Modeling devices, computational methods, and non-transitory computer-readable storage media

Abstract

This invention relates to modeling apparatus, computational methods, and non-transitory computer-readable storage media. A modeling apparatus is provided that performs calculations using an amplifier model that models an amplifier, wherein the internal states of the amplifier's influence distortion characteristics are changing. The amplifier model includes: a plurality of computational models that model the amplifier in different internal states; and a combiner that combines the plurality of computational models at a combination ratio corresponding to the changed internal states.

Description

Technical Field

[0001] This disclosure relates to a modeling apparatus, a computational method, and a non-transitory computer-readable storage medium. Background Technology

[0002] Amplifiers exhibit distortion characteristics. These distortion characteristics are also known as nonlinear characteristics. Distortion characteristics are represented, for example, by AM-AM or AM-PM characteristics.

[0003] In an amplifier, the internal states affecting distortion characteristics can change. These internal states can be, for example, the state in which Idq drift occurs, a phenomenon caused by changes in the idle current Idq flowing through the amplifier. The generation of Idq drift alters the amplifier's distortion characteristics.

[0004] Idq drift is a phenomenon caused by the capture of charge carriers that have entered the impurity energy levels of a semiconductor, and it is caused in devices using compound semiconductors such as GaN, particularly in high electron mobility transistors (HEMTs). Fluctuations in Idq drift alter the distortion characteristics of an amplifier.

[0005] A method for compensating for distortion characteristics caused by Idq drift is disclosed in Japanese Patent Application Publication Nos. 2017-220744 and 2014-17670. Summary of the Invention

[0006] In various situations, it is desirable to simulate the distortion characteristics of an amplifier. For example, when developing distortion compensation devices for amplifiers, simulating the amplifier's distortion characteristics allows for the evaluation of distortion compensation performance, thereby making the development process more efficient. Furthermore, it makes it possible to design the amplifier's peripheral circuitry by taking the amplifier's distortion characteristics into account in advance.

[0007] However, as mentioned above, the distortion characteristics of an amplifier can change depending on the amplifier's internal state. Amplifiers with varying distortion characteristics are difficult to model using a single amplifier model. To properly simulate the distortion characteristics of an amplifier, it is desirable to have an amplifier model that appropriately models the changes in the amplifier's internal state that affect its distortion characteristics.

[0008] According to one aspect of this disclosure is a modeling apparatus. The modeling apparatus of this disclosure is a modeling apparatus that performs calculations using an amplifier model that models an amplifier, wherein the internal state of the amplifier's influence distortion characteristics is changing, wherein the amplifier model includes: a plurality of computational models that model the amplifier in different internal states; and a combiner that combines the plurality of computational models with a combination ratio corresponding to the changed internal states.

[0009] According to another aspect of this disclosure is a calculation method. This calculation method is a method for modeling an amplifier whose internal states affecting distortion characteristics are changing. The calculation method includes the step of a combiner combining multiple calculation models of the amplifier in different internal states according to a combination ratio based on the changing internal states.

[0010] According to another aspect of this disclosure, there is a non-transitory computer-readable medium storing a computer program. The computer program of this disclosure is a computer program for calculating and modeling an amplifier whose internal states affecting distortion characteristics are changing. The computer program causes a computer to perform a process comprising: combining multiple computational models of the amplifier in different internal states with combination ratios corresponding to the changed internal states. Attached Figure Description

[0011] Figure 1 This is a block diagram of the modeling device according to the first embodiment;

[0012] Figure 2 It is a configuration diagram of the communication equipment;

[0013] Figure 3 This is a hardware configuration diagram of the modeling equipment;

[0014] Figure 4 This is an explanatory diagram of the amplifier model;

[0015] Figure 5 The diagram illustrates the relationship between input power and internal state parameters.

[0016] Figure 6 The diagram illustrates the relationship between internal state parameters and the computational model.

[0017] Figure 7 It is an explanatory diagram used to generate the computational model;

[0018] Figure 8 This is a flowchart of the process for generating the amplifier model;

[0019] Figure 9 This is an illustration of the α[n] generator;

[0020] Figure 10 This is a block diagram of the modeling device according to the second embodiment;

[0021] Figure 11 This is an illustrative diagram of another amplifier model;

[0022] Figure 12 The diagram illustrates the relationship between internal state parameters and the computational model.

[0023] Figure 13This is a flowchart of the process for generating the amplifier model; and

[0024] Figure 14 It is α p [n] An explanatory diagram of generators. Detailed Implementation

[0025] [Description of embodiments of this disclosure]

[0026] (1) The modeling apparatus according to the embodiment uses an amplifier model to perform calculations, which models an amplifier whose internal state affecting distortion characteristics is changing. Distortion characteristics are also referred to as nonlinear characteristics. The internal state is, for example, a state that generates Idq drift. The amplifier model includes: multiple computational models that model amplifiers in different internal states; and a combiner that combines the multiple computational models at a combination ratio corresponding to the changed internal state. An amplifier model corresponding to the changed internal state can be obtained by combining the computational models at a combination ratio corresponding to the changed internal state.

[0027] (2) The modeling device may also include a generator configured to generate parameters indicating the internal state. In this case, the modeling device is capable of generating parameters indicating the internal state. The generator may be provided to an external device of the modeling device. The combination ratio may be based on the parameters.

[0028] (3) Preferably, the combination ratio is determined based on a parameter indicating the internal state. Preferably, the parameter is calculated based on the level of the signal to be amplified by the amplifier. In this case, the parameter can indicate the internal state that changes depending on the level of the signal to be amplified. Since the combination ratio is determined based on the parameter indicating the internal state, a combination ratio corresponding to the level of the signal to be amplified is obtained.

[0029] (4) The parameters are preferably based on past values. Past values ​​of the parameters indicate past internal states. Therefore, the parameters can reflect past internal states. As a result, the memory effect of the amplifier is reflected in the parameters.

