, applying the LUT includes looking up the samples in the sequence in the LUT in order to generate linearized samples, and applying the FIR includes multiplying the linearized samples by the first coefficients to generate respective tap products, and summing the tap products to generate the predistorted samples. Additionally or alternatively, optimizing the operation of the transmitter includes identifying optimal values of the first and second memory depths and setting respective numbers of the first and second coefficients to be used in the digital predistortion according to the optimal values of the first and second memory depths.
- In alternatively embodiments, optimizing the operation of the transmitter includes estimating a dependence of the second coefficients on an operating parameter of the transmitter, and setting the operating parameter responsively to the estimated dependence. In one embodiment, setting the operating parameter includes adjusting a bias voltage applied to the transmitter. Additionally or alternatively, setting the operating parameter includes adjusting a modulation range of the transmitter. Further additionally or alternatively, optimizing the operation of the transmitter includes analyzing a compatibility of the transmitter with a physical link over which the output signal is to be transmitted, and modifying a property of the physical link in order to rectify a detected incompatibility.
- In a disclosed embodiment, the transmitter includes an optoelectronic component to which the driving signal is applied.
- There is also provided, in accordance with an embodiment of the invention, a system for communication analysis, including a memory configured to record an error signal produced by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal. A processor is configured to decompose the error signal into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth, to compute first coefficients, up to the first memory depth, of a linear predistortion kernel for application to the driving signal so as to compensate for the linear component of the error signal, to compute second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel for application to the driving signal so as to compensate for the nonlinear component of the error signal, and to optimize operation of the transmitter using the first and second coefficients.
- There is additionally provided, in accordance with an embodiment of the invention, a computer software product, including a non-transitory computer-readable medium in which program instructions are stored, which instructions, when read by a processor, cause the processor to receive an error signal produced by comparing a driving signal applied to a transmitter to an output signal generated by the transmitter in response to the driving signal, to decompose the error signal into a linear component having a first memory depth and a nonlinear component having one or more polynomial orders and a second memory depth that is less than the first memory depth, to compute first coefficients, up to the first memory depth, of a linear predistortion kernel for application to the driving signal so as to compensate for the linear component of the error signal, to compute second coefficients for the one or more polynomial orders, up to the second memory depth, of a nonlinear predistortion kernel for application to the driving signal so as to compensate for the nonlinear component of the error signal, and to optimize operation of the transmitter using the first and second coefficients.
- The present invention will be more fully understood from the following detailed description of the embodiments thereof, taken together with the drawings in which:
-
FIG. 1 is a block diagram that schematically illustrates a communication system with digital predistortion (DPD), in accordance with an embodiment of the invention; -
FIG. 2 is a block diagram that schematically illustrates components involved in adjustment of a transmitter with DPD, in accordance with an embodiment of the invention; -
FIG. 3 is a flow chart that schematically illustrates a method for optimizing DPD coefficients, in accordance with an embodiment of the invention; -
FIG. 4 is a schematic plot of SNR as a function of the number of taps in a linear component of a DPD circuit, in accordance with an embodiment of the invention; -
FIG. 5 is a schematic plot of SNR as a function of the numbers of taps and nonlinear orders in a nonlinear component of a DPD circuit, in accordance with an embodiment of the invention; and -
FIGS. 6A and 6B are schematic plots of SNR as a function of nonlinear order in an MZM, showing the effect of bias on nonlinear distortion, in accordance with an embodiment of the invention. - Embodiments of the present invention that are described herein address the problem of characterizing the nonlinear distortion in a communication system, in a way that enables efficient compensation for nonlinear effects in optimizing the operation of the system. (Although the description that follows focuses on nonlinear distortion introduced by a transmitter in such a system, the principles of the present invention may alternatively be applied in dealing with other nonlinearities.) This optimization may take the form, for example, of selecting the number of nonlinear orders to handle in designing an equalization filter, as well as selecting the memory depth (i.e., the number of taps or coefficients) of the filter that should be applied to each order. Additionally or alternatively, such optimization may be applied in setting coefficient values and other operating parameters in order to improve the performance of an existing circuit.
