A method for generating radio frequency (RF) impairments model and a transmission system

Abstract

The invention is related to a computer-implemented method for generating Radio Frequency (RF) impairments model to be used to design a transmission system and the transmission system designed according to the model.

Description

[0001] A METHOD FOR GENERATING RADIO FREQUENCY (RF) IMPAIRMENTS MODEL AND A TRANSMISSION SYSTEM

[0002] Technical Field

[0003] The invention is related to a computer-implemented method for generating Radio Frequency (RF) impairments model to be used to design a transmission system and the transmission system designed according to the model.

[0004] Prior Art

[0005] The rapid growth of wireless devices and the increasing variety of wireless services, including virtual / augmented reality and the Internet of Things (loT), pose unprecedented challenges for next-generation wireless networks. These challenges stem from factors such as diverse data traffic, massive device connectivity, ultra-high bandwidth efficiency, and ultra-low latency requirements, alongside the demand for cost-effective devices like loT devices. To reduce costs, companies often opt for low-cost components in the Radio Frequency (RF) chains of these devices. However, the use of these inexpensive components in RF chains introduces severe impairments and signal degradation in addition to wireless medium effects, resulting in significant setbacks in communication and sensing performance. This poses a significant contradiction with the stringent requirements of emerging technologies like 6G.

[0006] In the pursuit of advancing wireless communication technologies, our research involves developing innovative solutions to address real-world challenges. While many problems are tackled through simulations that consider channel impairments and established models based on field measurements, some simulations demand a closer look at hardware impairments. In these cases, we utilize existing models, which may not always be entirely accurate or comprehensive in covering the collective impact of RF issues.

[0007] Traditionally, modeling RF impairments involved creating separate models for each type of impairment. Understanding how these impairments collectively affected system performance required the use of various tools and measurements, resulting in a cumbersome and less accurate approach. The lack of precision in representing the complex interactions among different impairments further hindered the accuracy of the combined RF impairment model. In the prior art, RF impairment modeling predominantly relied on traditional methods which are measurements-based or Al-based which also can be called black box models. Measurements-based models involve using measurements, such as modeling power amplifier non-linearities [1], Phase Noise and Time Jitter can also be modeled using mathematical approximations [8],

[0008] Existing models, derived from (field) experimental measurements, tended to oversimplify the interactions among different impairments or complicate this interaction by representing everything in a black box Al model. The lack of accuracy or the complexity in these models made it challenging to comprehensively understand the collective impact of RF impairments on system performance and makes it hard to generalize. In contrast to that, Al-driven models may lack transparency in understanding the underlying processes.

[0009] Using measurements such as modeling power amplifier non-linearities disclosed in study of Salah, et al. [1], Using Al is another aspect of looking at these impairments such as in studies of Fernandez et al. [2] and Sygletos et al. [3] which examine these impairments through a decoding aspect, emphasizing their influence on decoding processes. In contrast, Jaraut et al. [4] and Paul et al. [5] explore harnessing these impairments for distinct purposes, including predistortion, hardware issue identification, and authentication. Notably, within the context of studies of Fernandez et al. [2] and Sygletos et al. [3], the primary goal is to eliminate the impact of these impairments on the receiver, rendering estimation procedures unnecessary. Moreover, Mohammadian et al. [6] and Aygul et al [7] used deep learning to look at the combined effect of RF impairments, However, the previously mentioned methods don’t tend to give a mathematical model to the combined RF impairments.

[0010] Some attempts were made to incorporate hardware impairments into simulations, but these models were not entirely accurate and fell short in capturing the general influence of RF impairments.

[0011] As can be seen, the conventional methods struggled with accuracy and efficiency, especially when it came to understanding the combined effects of various RF impairments. The limitations of these methods highlighted the need for a more advanced and comprehensive solution As a result, all the problems mentioned above have made it necessary to provide a novelty in the related field.

[0012] Brief Description and Objects of the Invention

[0013] The main object of the present invention is to establish a method generating an analytical model with improved accuracy by not oversimplify the interactions among different impairments and transparency and is to design a transmission system according to such a model.

[0014] Another object of the present invention is enabling a more nuanced and comprehensive analysis of RF impairments to facilitate better-informed decision-making in system design and operation.

[0015] Another object of the present invention is to establish a method that provides comprehensive understanding of how RF impairments collectively impact system performance.