[0030] (5) Each of the plurality of computational models preferably expresses a first memory effect having a first response time in the amplifier. The parameter computational model used to calculate the parameters preferably expresses a second memory effect having a second response time in the amplifier that is longer than the first response time. In this case, the parameters become parameters that take into account the second memory effect.

[0031] (6) Preferably, the combination ratio is calculated based on the level of the signal to be amplified by the amplifier. In this case, a combination ratio corresponding to the level of the signal to be amplified is obtained.

[0032] (7) Preferably, the combination ratio is calculated using temperature-dependent parameters that vary depending on temperature conditions. In this case, the changes in amplifier characteristics due to temperature are supported.

[0033] (8) Multiple computational models can be two computational models. In this case, the amplifier model becomes simple.

[0034] (9) The modeling device may also include a selector. Multiple computational models may include three or more computational models. The presence of three or more computational models makes it possible to more appropriately represent the characteristics of the amplifier. The selector is preferably configured to select two or more computational models from the three or more computational models to be combined by the combiner. In this case, it is possible to select computational models that more appropriately represent the characteristics of the amplifier.

[0035] (10) The selector is preferably configured to select two or more computational models based on parameters indicating the internal state. In this case, the computational model suitable for the internal state is selected.

[0036] (11) The multiple computational models may include: a first computational model that models the amplifier in a first internal state; and a second computational model that models the amplifier in a second internal state different from the first internal state. The combination ratio preferably has a value corresponding to the transitional internal state between the first and second internal states. In this case, the amplifier model is able to express the distortion characteristics of the amplifier in the transitional internal state.

[0037] (12) The multiple computational models may include: a first computational model that models the amplifier in a first internal state; a second computational model that models the amplifier in a second internal state different from the first internal state; and a third computational model that models the amplifier in a third internal state different from the first and second internal states. The second internal state is preferably an intermediate internal state between the first and third internal states. In this case, computational models corresponding to at least three internal states can be used.

[0038] (13) The calculation method according to the embodiment is a calculation method for modeling an amplifier whose internal state of influence distortion characteristics is changing. The calculation method includes the step of combining multiple calculation models of an amplifier in different internal states with a combination ratio corresponding to the changed internal state.

[0039] (14) The computer program according to an embodiment is a computational computer program for modeling an amplifier whose internal states affecting distortion characteristics are changing. The computer program causes a computer to perform a process comprising combining multiple computational models of the amplifier in different internal states with combination ratios corresponding to the changed internal states. The computer program is stored in a non-transitory computer-readable storage medium.

[0040] [Details of the embodiments of this disclosure]

[0041] [First Embodiment]

[0042] Figure 1 The figure shows a modeling device 100 according to the first embodiment. The modeling device 100 uses an amplifier model 200 that models the amplifier 10 to perform calculations. When an input signal u[n] is provided, the modeling device 100 calculates the output signal y[n] obtained by amplifying the input signal u[n] based on the amplifier model 200, and outputs the calculated output signal y[n].

[0043] The calculations performed by the modeling device 100 are calculations used to simulate, for example, signal amplification operations in amplifier 10. The simulation of signal amplification operations includes the simulation of the amplifier's distortion characteristics. The simulation of amplifier 10 makes it possible to evaluate the distortion compensation performance of amplifier 10, thereby making the development process more efficient. Furthermore, it is also possible to design the peripheral circuitry of amplifier 10 in advance, taking into account the distortion characteristics of amplifier 10.

[0044] Amplifier 10 is, for example, a power amplifier. Amplifier 10 is, for example, a gallium nitride (GaN)-HEMT. Amplifier 10 is not limited to GaN amplifiers, and can be a HEMT device using compound semiconductors such as aluminum nitride (AIN) or indium nitride (InN), or AlGaN, InAIN, or InGaN as a mixed crystal of AIN and InN.

[0045] The amplifier 10 using a compound semiconductor such as GaN has a transient response called Idq drift. Idq drift is a phenomenon in which the free current Idq decreases and the distortion characteristics change as charge carriers are trapped in the impurity energy levels of the semiconductor.

[0046] In devices that cause Idq drift, distortion changes immediately according to fluctuations in the electrical power of the signal. These fluctuations are particularly likely to occur in communication systems that alternate between time-division duplex (TDD) transmission and reception. It is difficult to model an amplifier that causes Idq drift using a single amplifier model because its distortion characteristics change.

[0047] The likelihood of Idq drift occurring depends on the input power (input signal level) and the degree of drift. Drift is more likely to occur as input power increases. In the state where Idq drift has occurred, the gain decreases in the region of low input power. In contrast, as input power decreases, the reduced idle current Idq begins to recover, and the gain recovers over time.

[0048] Typically, amplifier 10 exhibits distortion known as the memory effect. The memory effect is the phenomenon where the amplifier's output signal is influenced by past input signals. Memory effects can include memory effects with short response times (first memory effect with a first response time; short-term memory effect) and memory effects with long response times (second memory effect with a second response time longer than the first response time; long-term memory effect). Here, response time is the time it takes for the output to respond to changes in the input, and response time is also referred to as the time constant.

[0049] The characteristic change caused by Idq drift represents a state with a second memory effect having a long response time. Constructing an amplifier model that only expresses a first memory effect with a short response time is relatively easy, but it is difficult to construct a single amplifier model that expresses a second memory effect with a long response time in addition to the first memory effect with a short response time. In particular, modeling becomes more difficult when the extent of the second memory effect depends on the internal state of the amplifier, such as the degree to which Idq has decreased.

[0050] Therefore, this embodiment uses an amplifier model 200 that can appropriately express the characteristics of amplifier 10 based on changes in the internal state (the state of the second memory effect) of the generated state, such as Idq drift.

[0051] Amplifier 10 outputs an output signal y(t) obtained by amplifying the input signal u(t). Here, u(t) represents the signal at time t, which is a continuous value. That is, u(t) represents the input signal as a continuous value, and y(t) represents the output signal as a continuous value.

[0052] In the following text, the signal can be represented by a discrete value *[n] processed in digital circuitry. Here, *[n] is a complex baseband IQ signal sampled at time n×T in a system with a sampling interval T [seconds].