- A variety of methods are known in the art for creating mathematical models of nonlinear effects. Some methods of this sort use Volterra series, which is an expansion of a Taylor series to include temporal effects. Discrete-time, real-valued Volterra series have the following general form, which assumes a uniform memory depth for all the nonlinear orders:
-
- In this formula, x(n) is the input to the transmitter, y(n) is the output, K is the number of nonlinear orders in the model, and M is the memory depth, which is assumed to be constant for all orders. The set of coefficients of the Volterra series {hk(tl . . . tk)} is also referred to as the kernel, whose values characterize the nonlinear response of the transmitter (or other system components that are subjects of the model). Due to cross-terms in the series, the size of the kernel ρ grows rapidly with the number of nonlinear orders and the memory depth. Specifically,
-
- In working with bandwidth-limited systems, such as communication links, however, the inventors realized that the actual memory depth that is required in the higher orders (k>1) of the kernel decreases rapidly with increasing order. In other words, the number of coefficients M(k) that is required to represent each of the nonlinear components k of distortion in the transmitter decreases monotonically with increasing k:
-
M(k i)>M(k j)∀k i<k j - Moreover, in many practical applications, transmitters can be represented accurately by a Volterra series that is constrained by the assumption that all of the higher polynomial orders have the same memory depth, and thus the same number of coefficients M(k>1), which is less than M(1). These constraints reduce the number of degrees of freedom sufficiently so that the computation of the optimal kernel {hk(t1 . . . tk)} becomes mathematically tractable, without substantially degrading the accuracy of the model. Alternatively, this sort of analysis may be applied with fewer constraints, though at the expense of higher computational complexity.
- On this basis, the Volterra series can be decomposed into linear and nonlinear parts, and the corresponding linear and nonlinear components of the kernel can be defined and optimized separately. For example, using M(1)=3, M(k>1)=1 and K=2, the Volterra series becomes:
-
- Using this analysis, some embodiments of the present invention define and apply two different compensators: a simple linear compensator with longer memory, and a nonlinear compensator with shorter memory. Although the nonlinear compensator in the above example includes only a single nonlinear order (k=2), higher orders can be incorporated in the nonlinear compensator in similar fashion.
- Distortion compensation can be applied either by an equalizer on the receive side of the communication line or by predistortion on the transmit side. Although receive-side equalization may be preferable when distortion is linear, the inventors have found that the most efficient and effective way to compensate for nonlinear distortion originating in the transmitter is by adding digital predistortion (DPD) to the input signal that is used to drive the transmitter. In this case, the method of analysis described above is reversed in order to find the coefficients of a predistortion kernel for application to the input driving signal x(n), to give a predistorted digital signal {circumflex over (x)}(n) for input to the transmitter. The predistortion kernel is optimized, as described below, so as to minimize the error between the original input x(n) and the transmitter output signal y(n).
- Embodiments of the present invention use these principles of analysis and predistortion in optimizing the operation of a transmitter. For this purpose, an error signal is produced (either by actual measurement or simulation) by comparing a driving signal x(n) that is applied to the transmitter to an output signal y(n) that is generated by the transmitter in response to the driving signal. A processor decomposes the error signal into a linear component having a given memory depth and a nonlinear component having one or more polynomial orders and a memory depth that is smaller than that of the linear component. The processor computes coefficients, up to the given memory depth, of a linear predistortion kernel for application to the driving signal. The processor also computes coefficients for the one or more polynomial orders, up to the smaller memory depth, of a nonlinear predistortion kernel.
- Application of the linear and nonlinear predistortion kernels to the driving signal compensates respectively for the linear and nonlinear component of the error signal. The characteristics of these kernels, in terms of their memory depths, number of significant nonlinear orders, and coefficient values, are used, as noted earlier, in optimizing the operation of the transmitter, for example in selecting filter characteristics the design phase of the transmitter or in setting operating parameters of an existing transmitter of system.