[0016] Another object of the present invention is directly deriving analytical equations from data, eliminating the need for separate models for each impairment.

[0017] Another object of the present invention is facilitating proper identification of signal anomalies and aids in recovering the unimpaired signal.

[0018] To achieve such an object, the method of invention employs Symbolic Regression (SR), a machine learning-based regression method, directly deriving analytical equations from real transceiver data. This enhances precision and provides a transparent understanding of combined impairments, overcoming the oversimplification seen in traditional models. Moreover, the inefficiency in representing combined effects is resolved through SR, acknowledging intricate interactions among impairments, and presenting a model that authentically mirrors the complexity of RF systems. Simultaneously, challenges in incorporating hardware impairments are addressed by streamlining the modeling process with SR, eliminating the need for separate models, and enhancing overall efficiency. This approach ensures a more realistic representation of RF system behavior, offering a comprehensive and accurate solution to the challenges posed by traditional RF impairment modeling methods.

[0019] More clearly, the predefined parameters of the transmitted signal and signal itself, the impaired signal and the impairment and the error, which are determined according to the transmitted and received data, to a symbolic regression block to obtain a Radio Frequency (RF) impairments model in the invention.

[0020] Description of the Figures of the Invention

[0021] The figures and related descriptions necessary for the subject matter of the invention to be understood better are given below.

[0022] Figure 1. A schematic view represents the fundamental components of a digital radio system.

[0023] Figure 2. A flowchart represents the method of present invention.

[0024] Figure 3. Comparison of the generated models of the invention and real world used model.

[0025] Reference Numbers

[0026] The parts and components given in the figures are referenced for the subject matter of the invention to be understood better.

[0027] 101. Processing unit

[0028] 102a. Digital to analog converter

[0029] 102b. Analog to digital converter

[0030] 103. Filter

[0031] 104. Modulator block

[0032] 105. Radio Front block

[0033] 106a. IF local oscillator

[0034] 106b. RF local oscillator

[0035] 107. Mixer

[0036] 108 I / Q modulator

[0037] 109. Combiner

[0038] 110. Power amplifier

[0039] 111. Transmitter antenna

[0040] 112. Medium

[0041] 113. Receiver antenna

[0042] 114. Low noise amplifier Txc. Transmitter chain

[0043] Rxc. Receiver chain

[0044] 201. Defining parameters

[0045] 202. Generating signal

[0046] 203. Passing the signal through transmitter chain

[0047] 204. Transmitting the signal through the channel

[0048] 205. Receiving signal at receiver chain

[0049] 206. Time synchronization

[0050] 207. Detecting the impaired signal

[0051] 208. Detecting impairment

[0052] 209. Detecting error

[0053] 210. Inputting symbolic regression block

[0054] 211. Generating model

[0055] Detailed Description of the Invention

[0056] The invention is related to a computer-implemented method for generating Radio Frequency (RF) impairments model to be used to design a transmission system and the transmission system designed according to the model.

[0057] Radio Frequency (RF) impairments occur in digital radio transmission systems. The digital radio transmission systems comprise at least one transmitter chain (Txc) having at least one transmitter antenna (111) and at least one receiver chain (Rxc) having at least one receiver antenna (113). The transmitter chain (Txc) comprises means for generating a digital signal and converting means for converting the digital signal to an analog signal. The receiver chain (Rxc) comprises converting means for converting the received analog signal to a digital signal. Furthermore, the digital radio transmission system also comprises a processing device to receive and feed parameters (will be explained later) to a symbolic regression block and executing the symbolic regression block for generating Radio Frequency (RF) impairments model. The transmitter antenna (111) and the receiver Antenna (113) serves as the interface between electrical signals and electromagnetic waves, antennas play a vital role in signal propagation. Various performance parameters, such as bandwidth, gain, and radiation pattern, define antenna characteristics. Directional antennas and beamforming become especially significant in higher frequencies where path loss is severe.

[0058] Figure 1 represent components of a digital radio transmission system. The component may be configured differently than shown in Fig. 1 or some component may be removed from system.

[0059] At transmitter chain (Txc) side, a digital processing unit (101) provided. The digital processing unit (101) has function of generating a digital signal which is transmitted to the receiver chain (Rxc) later. The digital processing unit (101) configured to generate signal according to predetermined parameters. The predetermined parameters can be selected between operating frequency, sampling rate, symbol duration, modulation order, filter type, filter parameters.