[0053] For example, x[n] is the input signal before distortion compensation, and x[n] = x I [n]+j×x Q [n] expression. Here, x I [n] is the real part (I-channel) of x[n], while x Q[n] is the imaginary part (Q-channel) of x[n]. In contrast, u[n] is the distortion-compensated input signal, and is expressed as u[n] = u I [n]+j×u Q [n] expression. Here, u I [n] is the real part (I-channel) of u[n], while u Q [n] is the imaginary part (Q-channel) of u[n]. Additionally, y[n] is the output signal, and is given by y[n] = y I [n]+j×y Q [n] expression. Here, y I [n] is the real part (I-channel) of y[n], and y Q [n] is the imaginary part (Q-channel) of y[n].

[0054] Amplifier 10 is used in, for example Figure 2 In the communication device 50 shown, amplifier 10 is, for example, a power amplifier that amplifies the signal to be transmitted from the communication device 50.

[0055] The communication device 50 includes a distortion compensation device 20. The distortion compensation device 20 uses digital signal processing to compensate for distortion. The distortion compensation device 20 can be constructed by a computer including a storage device (not shown) and a processor (not shown) connected to the storage device, or it can be constructed by wired logic circuitry.

[0056] The distortion compensation device 20 includes a distortion compensation processing unit 21 and an inverse characteristic estimation unit 22. When the distortion compensation device 20 is constructed by a computer, the distortion compensation processing unit 21 and the inverse characteristic estimation unit 22 are implemented by causing the processor to read and execute a computer program stored in a storage device. The computer program contains code that enables the computer to operate as the distortion compensation processing unit 21 and the inverse characteristic estimation unit 22.

[0057] Inverse characteristic estimation unit 22 estimates the inverse model representing the distortion compensation characteristic. The inverse model is configured, for example, as a distortion compensation function. The inverse model (distortion compensation function) indicates the inverse characteristic G of the amplification characteristic G of amplifier 10. -1 It can indicate the inverse characteristic G. -1 The inverse model is obtained as the inverse function of, for example, the inverse model of the characteristic G of amplifier 10. The inverse characteristic estimation unit 22 copies the estimated inverse model to the distortion compensation processing unit 21. The inverse model to be copied is specifically the parameters expressing the inverse model, and more specifically, the distortion compensation coefficients constituting the distortion compensation function.

[0058] Distortion compensation processing unit 21 uses the inverse model copied from inverse characteristic estimation unit 22 to process the input signal x[n]=x I [n]+j×xQ [n] undergoes pre-distortion compensation processing. Pre-distortion compensation processing involves applying a distortion compensation function to the input signal x[n]. The distortion compensation processing unit 21 then outputs the signal u[n] = u generated through the pre-distortion compensation processing. I [n]+j×u Q [n]. In the following text, the signal generated by the predistortion compensation process is sometimes referred to as the "compensated signal". Here, the input signal x[n] and the compensation signal u[n] are digital signals.

[0059] The communication device 50 includes DA converters (DACs) 32A and 32B, a quadrature modulator 33, a frequency converter 34, and a driver amplifier 35.

[0060] DACs 32A and 32B convert signal u[n] from a digital signal to an analog signal. Quadrature modulator 33 outputs a modulated signal obtained by quadrature modulation of the analog compensation signal (analog IQ baseband signal) output from DACs 32A and 32B. Frequency converter 34 is an up-converter and up-converts the modulated signal output from quadrature modulator 33. Driver amplifier 35 amplifies the up-converted modulated signal.

[0061] The signal output from the driver amplifier 35 is provided to amplifier 10 as the input signal u(t). Amplifier 10 outputs the signal y(t) obtained by amplifying the input signal u(t). Peripheral circuit 10A is disposed in the pre-stage of amplifier 10, while peripheral circuit 10B is disposed in the post-stage of amplifier 10. Each of peripheral circuits 10A and 10B includes electronic circuit elements such as resistors or capacitors.

[0062] The communication device 50 also includes a coupler 36, a variable attenuator 37, a quadrature demodulator 42, filters 41A and 41B, and analog-to-digital converters (ADCs) 40A and 40B.

[0063] Coupler 36 outputs an analog monitoring signal obtained from the output signal y(t) of monitoring amplifier 10. Frequency converter 38 is a down-converter and down-converts the analog monitoring signal provided from coupler 36 via variable attenuator 37. Quadrature demodulator 42 performs quadrature demodulation on the analog monitoring signal output from frequency converter 38.

[0064] Filters 41A and 41B are low-pass or band-pass filters. The demodulated signal output from the quadrature demodulator 42 is provided to ADCs 40A and 40B after passing through filters 41A and 41B. ADCs 40A and 40B convert the demodulated signal provided by the quadrature demodulator 42 from an analog signal to a digital signal. ADCs 40A and 40B convert the digital demodulated signal y[n] = y(t) corresponding to the output signal y(t) of amplifier 10. I [n]+j×y Q [n] is provided to the inverse characteristic estimation unit 22.

[0065] Inverse characteristic estimation unit 22 uses the estimated inverse characteristic G -1 The digital demodulated signal y[n] = y is derived from the output signal y(t) of amplifier 10. I [n]+j×y Q [n] Obtain the distortion compensation signal u[n]=u I [n]+j×u Q The replica signal u'[n] = u' I [n]+j×u' Q [n].

[0066] Inverse characteristic estimation unit 22 obtains the indication distortion compensation signal u[n] = u I [n]+j×u Q [n] and the replica signal u'[n]=u' I [n]+j×u' Q The error signal (u') between [n] I [n]-u I [n]、u' Q [n]-u Q [n]). The inverse characteristic estimation unit 22 successively updates the inverse characteristic G that constitutes the distortion compensation characteristic. -1 The parameters (distortion compensation coefficients) are adjusted to reduce the error signal. The updated distortion compensation coefficients are then copied to the distortion compensation processing unit 21.

