# Practical Demonstration of 5G NR Transport Over-Fiber System with Convolutional Neural Network

## Abstract

**:**

## 1. Introduction

- In the experimental testbed, multiband 5G NR signals are used to cover enhanced mobile wide band (eMBB) scenarios and small cells for 3 GHz and 10 GHz, respectively.
- An abled DPD method based on a Convolutional Neural Network (CNN) is proposed and demonstrated. In comparison to existing learning architectures, the proposed DPD identification approach has a better performance and lower complexity than our previous machine learning approach.
- Finally, a simple CNN-based DPD algorithm is proposed as an upgrade to our previously published DPD-based technique based on deep learning for 20 MHz with 5G New Radio (NR) based RoF links. A new sort of training is used to implement the CNN DPD technique, which does not use In-direct Learning Architecture (ILA). We first use an RoF CNN to simulate the generic RoF connection and then use this to train the proposed DPD CNN by backpropagating the mistakes.
- A comparative experimental investigation was conducted in which the previously proposed ILA-based GMP method was benchmarked against the CNN technique and compared utilising a 5G NR multiband signal. Error Vector Magnitude (EVM) and multiplications and coefficients required measuring complexities are used to evaluate the performance.

## 2. Literature Review

## 3. Convolutional Neural Network Architecture

#### 3.1. CNN Model Salient Features

#### 3.1.1. Optimizer

#### 3.1.2. Activation Functions

#### 3.1.3. Regularization

_{1}represents the primary hidden output layer, $f$ is the nonlinear function of activation and ${W}_{1}$ is the weight and ${b}_{1}$ is the bias for the first output layer in the network.

^{th}layer is represented as:

#### 3.1.4. Training Algorithm

Algorithm 1. Training for linearization (DPD) |

$\widehat{x}\left(n\right)$ ←
$x\left(n\right)$for i $\le $ Z do$y\left(n\right)$ ← $I\left(\widehat{x}\left(n\right)\right)$: //Radio over Fiber -Transmission $\widehat{I}$ ← Train on $\widehat{x}\left(n\right)$, $\frac{y\left(n\right)}{G}$ //Update Radio over Fiber CNN //Fixed NN weights of $\widehat{I}$ $\widehat{I}$ ← Train on $x\left(n\right)$. //Utilize ${\widehat{I}}^{-1}$($\widehat{I}\left(x\left(n\right)\right)$) $\widehat{x}\left(n\right)$ ← ${\widehat{I}}^{-1}\left(x\left(n\right)\right)$: //Predistort end for |