-
FIG. 1 is a block diagram that schematically illustrates acommunication system 20 with digital predistortion (DPD), in accordance with an embodiment of the invention. This figure shows a typical operating environment in which principles of the present invention may be implemented. In the pictured example, anoptical transmitter 22 transmits a digital signal to anoptical receiver 24 over acommunication link 26, such as a fiberoptic cable.Optical transmitter 22 is assumed to include an optoelectronic component, such as a VCSEL or MZM, whose modulation gives rise to nonlinear distortion at the transmitter output. Alternatively, the principles of the present invention may be applied to transmitters that operate over links of other sorts, such as RF cables and wireless links. - A
signal generator 28 insystem 20 receives digital data for transmission overlink 26, digitally modulates the data to generate a corresponding sequence of symbols, and converts the sequence of symbols into a digital driving signal having the form of a stream of digital samples x(n), as is known in the art. The digital samples are input to anonlinear DPD block 30, which applies a digital predistortion to the samples, including both linear and nonlinear components, as explained above. The structure and operation ofblock 30 are described in greater detail hereinbelow. An analog drive circuit converts the predistorted sample stream into an analog driving signal, which it applies to drivetransmitter 22. - A
sensor 34, comprising a suitable photodiode, for example, samples the optical signals that that are output fromtransmitter 22 and generates a corresponding stream of output samples y(n). Asystem analyzer 36 produces an error signal by comparing the samples x(n) of the driving signal that is applied totransmitter 22 to the samples y(n) of the output signal. As explained above,system analyzer 36 decomposes the error signal into linear and nonlinear components, and based on these components, computes the coefficient values of the linear and nonlinear predistortion kernels for application bynonlinear DPD block 30 to the samples x(n).System analyzer 36 typically performs this kernel computation as a preliminary step, beforetransmitter 22 begins actual data transmission, in order to set the coefficient values inDPD block 30 and/or set other operating parameters oftransmitter 22. Additionally or alternatively,system analyzer 36 may continue to evaluate and update the coefficient values during operation ofsystem 20 in order to compensate for changes in the distortion characteristics of the transmitter over time. - In the pictured example,
system analyzer 36 comprises a general-purpose computer processor 38, along with amemory 40, which records the samples and the error signal that are used by the processor in its kernel computation.Processor 38 performs the functions that are described herein under the control of software, which may be downloaded to the system analyzer in electronic form, over a network, for example. Additionally or alternatively, the software may be stored in tangible, non-transitory computer-readable media, such as optical, magnetic, or electronic memory media. Further additionally or alternatively, at least some of the functions ofprocessor 38 may be performed by a digital signal processor or by digital logic circuits, which may be hard-wired or programmable. -
FIG. 2 is a block diagram that schematically shows details of the structure and adjustments ofnonlinear DPD block 30, in accordance with an embodiment of the invention. In this embodiment,DPD block 30 comprises a look-up table (LUT) 46 with entries based on the nonlinear kernel coefficients, and a finite impulse response (FIR)filter 48 with taps based on the linear kernel coefficients. In other words,DPD block 30 first looks up each successive sample or sub-sequence of samples (with the number of samples depending on the memory depth M(k>1)) inLUT 46 to read out linearized sample values. The DPD block then inputs these linearized samples toFIR 48, in which the linearized samples are multiplied by the linear kernel coefficients to generate respective tap products, which are summed to generate the predistorted samples for application totransmitter 22. - The elements of
DPD block 30 in this embodiments, includingfilter 48, are typically implemented in digital hardware logic, which may be hard-wired or programmable, along with a suitable memory to holdLUT 46. Alternatively, at least some of the functions of DPD block may be implemented in a programmable DSP or in software or firmware running on a microcontroller or other processor. - At the design stage of
DPD block 30, the value of the memory depth M(1) of the linear part of the DPD can be optimized, in the sense that the number of taps inFIR filter 48 is chosen so as to be sufficient to achieve a certain level of SNR, without exceeding the number of taps necessary for this purpose. The value of the memory depth M(k>1) for the nonlinear part of the DPD is likewise optimized. Because the optimal value of M(k>1) is generally small (for example, equal to 2 or 3), the number of entries inLUT 46 is also manageable, with one table entry for each sub-sequence of M(k>1) input samples. Another advantage of this approach is that because all the nonlinear orders are constrained to have the same memory depth, the size of the LUT is not affected by the number of nonlinear orders that are included in the kernel. - Although