[0067] Return to reference Figure 1 The modeling device 100 according to an embodiment includes an amplifier model 200 that models the amplifier 10 as described above. The amplifier model 200 has characteristics G for modeling the amplifier 10. The amplifier model 200 includes multiple computational models G1 and G2. As an example, Figure 1The amplifier model 200 shown includes two computational models G1 and G2. That is, the amplifier model 200 includes a first computational model G1 and a second computational model G2. The characteristic G of the amplifier model 200 is expressed as a composite characteristic of the computational models G1 and G2. Each of the computational models G1 and G2 is expressed by, for example, a function. The number of computational models can be three or more, and the characteristic G of the amplifier model 200 can be expressed as a composite characteristic of three or more computational models.

[0068] In this embodiment, the calculation of the first calculation model G1 is performed by the first arithmetic unit 210. The first arithmetic unit 210 applies the first calculation model G1 to the input signal u[n] and outputs the calculation result G1(u[·]). The calculation of the second calculation model G2 is performed by the second arithmetic unit 220. The second arithmetic unit 220 applies the second calculation model G2 to the input signal u[n] and outputs the calculation result G2(u[·]).

[0069] Amplifier model 200 also includes a combiner 230. Combiner 230 includes multipliers 231 and 232 and adder 233. Combiner 230 obtains the synthesis characteristics obtained by combining computational models G1 and G2 with a dynamically changing combination ratio. The synthesis characteristics of computational models G1 and G2 are characteristic G of amplifier model 200. Characteristic G changes according to the combination ratio.

[0070] In this embodiment, the combiner 230 combines the calculation result G1(u[·]) of the first arithmetic unit 210 and the calculation result G2(u[·]) of the second arithmetic unit 220 with a changed combination ratio. The synthesized function can be obtained by combining the coefficients of the functions constituting the calculation models G1 and G2 in advance with a combination ratio, and the obtained synthesized function can be applied to the input signal u[n].

[0071] Modeling device 100 includes a generator 300 for generating a parameter α[n] for determining a dynamically changing combination ratio. In this embodiment, generator 300 generates parameter α[n] based on an input signal u[n]. Here, as an example, α[n] is a real number ranging from 0 to 1. Amplifier model 200 obtains parameter α[n] from generator 300. Generator 300 can be provided to an external device of modeling device 100. In this case, modeling device 100 obtains α[n] generated by external generator 300 and provides the obtained α[n] to amplifier model 200.

[0072] In this embodiment, the dynamically changing combination ratio is represented by α[n]:1-α[n]. Therefore, the combination ratio is automatically determined when the parameter α[n] is dynamically determined. Here, α[n] indicates the ratio (weight) used for the first computational model G1, and the multiplier 231 multiplies the first computational model G1 by α[n] (see...). Figure 1 and Figure 4 Additionally, 1-α[n] indicates the ratio (weight) used for the second computational model G2, and multiplier 232 multiplies the second computational model G2 by 1-α[n] (see...). Figure 1 and Figure 4 The modeling device 100 includes a constant generator 250 for generating “1” in 1-α[n].

[0073] Adder 233 adds the outputs of multipliers 231 and 232. The result of the addition by adder 233 represents the calculation result (output signal y[n]) obtained using amplifier model 200.

[0074] In this embodiment, Figure 1 The modeling device 100 shown consists of, for example, the following components: Figure 3 The computer configuration shown includes a processor 101 and a storage device 102. The processor 101 is coupled to the storage device 102. The processor 101 is, for example, a central processing unit (CPU). The storage device 102 includes, for example, a primary storage device and a secondary storage device. The primary storage device is, for example, random access memory (RAM). The secondary storage device is, for example, a hard disk drive (HDD) or a solid-state drive (SSD). The modeling device 100 may be constructed using wired logic circuitry.

[0075] Storage device 102 stores a computer program 102A that enables the computer to operate as a modeling device 100. The computer program 102A is configured to cause processor 101 to execute amplifier model calculation processing 101A. The computer performing amplifier model calculation processing 101A acts as... Figure 1 The modeling device 100 shown is in operation.

[0076] The processor 101 reads the computer program 102A from the storage device 102 and executes the computer program 102A. The computer program 102A enables the computer to function as... Figure 1 The code for the operation of the amplifier model 200 and generator 300 included in the modeling device 100 shown.

[0077] The computer operating as modeling device 100 includes interface 103. Interface 103 includes at least one interface selected from the group consisting of: a communication interface for communicating with other computers, an input interface for connecting to an input device such as a keyboard or mouse, and an output interface for connecting to an output device such as a display.

[0078] Analog data used in simulating amplifier 10 is input to modeling device 100 via interface 103. The analog data is, for example, the input signal u[n] to amplifier 10. The analog data may include the temperature of amplifier 10.

[0079] According to the embodiment, amplifier model 200 is composed of... Figure 4 Equation (1) is presented in the text. Equation (1) is equivalent to Figure 1 The amplifier model 200 is shown. Additionally, in... Figure 4 In this embodiment, the amplifier model 200 is represented as a model obtained by combining and calculating models G1 and G2 with a combination ratio determined by α[n].

[0080] Computational models G1 and G2 model amplifier 10 in different internal states. Specifically, the first computational model G1 models the characteristics of amplifier 10 in a first internal state, while the second computational model G2 models the characteristics of amplifier 10 in a second internal state. The first internal state and the second internal state are different.

[0081] The internal state is not particularly limited, as long as it is the state of the amplifier that affects the distortion characteristics of amplifier 10. For example, the internal state is based on the value of the idle current Idq flowing through amplifier 10. Idq changes according to the amplitude of the input power (input signal level). In this embodiment, as the internal state parameter α[n] gets closer to 0, it indicates a larger decrease in Idq, and as the internal state parameter α[n] gets closer to 1, it indicates a smaller decrease in Idq.

[0082] Figure 5 The diagram illustrates the relationship between the input power value and the internal state parameter α[n] of the generated state based on Idq drift. Figure 5 As shown, when a large input power is applied to amplifier 10, an Idq drift is generated, and α[n] becomes closer to 0. In contrast, as the input power decreases, no Idq drift is generated, and α[n] becomes closer to 1. As seen above, the internal state parameter α[n] changes depending on the input power value.