[0060] The digital processing unit (101) preferably generates In-Phase (I) and Quadrature-Phase (Q) channels, customized to meet waveform specifications and standard requirements.

[0061] The generated digital signal by the digital processing unit (101) is converted from digital to analog by a digital to analog convertor (102a) and filtered by a transverse filter (103) which resulting in filtered I / Q analog signals. The resolution of converter is defined by the number of bits, with higher bit resolutions providing finer details. The filter (103) filters are isolating desired signals while blocking undesired ones. While perfect filters (103) exhibit a flat response within the passband and infinite attenuation in the stopband, practical filters showcase a gradual roll-off, passband ripples, and stopband leakage.

[0062] The transmitter chain (Txc) has a modulator block (104) and the filtered I / Q analog signal signals undergo modulation in an I / Q modulator block (104). The modulator block (104) comprises IF (Intermediate Frequencies) local oscillator (106a) and a mixer (107).

[0063] IF local oscillator (106a) shapes the In-Phase and Quadrature-Phase carriers at Intermediate Frequencies.

[0064] At least one, preferably two mixers (107) are provided in the modulator block (104). The mixers (107) components facilitate frequency translation through signal multiplication, utilizing the IF local oscillator (106a). Subsequent filtering selects the desired output frequency, allowing for either up-conversion (higher frequency) or down-conversion (lower frequency).

[0065] A combiner (109) is provided to sum the resulting real (In-phase) and imaginary (Quadrature- Phase) sinusoidal signals from the modulator block (104).

[0066] The transmitter chain (Txc) further comprising another filter (103) positioned after the combiner (109) and the filtered signal is forwarded to radio front block (105), which contains RF local oscillator (106b), a mixer (107) and a power amplifier (110). IF local oscillator (106a) and RF local oscillator (106b) are instrumental in producing periodic sinusoidal waves, crucial for mixer (107) operations that enable frequency translation. The precision of the transceiver's frequency hinges on the accuracy and stability of this IF local oscillator (106a) and RF local oscillator (106b). Any imperfections in carrier synchronization during reception can lead to constant or gradually changing phase errors, often rectifiable through baseband processing. Short-term oscillation instabilities manifest as phase noise, exerting a substantial impact on the RF performance of both the transmitter chain (Txc) and receiver chain (Rxc).

[0067] The signal forwarded to RF local oscillator (106b) goes to another filter (103) and a power amplifier (110) which is connected to the transmitter antenna (111). The transmitter antenna (111) transmits the RF signal through a medium (112) to the receiver chain (Rxc).

[0068] The medium (112) can be a wireless channel or a cable for all the embodiments.

[0069] The RF signal is received by the receiver antenna (113) and the received signal is filtered and sent to Low-Noise Amplifier (114).

[0070] The power amplifier (110) and low-noise amplifier (114) contribute to signal amplification, low-noise amplifier (114) focuses on enhancing low-level signals with minimal noise, while power amplifier (110) amplifies signals for transmission. The nonlinear behavior of amplifiers can affect signal quality and result in legally restricted signal leakage into adjacent channels, as characterized by the adjacent channel power ratio (ACPR). ACPR is a measure of the power leakage from the desired channel into adjacent channels. It quantifies the level of interference and is crucial for evaluating the quality of transmitted signals. The Digital to Analog Converter (102a), Analog to Digital Converter (102b) bridge the analog and digital realms, with ADC converting analog signals to digital and DAC performing the reverse.

[0071] The signal is transferred to another filter through the mixer (107) and is sent to the modulator block (104) of the receiver chain (Rxc) and filtered by another filter and lasty, converted from digital to analog by an analog to digital convertor (102b) before reach a processing unit (101) of the receiver chain (Rxc).

[0072] The processing unit (101) serves as hardware platforms for implementing radio functionalities, DSPs and FPGAs employ digital signal processing techniques. DSPs are specialized processors optimized for DSP algorithms, while FPGAs offer configurable hardware with a range of logic resources.

[0073] In the method of the present invention, the digital signal used for method is generated by the processing unit (101) of the transmitter chain (Txc). The signal generating (202) is carried out according to the predetermined parameters such as an operating frequency, a sampling rate, a symbol duration, a modulation order, a filter type, filter parameters which all parameters are defined (201) before. The signal is passed through transmitter chain (203) and transmitting through the wireless channel (204). During the process, the digital signal is converted to analog signal before transmitting by the transmitter antenna (111). Preferably, the transmitted signal is a single carrier with QAM symbols and serves the identification of the impairments at the receiver chain (Rxc) side.