## 4. Experimental Setup

## 5. Results and Discussion

#### Error Vector Magnitude

## 6. Complexity Considerations

## 7. Conclusions

## Funding

## Institutional Review Board Statement

## Data Availability Statement

## Conflicts of Interest

## References

- Hadi, M.; Awais, M.; Raza, M.; Khurshid, K.; Jung, H. Neural Network DPD for Aggrandizing SM-VCSEL-SSMF-Based Radio over Fiber Link Performance. Photonics
**2021**, 8, 19. [Google Scholar] [CrossRef] - Hadi, M.U.; Awais, M.; Raza, M.; Ashraf, M.I.; Song, J. Experimental Demonstration and Performance Enhancement of 5G NR Multiband Radio over Fiber System Using Optimized Digital Predistortion. Appl. Sci.
**2021**, 11, 11624. [Google Scholar] [CrossRef] - Nanni, J.; Polleux, J.-L.; Algani, C.; Rusticelli, S.; Perini, F.; Tartarini, G. VCSEL-based radio-over-G652 fiber system for short-/medium-range MFH solutions. J. Lightwave Technol.
**2018**, 36, 4430–4437. [Google Scholar] [CrossRef] - Khurshid, K.; Khan, A.A.; Siddiqui, H.; Rashid, I.; Hadi, M.U. Big Data Assisted CRAN Enabled 5G SON Architecture. J. ICT Res. Appl.
**2019**, 13, 93–106. [Google Scholar] [CrossRef] - Nanni, J.; Baschieri, L.; Polleux, J.-L.; Tartarini, G.; Hadi, M.U. Effective Digital Pre-Distortion Loop for Front Hauls based on short-λ-VCSELs over pre-existent G-652 Infrastructures. In Proceedings of the 2021 International Topical Meeting on Microwave Photonics (MWP), Pisa, Italy, 15–17 November 2021; pp. 1–4. [Google Scholar] [CrossRef]
- Aslam, N.; Xia, K.; Haider, M.T.; Hadi, M.U. Energy-Aware Adaptive Weighted Grid Clustering Algorithm for Renewable Wireless Sensor Networks. Future Internet
**2017**, 9, 54. [Google Scholar] [CrossRef] [Green Version] - Zhang, X.; Zhu, R.; Shen, D.; Liu, T. Linearization Technologies for Broadband Radio-Over-Fiber Transmission Systems. Photonics
**2014**, 1, 455–472. [Google Scholar] [CrossRef] - Li, H.; Bauwelinck, J.; Demeester, P.; Torfs, G.; Verplaetse, M.; Verbist, J.; Van Kerrebrouck, J.; Breyne, L.; Wu, C.-Y.; Bogaert, L.; et al. Real-Time 100-GS/s Sigma-Delta Modulator for All-Digital Radio-Over-Fiber Transmission. J. Lightwave Technol.
**2020**, 38, 386–393. [Google Scholar] [CrossRef] [Green Version] - Hadi, M.U.; Jung, H.; Ghaffar, S.; Traverso, P.A.; Tartarini, G. Optimized digital radio over fiber system for medium range communication. Opt. Commun.
**2019**, 443, 177–185. [Google Scholar] [CrossRef] - Wang, J.; Jia, Z.; Campos, L.A.; Knittle, C.; Jia, S. Delta-Sigma Modulation for Next Generation Fronthaul Interface. J. Lightwave Technol.
**2018**, 37, 2838–2850. [Google Scholar] [CrossRef] - Nanni, J.; Giovannini, A.; Hadi, M.U.; Lenzi, E.; Rusticelli, S.; Wayth, R.; Perini, F.; Monari, J.; Tartarini, G. Controlling Rayleigh-Backscattering-Induced Distortion in Radio Over Fiber Systems for Radioastronomic Applications. J. Lightwave Technol.
**2020**, 38, 5393–5405. [Google Scholar] [CrossRef] - Van Kerrebrouck, J.; Breyne, L.; Li, H.; Bauwelinck, J.; Torfs, G.; Demeester, P.; Bohn, T. Real-Time All-Digital Radioover-Fiber LTE Transmission. In Proceedings of the 2017 Advances in Wireless and Optical Communications (RTUWO), Riga, Latvia, 2–3 November 2017; pp. 83–86. [Google Scholar]
- Li, H.; Hajipour, J.; Attar, A.; Leung, V.C.M. Efficient HetNet implementation using broadband wireless access with fiber-connected massively distributed antennas architecture. IEEE Wirel. Commun.
**2011**, 18, 72–78. [Google Scholar] [CrossRef] - Wang, D.; Zhang, M.; Fu, M.; Cai, Z.; Li, Z.; Han, H.; Cui, Y.; Luo, B. Nonlinearity mitigation using a machine learning detector based on k-nearest neighbors. IEEE Photonics Technol. Lett.
**2016**, 28, 2102–2105. [Google Scholar] [CrossRef] - Huang, Y.; Chen, Y.; Yu, J. Nonlinearity mitigation of RoF signal using machine learning based classifier. In Proceedings of the Asia Communications and Photonics Conference, Guangzhou, China, 10–13 November 2017; p. Su2A.28. [Google Scholar]
- Cui, Y.; Zhang, M.; Wang, D.; Liu, S.; Li, Z.; Chang, G.K. Bit-based support vector machine nonlinear detector for millimeter-wave radio-over-fiber mobile fronthaul systems. Opt. Express