[0083] In this embodiment, different computational models G1 and G2 are prepared corresponding to different generated states of Idq drift. The amplifier model 200 representing distortion characteristics in any generated state of Idq drift is obtained by combining computational models G1 and G2 according to the changed Idq drift state (internal state). In this embodiment, the combination of computational models G1 and G2 is a linear combination of computational models G1 and G2 (see [link to documentation]). Figure 4 The expression (1) in the text.

[0084] With the first computational model Gl The corresponding first internal state indicates that no Idq drift has been generated in amplifier 10. For example... Figure 6 As shown, the first internal state is the state where the internal state parameter α[n] = 1. For example... Figure 7 As shown, in state S1 where Idq drift is not generated, the coefficients constituting the first calculation model G1 are determined based on the input and output signals of amplifier 10.

[0085] The second internal state indicator, corresponding to the second computational model G2, generates a state in amplifier 10 with an Idq drift greater than that in the first internal state. For example... Figure 6 As shown, the second internal state is the state where the internal state parameter α[n] = 0. For example... Figure 7 As shown, in state S2 where Idq drift is generated, the coefficients constituting the second calculation model G2 are determined based on the input and output signals of amplifier 10.

[0086] Figure 8 This involves generating computational models G1 and G2 that constitute amplifier model 200. First, an input signal is provided to amplifier 10, and the output signal is measured (step S11). The input signal is, for example, a communication signal transmitted by communication device 50. Through the measurement in step S11, a data pair of the input and output signals of amplifier 10 is obtained.

[0087] In step S12, the internal state parameter α[n] is calculated based on the input signal level provided to amplifier 10. The internal state parameter α[n] is calculated by α[n] generator 300 based on the level of input signal u[n].

[0088] In step S13, the time T1 when the value of α[n] becomes 1 is determined (see...). Figure 7 Time T1 is the time when the speculative amplifier 10 is in its first internal state without generating Idq drift.

[0089] In step S14, the coefficients h of the first computational model G1 are determined from the data of the input and output signals during a predetermined past time period starting from time T1. m,l,k Through this process, the first computational model G1 was obtained.

[0090] In step S15, the time T2 when the value of α[n] becomes 0 is determined (see...). Figure 7 Time T2 is the time when the speculative amplifier 10 is in the second internal state that generates the Idq drift.

[0091] In step S16, the coefficients g of the second calculation model G2 are determined from the data of the input and output signals during a predetermined past time period starting from time T2. m,l,kThis process yielded the second computational model, G2.

[0092] Each of the computational models G1 and G2 is an equation representing the nonlinear characteristics (distortion characteristics) of amplifier 10. Typical expressions used to model the amplifier can be used as expressions for computational models G1 and G2. These expressions can be, for example, generalized memory polynomials, Winer-Hammerstein models, Sarah models, or Volterra series. Figure 4 The computational models G1 and G2 shown are expressed by generalized memory polynomials. Conventional models such as generalized memory polynomials, Winer-Hammerstein models, Sarah models, or Volterra series can express the nonlinear characteristics of an amplifier in a certain internal state, but they cannot adequately express the phenomenon that the nonlinear characteristics (distortion characteristics) change depending on the changes in the amplifier's internal state.

[0093] exist Figure 4 In the first computational model G1 shown (see equation (2)), h m,l,k These are the coefficients representing the first computational model G1, and the input signal u[·] is multiplied by h. m,l,k The values ​​of M1 and M2, defining the range of m, are set based on the length of the first response time (short response time) of the first memory effect (short-term memory effect), and the value of L, defining the range of l, is defined. 1,m and L 2,m The value of . That is, the first computational model G1 represents the first memory effect with a short first response time.

[0094] exist Figure 4 In the second computational model G2 shown (see equation (3)), g m,l,k These are the coefficients representing the second computational model G2, and the input signal u[·] is multiplied by g. m,l,k The values ​​of M3 and M4, defining the range of m, and L, defining the range of l, are determined based on the length of the first response time (short response time) of the first memory effect (short-term memory effect). 3,m and L 4,m The value of . That is to say, the second computational model G2 expresses a first memory effect with a short first response time. The first response time in the first computational model G1 does not have to be the same as the first response time in the second computational model G2.

[0095] The first computational model G1 represents the characteristics of amplifier 10 in the first internal state (α[n] = 1), while the second computational model G2 represents the characteristics of amplifier 10 in the second internal state (α[n] = 0). However, neither computational model G1 nor G2 adequately represents the characteristics of amplifier 10 in any transitional internal state (0 < α[n] < 1) between the first and second internal states.

[0096] Therefore, in this embodiment, the combination ratios corresponding to arbitrary transitional internal states are used to calculate models G1 and G2. The amplifier model 200 obtained by combining the combination ratios corresponding to transitional internal states using models G1 and G2 can represent the characteristics of the amplifier 10 in transitional internal states. An extrapolation method can be used to obtain the combination ratios corresponding to internal states outside the range between the first and second internal states. In this embodiment, the combination ratio is calculated based on the level of the input signal u[n] (see...). Figure 1 ).

[0097] The combination ratio of computational models G1 and G2 can have values ​​corresponding to any transitional internal state between the first internal state and the second internal state (0 < α[n] < 1). In the transitional internal state, the state in which the Idq drift is generated is between the state in which the Idq drift is generated in the first internal state and the state in which the Idq drift is generated in the second internal state.

[0098] Figure 9 The diagram illustrates the α[n] generator 300. The α[n] generator 300 calculates a first parameter R[n] and a second parameter α[n] that indicate the internal state, and outputs the second parameter α[n] used to determine the combination ratio. The first parameter R[n], indicating the internal state, represents the state of the Idq drift.