FIG. 2 shows a particular serial implementation ofDPD block 30, other implementations can be used in alternative embodiments of the present invention. For example, when the nonlinear component of the distortion is small,FIR 48 may precedeLUT 46 in the processing order. As another example, the input samples may be applied toLUT 46 andFIR 48 in parallel, and the outputs may then be summed. Further alternatively,DPD block 30 may apply other filtering architectures that are known in the art, such as frequency-domain filtering. - As explained above,
system analyzer 36 computes an error signal based on the difference between the respective samples output bysignal generator 28 and bysensor 34, and decomposes the error signal into linear and nonlinear components.Processor 38 uses these error components in computing the linear and nonlinear parts of the predistortion kernel. For example,processor 38 may apply aWiener filter 42 in computing the linear predistortion coefficients, while applying apolynomial Volterra filter 44, as defined above, in computing the nonlinear predistortion coefficients. The kernel computations may be carried out using any suitable method of optimization that is known in the art, typically using the mean squared error (MSE) as the cost function. For example,processor 38 may apply a single-step linear least squares estimation algorithm for this purpose, or an adaptive least mean squares algorithm. Alternatively, the kernel computations may use other cost functions, such as the bit error rate (BER), as well as other sorts of estimation algorithms that are known in the art, such as recursive least squares (RLS) estimation. -
FIG. 3 is a flow chart that schematically illustrates a method for optimizing DPD coefficients, in accordance with an embodiment of the invention. For the sake of concreteness and clarity, it is assumed that this method is carried out bysystem analyzer 36 insystem 20. The principles of this method may similarly be applied, mutatis mutandis, in correcting for nonlinearities in other systems, as well as in optimizing other operating parameters. - The method of
FIG. 3 begins with a process ofkernel estimation 50, in whichprocessor 38 computes coefficients of the linear and nonlinear distortion kernels, as explained above.Processor 38 may repeat the computation under different system conditions and using different values of the Volterra series parameters M(1), M(k>1) and K, for the purpose of extractingmodel properties 52. This step enables the designer or system operator to observe and understand the nonlinearities insystem 20 and how these nonlinearities are affected by the various parameters. -
Model properties 52 also enable the designer or system operator to choose the parameter values that will best mitigate nonlinear effects. In the present example, the model properties indicate anoptimal memory depth 54 for purposes of mitigating distortion inDPD block 30. Specifically, the model properties enable the designer or operator to find the optimal number of taps to include inFIR filter 48, as well as the optimal number of coefficients (and hence input samples) to use in generatingLUT 46. The model properties can also be used to find an optimalnonlinear order 56 to use in computing the entries ofLUT 46. As noted earlier, the numbers of taps, coefficients and orders are “optimal” in the sense that they are sufficient to achieve a certain level of SNR, without exceeding the number of taps necessary for this purpose, and thus represent the best tradeoff between the cost and performance of the target system. -
FIG. 4 is a schematic plot of SNR as a function of the number of taps (coefficients) inlinear FIR filter 48 inDPD block 30, in accordance with an embodiment of the invention. The SNR in this example represents the difference between the optimal non-distorted signal and the equalized received signal, and is formulated as: -
- wherein E[⋅] is the expectation operator. The plot in
FIG. 4 shows that the optimal number of taps is six, since the SNR increases significantly with each tap that added up to the sixth, but only negligibly with the seventh and subsequent taps. These circuit characteristics, however, are presented solely by way of example, to illustrate the optimization of a certain type of transmitter design; and other systems and components will typically exhibit different characteristics. -
FIG. 5 is a schematic plot of SNR as a function of the numbers of taps and nonlinear orders used in generatingLUT 46 inDPD block 30, in accordance with an embodiment of the invention. The starting point for this plot is the SNR level of about 22 dB that is achieved by application ofFIR 48, as illustrated inFIG. 4 . As in the case ofFIG. 4 , the characteristics inFIG. 5 are shown solely by way of example. The characteristics in this case were computed subject to the constraint that all orders share the number of taps (i.e., have the same memory depth). Alternatively, plots of this sort can be computed and used without this constraint, though at the cost of greater complexity. - In the example shown in