[0074] After receiving the transmitted analog signal by the receiver antenna (205), the received signal is time synchronized.

[0075] Preferably, a known sequence may be generated at transmitter chain (Txc) side, two. The known sequence is used for the time and frequency synchronization and used to detect RF impairments at the receiver chain (Rxc) side.

[0076] After the synchronization, detecting impaired signal (206) is carried out by comparing the received signal to the transmitted signal and impairments are detected thus we can say that the received signal is an impaired signal. The generated signal at the transmitter chain (Txc) and the impaired signal (206) is used to determine an impairment (207) and determine an error (208) by comparing the received signal to the transmitted signal, errors are detected.

[0077] For determination of impairment (207) and an error (208), comparation of generated signal at the transmitter chain (Txc) and the impaired signal (206) can be used. For comparison, many techniques can be used. For example correlation based, Maximum likelihood estimation, Phase locked loops PLL, machine learning techniques by comparing the known sequence with the received one can be used to estimate the frequency offset, IQ imbalance, phase noise and jitter, Non-linear distortions.

[0078] Finally, the predefined parameters that is used to generate signal, the transmitted signal, the impaired signal and the impairment and the error is inputted / fed to a Symbolic regression block (210) to generating Radio Frequency (RF) impairments model (211). The Symbolic regression block is trained to create a model for both individual and combined RF impairments effects, such as Phase Noise (PN), in-phase (I) and quadrature (Q) signal components, Gain Imbalance, Quadrature Offset (QO), Frequency Offset (FO), and Power Amplifier non-linearities. Each impairment is meticulously modeled to provide a comprehensive understanding of its specific behavior.

[0079] Crucially, the combined modeling of these impairments goes beyond a simple summation of individual models. The approach recognizes and incorporates the complicated interactions between different impairments, resulting in a combined model that better reflects the real complexity of RF systems.

[0080] This invention not only gives accurate models for RF impairments but also simplifies the entire process, offering a more comprehensive understanding of how these impairments collectively impact system performance.

[0081] One of the most critical components in an RF chain is the power amplifier (PA). To increase the propagation distance of a signal, PA is used to amplify it. Thus, the desired spectral efficiency can be achieved. However, increasing the PA input power can lead to exceeding the linear range of input and output power which can cause non-linearities and clipping in the signal which can be interpreted as signal distortion and may lead to severe degradation in a communication system performance. To combat these non-linearities, first understanding the behavior and representing it in a mathematical equation is important.

[0082] One of the early and famous attempts to model PA non-linearities is Saleh’s model (ref) which was studied through experimental results on real amplifiers. A model is chosen from his work [ref] is the following: P(r)

[0083] Where r(t) represents the envelope of the input signal Sin(t) = r(t)cos (2iifct + <p(t)) and fcis the carrier frequency and <p(t) is the phase of the input signal. The output of PA, sout(t), can be written as:

[0084] As an example, the model to generate the data is used in a python script and giving the data to a symbolic regressor of the following parameters: “population_size”=5000, “generations”=20, “stopping_criteria”=0.000001. where “population size” is the number of pseudo-random equations that a symbolic regressor generates at the start, “generations” is the number of generations to make before stopping in order to approach the an accurate yet simple model and “stopping criteria” is the threshold value of mean square error to stop at even before the number of generations is not finished. The given values can’t be interpreted as limitation for the invention. It is obvious to use different values for the skilled person in the art to adapt model for different situations.

[0085] After training the SR model on the generated data of a chosen model:

[0086] P(r) = 2.1578 r / (l + 1.1517 r2)

[0087] We got 2 models by choosing different number of generations:

[0088] Pi(r) = 1.85873605947955 r / (0.848 + 1 r2)

[0089] P2(r) = 1.9996525949993265 r / (0.9263226145315 + 1.066845726468 r2)

[0090] Figure 3 shows that comparison of the generated models of the invention and real world used model. As can be seen in graphic, the generated models of the invention are substantially reduced non-linearities.

[0091] Another aspect of invention is to design a transmission system, according to the model obtained by above mentioned methods. The means or parameters of the transmission system (for both or one of the transmitter chain (Txc) and the receiver chain (Rxc)) may be adjusted, configured or removed according to the model to improve performance against the impairments.