[0099] according to Figure 9 Equation (4) in the equation calculates the first parameter R[n]. Equation (4) constitutes the parameter calculation model for calculating the parameter R[m]. As presented in equation (4), the first parameter R[n+1] at n+1 is calculated based on the level of the input signal u[n] at n. More specifically, the first parameter R[n+1] is calculated based on the past value (previous value) R[n] of the parameter and the input signal u[n].

[0100] In equation (4), R[n] and R[n+1] are m×1 scalar matrices. I is an m×m identity matrix. A is an m×m scalar matrix. B is an m×m scalar matrix and is the coefficient matrix multiplied by the input signal u[n]. Here, m is a positive integer, and the same applies to the description below. As m increases, the generated state of Idq drift is expressed more precisely.

[0101] Equation (4) indicates that as the level of the input signal u[n] increases, Idq drift is more likely to occur, and the first parameter R[n+1] increases. Equation (4) also indicates that as the level of the input signal u[n] decreases, Idq drift is less likely to occur, and the first parameter R[n+1] decreases.

[0102] In equation (4), when the input signal u[n] is assumed to be zero, R[n+1] = (IA)R[n]. In equation (4), -A represents the time constant for the recovery of the reduced idle current Idq, and expresses the recovery of the idle current Idq at regular intervals (sampling interval T[seconds]) when the input signal u[n] is zero. The value of each element in matrix A is set to be sufficiently small (e.g., approximately 1 / 1000 to 1 / 1000000). The value of each element in matrix A is set sufficiently small so that (IA) to be multiplied by R[n] is almost identical to the identity matrix, and the degree to which R[n] decreases from R[n] to R[n+1] is reduced. That is, the value of each element in matrix A is set sufficiently small so that the change of the first parameter R[n] becomes gradual. As seen above, the parameter calculation model indicated by equation (4) expresses a second memory effect with a long response time.

[0103] In this embodiment, the magnitude of each element in matrix A is set small enough that the second memory effect expressed by equation (4) has a longer response time than the first memory effect expressed by equations (2) and (3).

[0104] Here, A and B in equation (4) are set according to the physical properties and characteristics of amplifier 10. When the physical properties and characteristics of amplifier 10 are temperature-dependent, A and B are temperature-dependent parameters that change depending on the temperature conditions of amplifier 10. For example, the ease with which Idq drift is generated depends on the temperature of amplifier 10. Drift is more likely to occur at lower temperatures, but less likely to occur at higher temperatures.

[0105] In equation (4), temperature-related parameters A and B are used to calculate parameter R[n]. Temperature-related parameters A and B are variable, and their values ​​are adjusted by temperature-related parameter adjuster 310. Temperature-related parameter adjuster 310 adjusts temperature-related parameters A and B based on the temperature conditions provided as simulation data. In this embodiment, temperature-related parameters A and B are used to calculate the combination ratio. Therefore, the combination ratio is affected by temperature conditions.

[0106] exist Figure 9In equation (5), the first parameter R[n], which indicates the state of Idq drift, is transformed into a second parameter α[n], whose value is in the range of 0 and below 1. In equation (5), F(·) is a normalization function that normalizes the first parameter R[n] to the range of 0 and below 1. The second parameter α[n] is also a parameter indicating the internal state.

[0107] The normalization function F(·) is appropriately set such that as the value of the first parameter R[n] increases (as more Idq drifts are generated), α[n] becomes closer to 0, and as the value of the first parameter R[n] decreases (as fewer Idq drifts are generated), α[n] becomes closer to 1.

[0108] The generation of Idq drift can be expressed using internal states as follows. The initial internal state is defined as the state without Idq drift, obtained when the input signal level to the amplifier is zero for a sufficiently long period. As the input signal level increases, the internal state changes more significantly from the initial internal state. That is, the difference between the internal state when the input signal level is a first level and the initial internal state is greater than the difference between the internal state when the input signal level is a second level lower than the first level and the initial internal state. As the input signal level decreases, the internal state returns to the initial internal state over time. In other words, when the input signal level changes from the first level to a second level lower than the first level (e.g., 0V), the internal state returns to the initial internal state over time.

[0109] like Figure 4 As presented in the diagram, the change from the initial internal state in the second internal state corresponding to computational model G2 is greater than the change from the initial internal state in the first internal state corresponding to computational model G1. The combination ratio α is the combination ratio of computational model G1 relative to the sum of computational models G1 and G2. Figure 9 In equation (4), n corresponds to a certain time (first time), and n+1 corresponds to a time after the first time (second time). In this case, in equations (4) and (5), according to the first term of equation (4), when the input signal level at time n is zero, the combination ratio α (second combination ratio) at time n+1 is less than the combination ratio α (first combination ratio) at time n. Additionally, according to the second term of equation (4), when the input signal level at time n is the first level, the combination ratio α at time n+1 is greater than the combination ratio α at time n+1 when the input signal level at time n is the second level, which is smaller than the first level. Therefore, it becomes possible to model an amplifier with Idq drift.

[0110] Furthermore, as presented in equation (4), the combination ratio α at n+1 is the sum of a first term that is smaller than the combination ratio α at n and a second term that increases with the input signal level. Therefore, it becomes possible to model an amplifier with Idq drift.

[0111] [Second Embodiment]

[0112] Figure 10 The illustration shows a modeling device 100 according to a second embodiment. The configurations in the second embodiment, unless otherwise specified, are the same as those in the first embodiment.

[0113] There may be situations where using two models does not yield sufficient representational accuracy. Therefore, depending on the state in which the Idq drift is generated, using three or more computational models allows for a more appropriate representation of the characteristics of amplifier 10. Thus, in the second embodiment, as an example, a coupled model obtained by combining two adjacent computational models among the three computational models corresponding to the three internal states is used as amplifier model 200.

[0114] The modeling device 100 according to the second embodiment includes a selector 260 for selecting computational models to be combined. The modeling device 100 according to the second embodiment includes three computational models G1, G2, and G3. That is, the amplifier model 200 includes a first computational model G1, a second computational model G2, and a third computational model G3. The modeling device 100 may include four or more computational models. In the second embodiment, the characteristic G of the amplifier model 200 is represented as the composite characteristic of the computational models selected from the computational models G1, G2, and G3.