FIG. 5 , the optimal number of nonlinear taps is three, since SNR increases significantly up to the third tap but only negligibly thereafter. Incorporating seven nonlinear orders gives the greatest improvement in the SNR. Although this high order of nonlinearity complicates the computation of the entries inLUT 46, the computation can be performed offline and does not increase the actual size of the LUT. -
FIGS. 6A and 6B are schematic plots of SNR as a function of nonlinear order in an MZM, showing the effect of bias on nonlinear distortion, in accordance with an embodiment of the invention. The MZM comprises a Mach-Zender interferometer with two parallel legs, in an electro-optical medium that applies a variable phase shift to (at least) one of the legs as a function of the applied voltage. The MZM is inherently nonlinear, with a sinusoidal response to variations in the applied voltage level about the default bias (Vπ/2). In the embodiment ofFIGS. 6A /B,system analyzer 36 computes the nonlinear predistortion kernel of an MZM in order to analyze its voltage response and bias voltage. - Since the sine is an odd function, its Volterra series representation should include only odd polynomial orders. This situation is illustrated in
FIG. 6A , in which the bias voltage of the MZM is set properly to Vπ/2. Thus, as expected, only the odd polynomial orders (mainly the third and fifth orders) contribute to the distortion. On the other hand, inFIG. 6B , the bias voltage is offset from Vπ/2, with the result that the contributions of the third and fifth orders are reduced, while the sixth polynomial order makes a significant contribution to the kernel. This effect can be seen in aregion 60 of the two plots. - Thus, the nonlinear coefficients derived by
system analyzer 36 provide an estimate of the dependence of the nonlinear coefficients on the operating parameters of the transmitter—specifically on the bias voltage in the present example. Based on this information, a system operator will adjust the bias voltage of the MZM in order to achieve the desirable characteristic shown inFIG. 6A . - By the same token, the system designer or operator may observe the effect of the range of modulation voltages on the nonlinear kernel coefficients computed by the system analyzer. As long as the distortion can be properly compensated by
DPD block 30, the voltage range can be increased in order to increase the depth of modulation and thus support higher-order digital modulation schemes. The limit of the modulation range may be set at the point at which it is no longer possible to achieve the desired SNR due to excessive nonlinear distortion. More generally, the limit of the modulation range is chosen based on the optimal tradeoff between cost and performance, which can include additional factors besides the distortion. - The techniques of nonlinear analysis that are described above are not limited to these specific operating parameters and DPD coefficients. Other applications will be apparent to those skilled in the art after reading the above description and are considered to be within the scope of the present invention. For example, estimation of the nonlinear kernel coefficients may be used in defining physical link properties, including nonlinear reflections. The compatibility of a particular physical link with a given transmitter and receiver can be analyzed in this fashion, and the link properties can then be modified as necessary to rectify problems that are detected. As another example, estimation of the nonlinear kernel coefficients can be used to analyze and optimize the equalization that is required to compensate for instabilities (such as temperature-related drift) of an external optical modulator, such as a micro-ring modulator.
- It will thus be appreciated that the embodiments described above are cited by way of example, and that the present invention is not limited to what has been particularly shown and described hereinabove. Rather, the scope of the present invention includes both combinations and subcombinations of the various features described hereinabove, as well as variations and modifications thereof which would occur to persons skilled in the art upon reading the foregoing description and which are not disclosed in the prior art.
Claims (29)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US16/779,610 US10756682B1 (en) | 2019-02-04 | 2020-02-02 | Compensating for transmitter nonlinearities |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US201962800579P | 2019-02-04 | 2019-02-04 | |
US16/779,610 US10756682B1 (en) | 2019-02-04 | 2020-02-02 | Compensating for transmitter nonlinearities |
Publications (2)
Publication Number | Publication Date |
---|---|
US20200252032A1 true US20200252032A1 (en) | 2020-08-06 |
US10756682B1 US10756682B1 (en) | 2020-08-25 |
Family
ID=71837928
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/779,610 Active US10756682B1 (en) | 2019-02-04 | 2020-02-02 | Compensating for transmitter nonlinearities |
Country Status (1)
Country | Link |
---|---|
US (1) | US10756682B1 (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114301521A (en) * | 2021-12-23 | 2022-04-08 | 中国电子科技集团公司第十四研究所 | Nonlinear predistortion method for microwave photon signal generation link |
US11342999B2 (en) * | 2019-11-15 | 2022-05-24 | Cisco Technology, Inc. | Method and apparatus for pre-distorting an input signal for an optical transmitter |