**2017**, 25, 26186–26197. [Google Scholar] [CrossRef] [PubMed] - Li, D.; Yu, S.; Jiang, T.; Han, Y.; Gu, W. An M-ary SVM-based detection for 16-QAM RoF system with data-dependent cross modulation distortion. In Proceedings of the Asia Communications and Photonics Conference, Hong Kong, 19–23 November 2015; p. ASu3J.4. [Google Scholar]
- Gonzalez, N.G.; Zibar, D.; Caballero, A.; Monroy, I.T. Experimental 2.5-Gb/s QPSK WDM Phase-Modulated Radio-Over-Fiber Link with Digital Demodulation by a KK-Means Algorithm. IEEE Photonics Technol. Lett.
**2010**, 22, 335–337. [Google Scholar] [CrossRef] - Fernandez, E.A.; Torres, J.J.G.; Soto, A.M.C.; Gonzalez, N.G. Radio-over-fiber signal demodulation in the presence of non-Gaussian distortions based on subregion constellation processing. Opt. Fiber Technol.
**2019**, 53, 102062. [Google Scholar] [CrossRef] - Khurshid, K.; Khan, A.A.; Siddiqui, M.H.; Hadi, M.U.; Rashid, I.; Imran, M. Optimality of Linear MIMO Detection for 5G Systems via 1-Opt Local Search. J. Electr. Eng. Technol.
**2021**, 16, 1099–1108. [Google Scholar] [CrossRef] - Hadi, M.U.; Mittal, I. On the Use of SVR based Machine Learning Method for Nonlinearities Mitigation in Short Range Fronthaul Links. In Proceedings of the 2021 10th IEEE International Conference on Communication Systems and Network Technologies (CSNT), Bhopal, India, 18–19 June 2021; pp. 628–631. [Google Scholar]
- Hadi, M.U.; Basit, A. Machine Learning for Performance Enhancement in Fronthaul Links for IOT Applications. In Proceedings of the 2021 International Conference on Digital Futures and Transformative Technologies (ICoDT2), Islamabad, Pakistan, 20–21 May 2021; pp. 1–5. [Google Scholar]
- Lee, J.; He, J.; Wang, Y.; Fang, C.; Wang, K. Experimental demonstration of millimeter-wave radio-over-fiber system with convolutional neural network and binary convolutional neural network (BCNN). arXiv
**2020**, arXiv:2001.02018. Available online: https://arxiv.org/abs/2001.02018 (accessed on 5 December 2021). - Lee, J.; He, J.; Wang, K. Neural networks and FPGA hardware accelerators for millimeter-wave radio-over-fiber systems. In Proceedings of the International Conference on Transparent Optical Networks, Bari, Italy, 19–23 July 2020; p. Mo.D1.5. [Google Scholar]
- Liu, S.; Wang, X.; Zhang, W.; Shen, G.; Tian, H. An Adaptive Activated ANN Equalizer Applied in Millimeter-Wave RoF Transmission System. IEEE Photonics Technol. Lett.
**2017**, 29, 1935–1938. [Google Scholar] [CrossRef] - Liu, S.; Xu, M.; Wang, J.; Lu, F.; Zhang, W.; Tian, H.; Chang, G.-K. A Multilevel Artificial Neural Network Nonlinear Equalizer for Millimeter-Wave Mobile Fronthaul Systems. J. Lightwave Technol.
**2017**, 35, 4406–4417. [Google Scholar] [CrossRef] - Liu, S.; Alfadhli, Y.M.; Shen, S.; Tian, H.; Chang, G.K. Mitigation of multi-user access impairments in 5G A-RoF-based mobile-fronthaul utilizing machine learning for an artificial neural network nonlinear equalizer. In Proceedings of the Optical Fiber Communication Conference, San Diego, CA, USA, 11–15 March 2018. [Google Scholar]
- Liu, E.; Yu, Z.; Yin, C.; Xu, K. Nonlinear distortions compensation based on artificial neural networks in wideband and multi-carrier systems. IEEE J. Quantum Electron.
**2019**, 55, 800305. [Google Scholar] [CrossRef] - Liu, J.; Zou, X.; Bai, W. Performance enhancement of UFMC based radio over fiber system using ANN equalizer. In Proceedings of the Asia Communications and Photonics Conference, Hangzhou, China, 26–29 October 2018. [Google Scholar]
- Safari, L.; Baghersalimi, G.; Karami, A.; Kiani, A. On the Equalization of an OFDM-Based Radio-over-Fiber System Using Neural Networks. Radioengineering
**2017**, 26, 162–169. [Google Scholar] [CrossRef] - Zhou, Q.; Lu, F.; Xu, M.; Peng, P.-C.; Liu, S.; Shen, S.; Zhang, R.; Yao, S.; Finkelstein, J.; Chang, G.-K. Enhanced Multi-Level Signal Recovery in Mobile Fronthaul Network Using DNN Decoder. IEEE Photonics Technol. Lett.
**2018**, 30, 1511–1514. [Google Scholar] [CrossRef] - Xu, T.; Shevchenko, A.N.; Zhang, Y.; Jin, C.; Zhao, J.; Liu, T. Information rates in Kerr non-linearity limited optical fiber communication systems. Opt. Express
**2021**, 29, 17428–17439. [Google Scholar] [CrossRef] [PubMed] - Draa, M.N.; Hastings, A.S.; Williams, K.J. Comparison of photodiode nonlinearity measurement systems. Opt. Express
**2011**, 19, 12635–12645. [Google Scholar] [CrossRef] [PubMed] - Chen, Z.; Yan, L.; Pan, W.; Luo, B.; Zou, X.; Guo, Y.; Jiang, H.; Zhou, T. SFDR enhancement in analog photonic links by simultaneous compensation for dispersion and nonlinearity. OSA Optics Expr.