[0115] In the second embodiment, selector 260 selects two computational models G1, G2, and G3 to be combined. p and G p+1 The number of computational models to be selected is not limited to two, and three or more computational models can be selected from four or more computational models.

[0116] The selected computational model G p The calculations are performed by the first arithmetic unit 210. The first arithmetic unit 210 calculates the model G. p It is applied to the input signal u[n] and outputs the calculation result G. p (u[·]). The selected computational model G p+1 The calculations are performed by the second arithmetic unit 220. The second arithmetic unit 220 calculates the model G. p+1 It is applied to the input signal u[n] and the output calculation result G is given. p+1 (u[·]).

[0117] Combiner 230 obtains the selected computational model G by combining it with dynamically changing combination ratios. p and G p+1 The synthesized properties obtained. The selected computational model G. p and G p+1 The synthesis characteristic is characteristic G of amplifier model 200. The amplifier model 200 in the second embodiment is composed of... Figure 11 Equation (7) in the equation is expressed here. The combination ratio α in equation (7) will be described later. p [n]:(1-α p [n]).

[0118] In this embodiment, the combiner 230 combines the calculation result G of the first arithmetic unit 210 with a changed combination ratio. p (u[·]) and the calculation result G of the second arithmetic unit 220 p+1 (u[·]). The selected computational model G can be constructed in advance based on the combination ratio. p and G p+1 The coefficients of the function are used to obtain the composite function, and the obtained composite function can be applied to the input signal u[n].

[0119] Computational models G1, G2, and G3 model the amplifier 10 in different internal states. More specifically, the first computational model G1 models the characteristics of the amplifier 10 in a first internal state. The second computational model G2 models the characteristics of the amplifier 10 in a second internal state. The third computational model G3 models the characteristics of the amplifier 10 in a third internal state.

[0120] like Figure 12 As shown, the first internal state, the second internal state, and the third internal state are different from each other. The first internal state, corresponding to the first computational model G1, indicates the state that generates at least Idq drift in amplifier 10. The first internal state corresponds to the internal state parameter α[n] = 1. The third internal state, corresponding to the third computational model G3, indicates the state that generates at most Idq drift in amplifier 10. The third internal state corresponds to the internal state parameter α[n] = 0.

[0121] The second internal state, corresponding to the second computational model G2, is an intermediate internal state between the first and third internal states. Figure 12 In this context, the second internal state corresponds to the internal state parameter α[n] = 0.5. In the second internal state, the state in which the Idq drift is generated lies between the state in which the Idq drift is generated in the first internal state and the state in which the Idq drift is generated in the third internal state. The model can be expressed more precisely by increasing the number of internal states corresponding to the computational model to three or more.

[0122] Selector 260 selects the computational model G to be combined based on the internal state of amplifier 10. p and G p+1 More specifically, such as Figure 11 and Figure 12 As shown, in the case of 0.5 (=β2) < α[n] ≤ 1 (=β1) (in the case of p=1), the first computational model G1 and the second computational model G2 are selected as the computational models to be combined. In this case, the amplifier model 200 has the synthesis characteristics of the first computational model G1 and the second computational model G2, and the synthesis characteristics are determined by… Figure 11 Equation (7-1) in the text expresses this. Furthermore, the combination ratio used for the first computational model G1 is α1[n], while the combination ratio used for the second computational model G2 is (1-α1[n]). Here, according to... Figure 11 Equation (8-1) in the equation is used to determine α1[n].

[0123] The synthesized characteristics expressed by equation (7-1) can represent the characteristics of amplifier 10 in any first transition internal state (0.5 < α[n] < 1) between the first internal state and the second internal state. In the second embodiment, the characteristics of amplifier 10 in the first transition internal state can be expressed more appropriately compared to the first embodiment (0.5 < α[n] < 1).

[0124] In the case where 0 (=β3)≤α[n]≤0.5 (=β2) (in the case where p=2), the second computational model G2 and the third computational model G3 are selected as the computational models to be combined. In this case, the amplifier model 200 has the synthesis characteristics of the second computational model G2 and the third computational model G3, and the synthesis characteristics are determined by… Figure 11 Equation (7-2) in the equation expresses this. Furthermore, the combination ratio used for the second computational model G2 is α2[n], while the combination ratio used for the third computational model G3 is (1-α2[n]). Here, α2[n] is determined according to equation (8-2).

[0125] The synthesis characteristics expressed by equation (7-2) can represent the characteristics of amplifier 10 in any second transition internal state (0 < α[n] < 0.5) between the second internal state and the third internal state. In the second embodiment, the characteristics of amplifier 10 in the second transition internal state (0 < α[n] < 0.5) can be expressed more appropriately compared to the first embodiment.

[0126] Figure 13 The diagram illustrates the process of generating computational models G1, G2, and G3. First, the input signal is provided to amplifier 10, and the output signal is measured (step S21).

[0127] In step S22, the internal state parameter α[n] is calculated from the input signal level applied to amplifier 10. The internal state parameter α[n] is calculated by α[n] generator 300 based on the level of input signal u[n].

[0128] In step S23, the time T1 when the value of α[n] becomes 1 is determined. Time T1 is the time when the speculative amplifier 10 is in the first internal state with the minimum generation of Idq drift.

[0129] In step S24, the coefficients of the first computational model G1 are identified from the data of the input and output signals within a predetermined past time period starting from time T1. Through this process, the first computational model G1 is obtained.

[0130] In step S25, the time T2 when the value of α[n] becomes 0.5 is determined. Time T2 is the time when the speculative amplifier 10 is in the second internal state as an intermediate internal state.

[0131] In step S26, the coefficients of the second calculation model G2 are identified from the data of the input and output signals over a predetermined past time period starting from time T2. Through this process, the second calculation model G2 is obtained.

[0132] In step S27, the time T3 when the value of α[n] becomes 0 is determined. Time T3 is the time when the speculative amplifier 10 is in the third internal state that generates at most Idq drift.