US11356181B2 (en) * | 2017-06-15 | 2022-06-07 | Nokia Of America Corporation | Backward propagation with compensation of some nonlinear effects of polarization mode dispersion |
US11451419B2 (en) | 2019-03-15 | 2022-09-20 | The Research Foundation for the State University | Integrating volterra series model and deep neural networks to equalize nonlinear power amplifiers |
US20230130805A1 (en) * | 2021-10-22 | 2023-04-27 | Qualcomm Incorporated | Directional digital pre-distortion for inter user equipment interference mitigation |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8767857B2 (en) * | 2009-05-14 | 2014-07-01 | Seyed Aidin BASSAM | Multi-cell processing architectures for modeling and impairment compensation in multi-input multi-output systems |
US10284356B2 (en) * | 2011-02-03 | 2019-05-07 | The Board Of Trustees Of The Leland Stanford Junior University | Self-interference cancellation |
US10263728B1 (en) * | 2014-03-24 | 2019-04-16 | L-3 Communications Corp. | Combining time-varying non-linear distortion with a communications signal |
CN107786174B (en) * | 2016-08-26 | 2021-02-05 | 中颖电子股份有限公司 | Circuit and method for predistortion |
US9935810B1 (en) * | 2017-03-07 | 2018-04-03 | Xilinx, Inc. | Method and apparatus for model identification and predistortion |
-
2020
- 2020-02-02 US US16/779,610 patent/US10756682B1/en active Active
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11356181B2 (en) * | 2017-06-15 | 2022-06-07 | Nokia Of America Corporation | Backward propagation with compensation of some nonlinear effects of polarization mode dispersion |
US11451419B2 (en) | 2019-03-15 | 2022-09-20 | The Research Foundation for the State University | Integrating volterra series model and deep neural networks to equalize nonlinear power amplifiers |
US11855813B2 (en) | 2019-03-15 | 2023-12-26 | The Research Foundation For Suny | Integrating volterra series model and deep neural networks to equalize nonlinear power amplifiers |
US11342999B2 (en) * | 2019-11-15 | 2022-05-24 | Cisco Technology, Inc. | Method and apparatus for pre-distorting an input signal for an optical transmitter |
US20230130805A1 (en) * | 2021-10-22 | 2023-04-27 | Qualcomm Incorporated | Directional digital pre-distortion for inter user equipment interference mitigation |
CN114301521A (en) * | 2021-12-23 | 2022-04-08 | 中国电子科技集团公司第十四研究所 | Nonlinear predistortion method for microwave photon signal generation link |
Also Published As
Publication number | Publication date |
---|---|
US10756682B1 (en) | 2020-08-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10756682B1 (en) | Compensating for transmitter nonlinearities | |
US20210006286A1 (en) | Systems and methods for intelligently-tuned digital self-interference cancellation | |
US7429892B2 (en) | Model based distortion reduction for power amplifiers | |
US11476809B2 (en) | Polyphase digital signal predistortion in radio transmitter | |
US20210075583A1 (en) | Systems and methods for efficiently-transformed digital self-interference cancellation | |
US9413302B2 (en) | Digital predistortion apparatus and method | |
US10230353B2 (en) | Nonlinear signal filtering | |
US20130089169A1 (en) | Low-Power and Low-Cost Adaptive Self-Linearization System with Fast Convergence | |
US20090121788A1 (en) | System and method for self-cancellation of nth-order intermodulation products | |
CN113875170A (en) | Optical transmission characteristic compensation method and optical transmission characteristic compensation system | |
US7693672B2 (en) | Adaptive self-linearization | |
US7688235B2 (en) | Composite analog to digital receiver with adaptive self-linearization | |
US20210384891A1 (en) | Method and apparatus for nonlinear signal processing | |
US7917337B2 (en) | Adaptive self-linearization with separation filter | |
EP2067252B1 (en) | Adaptive self-linearization | |
Sena et al. | An autonomous identification and pre-distortion scheme for cognitive transceivers using Bayesian optimization | |
US11901951B2 (en) | Distorter coefficient updating apparatus, method and digital predistortion apparatus | |
US11683093B2 (en) | Wavelength dispersion compensation apparatus, optical receiving apparatus, wavelength dispersion compensation method and computer program | |
US11342999B2 (en) | Method and apparatus for pre-distorting an input signal for an optical transmitter | |
KR102177925B1 (en) | Method and appratus for electronic nonlinear equalization in intensity-modulation/direct-detection optical communication systems | |
KR100366289B1 (en) | Predistorter design and method using stochastic gradient method and indirect learning architecture | |
US20230238991A1 (en) | Communications device and method for compensating frequency response distortion of communications device | |
KR100625445B1 (en) | Prediction apparatus with a varying order and its control method | |
CN104104633A (en) | Non-linear compensation device, non-linear compensation method, transmitter and communication system | |
KR101105903B1 (en) | Apparatus and method for digital predistortion using adaptive noise cancelation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: MELLANOX TECHNOLOGIES, LTD., ISRAEL Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:FAIG, HANANEL;COHEN, SHAI;GANTZ, LIRON;SIGNING DATES FROM 20200129 TO 20200130;REEL/FRAME:051694/0688 |
|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 4 |