[0092] Further aspect of invention is to a system for generating such a model. The system comprises the transmitter chain (Txc) and the receiver chain (Rxc). The transmitter chain (Txc) comprises a processing unit (101) to generate signal, a digital analog converter (102a), a filter (103) connected to the converter and a modulator block (104) according above embodiments and another filter (103) connected to radio front block (105).

[0093] The transmitter comprises at least one transmitter antenna (111) or wired connection to send data to the receiver chain (Rxc) comprises at least one receiver antenna (113) or wired connection to receive said data and radio front block (105) and a modulator block (104) and an analog to digital converter (102b) and filters (103) connected between them.

[0094] REFERENCES

[0095] [1] Saleh, A. A., 1981. Frequency-independent and frequency-dependent nonlinear models of TWT amplifiers. IEEE Transactions on communications, 29(11), pp.1715-1720.

[0096] [2] Fernandez, E.A., Soto, A.M.C., Gonzalez, N.G., Serafino, G., Ghelfi, P. and Bogoni, A., 2019. Machine learning techniques to mitigate nonlinear phase noise in moderate baud rate optical communication systems. In Intelligent Syst. Comput. IntechOpen.

[0097] [3] Sygletos, S., Redyuk, A. and Sidelnikov, O., 2019, July. Nonlinearity compensation techniques using machine learning. In Signal Processing in Photonic Communications (pp. SpT2E-2). Optica Publishing Group.

[0098] [4] Jaraut, P., Rawat, M. and Ghannouchi, F.M., 2018. Composite neural network digital predistortion model for joint mitigation of crosstalk, $ i / q $ imbalance, nonlinearity in MIMO transmitters. IEEE Transactions on Microwave Theory and Techniques, 66(11), pp.5011- 5020.

[0099] [5] Paul, L.Y., Baras, J.S. and Sadler, B.M., 2008. Physical-layer authentication. IEEE Transactions on Information Forensics and Security, 3(1), pp.38-51. [6] Mohammadian, A., Tellambura, C. and Li, G.Y., 2021. Deep learning LMMSE joint channel, PN, and IQ imbalance estimator for multicarrier MIMO full-duplex systems. IEEE Wireless Communications Letters, 11(1), pp.111-115.

[0100] [7] Aygul, M.A., Memi§oglu, E. and Arslan, H., 2022, April. Joint estimation of multiple RF impairments using deep multi-task learning. In 2022 IEEE Wireless Communications and

[0101] Networking Conference (WCNC) (pp. 2393-2398). IEEE.

[0102] [8] Lee, T.H. and Hajimiri, A., 2000. Oscillator phase noise: A tutorial. IEEE journal of solid- state circuits, 35(3), pp.326-336.

Claims

CLAIMS1. A computer-implemented method for generating Radio Frequency (RF) impairments model for a transmission system having a transmitter chain (Txc) and a receiver chain (Rxc) communicates with each other, characterized by receiving the signal generated by predefined parameters at the transmitter chain (Txc) by the receiver chain (Rxc), synchronizing the received signal with the transmitted signal, detecting an impaired signal and detecting an impairment and an error, feeding the predefined parameters, the transmitted signal, the impaired signal and the impairment and the error to a symbolic regression block and obtaining a Radio Frequency (RF) impairments model by the Symbolic regression block.

2. A method according to Claim 1, characterized by further comprising step of receiving a known sequence at the transmitter chain (Txc) by the receiver chain (Rxc) synchronizing the received signal with the transmitted signal.

3. A method according to Claim 1, wherein the predefined parameters are at least one of operating frequency, sampling rate, symbol duration, modulation order, filter type, filter parameters. Or any parameter that can be used to distinguish impairments in a signal.

4. A method according to Claim 1, wherein the Symbolic regression block is trained to be generate model for improve performance against the both induvial and combined impairments.

5. A method according to Claim 1, wherein the impairment is at least one of Phase Noise (PN), in-phase (I) and quadrature (Q) signal components, Gain Imbalance, Quadrature Offset (QO), Frequency Offset (FO), and Power Amplifier non-linearities6. A data processing device comprising means for carrying out the steps of the method of any of claim 1 to 5.

7. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry carrying out the steps of the method of any of claim 1 to 3.

8. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method of any of claim 1 to 5.

9. A transmission system designed according to a Radio Frequency (RF) impairments model generated according to any of claim 1 to 5.

Images