[0133] In step S28, the coefficients of the third calculation model G3 are identified from the data of the input and output signals within a predetermined past time period starting from time T3. Through this process, the third calculation model G3 is obtained.

[0134] like Figure 10 As shown, the modeling device 100 of the second embodiment includes a generator 301, which generates a parameter α[n] for determining the combination ratio from a parameter α[n] indicating the internal state. p [n]. Generator 301 obtains α[n] from generator 300 and outputs α. p [n].

[0135] like Figure 14 As shown, generator 301 according to Figure 14 Equation (8) in the equation calculates the parameter α from the parameter α[n]. p [n]. Generator 301 will set the value in β. p ≥α[n]≥β p+1 Normalize α[n] within the range to 0≤α pValues ​​within the range [n]≤1. Equation (8) becomes as follows when 0.5 (=β2)<α[n]≤1 (=β1) (when p=1): Figure 11 Equation (8-1) in the equation becomes, in the case of 0 (=β3)≤α[n]≤0.5 (=β2) (in the case of p=2), that is... Figure 11 Equation (8-2) in the text.

[0136] Although embodiments of the invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the invention.

Claims

1. A modeling apparatus that uses an amplifier model to perform calculations, wherein the internal state of the amplifier's influence distortion characteristics is altered. in, The amplifier model includes: Multiple computational models model the amplifier in different internal states, and A combiner combines the plurality of computational models at a combination ratio corresponding to a changed internal state, wherein the combination ratio changes dynamically. Wherein, the combination ratio is based on the parameters of the internal state. Each of the plurality of computational models expresses a first memory effect with a first response time in the amplifier, and The parameter calculation model for calculating the parameters of the internal state is expressed as a second memory effect in the amplifier having a second response time that is longer than the first response time.

2. The modeling apparatus of claim 1 further includes a generator configured to generate the parameters indicating the internal state.

3. The modeling device according to claim 2, wherein: The combination ratio is determined based on the parameters indicating the internal state, and The parameters of the internal state are calculated based on the level of the signal to be amplified by the amplifier.

4. The modeling device according to claim 3, wherein, In addition to the level of the signal to be amplified by the amplifier, the parameters of the internal state are also calculated based on past values ​​of the parameters of the internal state.

5. The modeling apparatus according to any one of claims 1 to 4, wherein, The combination ratio is calculated based on the level of the signal to be amplified by the amplifier.

6. The modeling apparatus according to any one of claims 1 to 4, wherein, The combination ratio is calculated using temperature-dependent parameters that vary depending on temperature conditions.

7. The modeling apparatus according to any one of claims 1 to 4, wherein, The multiple computational models mentioned are two computational models.

8. The modeling apparatus according to any one of claims 1 to 4, further comprising a selector, wherein: The plurality of computational models includes three or more computational models, and The selector is configured to select two or more computational models from the three or more computational models to be combined by the combiner.

9. The modeling device according to claim 8, wherein, The selector is configured to select two or more computational models based on parameters indicating the internal state.

10. The modeling apparatus according to any one of claims 1 to 4, wherein: The plurality of computational models include: A first computational model models the amplifier in a first internal state, and A second computational model is used to model the amplifier in a second internal state that is different from the first internal state. The combination ratio has a value corresponding to the transitional internal state between the first internal state and the second internal state.

11. The modeling apparatus according to any one of claims 1 to 4, wherein: The plurality of computational models include: A first computational model is used to model the amplifier in a first internal state. A second computational model is used to model the amplifier in a second internal state that is different from the first internal state. A third computational model is used to model the amplifier in a third internal state that is different from the first and second internal states. The second internal state is an intermediate internal state between the first internal state and the third internal state.

12. The modeling apparatus according to any one of claims 1 to 4, wherein: As the level of the input signal to the amplifier increases, the internal state changes more significantly from its initial internal state, and as the level of the input signal decreases, the internal state returns to its initial internal state over time. The plurality of computational models include: A first computational model models the amplifier in a first internal state, and A second computational model is used to model the amplifier in a second internal state that is different from the first internal state. The change from the initial internal state when in the second internal state is greater than the change from the initial internal state when in the first internal state. The first combination ratio is the combination ratio of the first calculation model relative to the sum of the first calculation model and the second calculation model at the first time point. The second combination ratio is the ratio of the first calculation model to the sum of the first calculation model and the second calculation model at a second time after the first time. When the level of the input signal is zero at the first time point, the second combination ratio is less than the first combination ratio, and The second combination ratio when the level of the input signal at the first time is a first level is greater than the second combination ratio when the level of the input signal at the first time is a second level that is lower than the first level.

13. The modeling device according to claim 12, wherein, The second combination ratio is the sum of the first term and the second term, wherein the first term is less than the first combination ratio, and the second term increases as the level of the input signal increases at the first time.

14. A computational method for modeling an amplifier, wherein the internal state of the amplifier's influence distortion characteristics is changing, the computational method comprising the following steps: The combiner combines multiple computational models that model the amplifier in different internal states with a combination ratio corresponding to the changed internal states, wherein the combination ratio changes dynamically. Wherein, the combination ratio is based on the parameters of the internal state. Each of the plurality of computational models expresses a first memory effect with a first response time in the amplifier, and The parameter calculation model for calculating the parameters of the internal state is expressed as a second memory effect in the amplifier having a second response time that is longer than the first response time.

15. A non-transitory computer-readable storage medium storing a computer program for computation, the computation modeling an amplifier whose internal states affect distortion characteristics are altered, the computer program causing a computer to perform processing, the processing comprising: Multiple computational models modeling the amplifier in different internal states are combined using combination ratios corresponding to the changed internal states, wherein the combination ratios change dynamically. Wherein, the combination ratio is based on the parameters of the internal state. Each of the plurality of computational models expresses a first memory effect with a first response time in the amplifier, and The parameter calculation model for calculating the parameters of the internal state is expressed as a second memory effect in the amplifier having a second response time that is longer than the first response time.

Images