**2013**, 21, 20999–21009. [Google Scholar] [CrossRef] - Haas, B.M.; Murphy, T. A Simple, Linearized, Phase-Modulated Analog Optical Transmission System. IEEE Photonics Technol. Lett.
**2007**, 19, 729–731. [Google Scholar] [CrossRef] - Zhu, R.; Zhang, X. Linearization of Radio-Over-Fiber Systems by Using Two Lasers with Different Wavelengths. In Proceedings of the 2014 IEEE MTT-S International Microwave Symposium (IMS), Tampa, FL, USA, 1–6 June 2014. [Google Scholar]
- Ghannouchi, F.; Younes, M.; Rawat, M. Distortion and impairments mitigation and compensation of single- and multi-band wireless transmitters. IET Microw. Antennas Propag.
**2013**, 7, 518–534. [Google Scholar] [CrossRef] - Duan, R.; Xu, K.; Dai, J.; Cui, Y.; Wu, J.; Li, Y.; Dai, Y.; Li, J. Linearity improvement based on digital signal processing in intensity-modulated analog optical links incorporating photonic frequency down conversion. In Proceedings of the Optical Fiber Communication Conference and Exposition, Los Angeles, CA, USA, 6–8 March 2012. [Google Scholar]
- Pei, Y.; Xu, K.; Li, J.; Zhang, A.; Dai, Y.; Ji, Y.; Lin, J. Complexity-reduced digital predistortion for subcarrier multiplexed radio over fiber systems transmitting sparse multi-band RF signals. Opt. Express
**2013**, 21, 3708–3714. [Google Scholar] [CrossRef] - Vieira, L.; Gomes, N.J.; Nkansah, A.; van Dijk, F. Behavioral modeling of radio-overfiber links using memory polynomials. In Proceedings of the 2010 IEEE Topical Meeting on Microwave Photonics (MWP), Montreal, QC, Canada, 5–9 October 2010; pp. 85–88. [Google Scholar]
- Hekkala, A.; Hiivala, M.; Lasanen, M.; Perttu, J.; Vieira, L.C.; Gomes, N.J.; Nkansah, A. Predistortion of Radio Over Fiber Links: Algorithms, Implementation, and Measurements. IEEE Trans. Circuits Syst. I Regul. Pap.
**2011**, 59, 664–672. [Google Scholar] [CrossRef] - Fuochi, F.; Hadi, M.U.; Nanni, J.; Traverso, P.A.; Tartarini, G. Digital predistortion technique for the compensation of nonlinear effects in radio over fiber links. In Proceedings of the 2016 IEEE 2nd International Forum on Research and Technologies for Society and Industry Leveraging a Better Tomorrow (RTSI), Bologna, Italy, 7–9 September 2016; pp. 1–6. [Google Scholar]
- Vieira, L.C.; Gomes, N.J.; Nkansah, A. An experimental study on digital predistortion for radio-over-fiber links. In Proceedings of the Asia Communications and Photonics Conference and Exhibition, Shanghai, China, 8–12 December 2010; pp. 126–127. [Google Scholar] [CrossRef]
- Hekkala, A.; Lasanen, M.; Vieira, L.C.; Gomes, N.J.; Nkansah, A. Architectures for Joint Compensation of RoF and PA with Nonideal Feedback. In Proceedings of the 2010 IEEE 71st Vehicular Technology Conference, Taipei, Taiwan, 16–19 May 2010; pp. 1–5. [Google Scholar]
- Mateo, C.; Carro, P.L.; Garcia-Ducar, P.; De Mingo, J.; Salinas, I. Experimental evaluation of the feedback loop effects in digital predistortion of a radio-over-fiber system. In Proceedings of the 2017 Ninth International Conference on Ubiquitous and Future Networks (ICUFN), Milan, Italy, 4–7 July 2017; pp. 1039–1041. [Google Scholar]
- Mateo, C.; Clemente, J.; Garcia-Ducar, P.; Carro, P.L.; de Mingo, J.; Salinas, I. Digital predistortion of a full-duplex Radio-over-Fiber mobile fronthaul link with feedback loop. In Proceedings of the 2017 13th International Wireless Communications and Mobile Computing Conference (IWCMC), Valencia, Spain, 26–30 June 2017; pp. 1425–1430. [Google Scholar]
- Mateo, C.; Carro, P.L.; García-Dúcar, P.; De Mingo, J.; Salinas, Í. Minimization of Feedback Loop Distortions in Digital Predistortion of a Radio-Over-Fiber System with Optimization Algorithms. IEEE Photonics J.
**2017**, 9, 7904414. [Google Scholar] [CrossRef] - Roselli, L.; Borgioni, V.; Zepparelli, F.; Ambrosi, F.; Comez, M.; Faccin, P.; Casini, A. Analog laser predistortion for multiservice radio-over-fiber systems. J. Lightwave Technol.
**2003**, 21, 1211–1223. [Google Scholar] [CrossRef] [Green Version] - Hadi, M.U.; Jung, H.; Traverso, P.A.; Tartarini, G. Experimental evaluation of real-time sigma-delta radio over fiber system for fronthaul applications. Int. J. Microw. Wirel. Technol.
**2020**, 13, 756–765. [Google Scholar] [CrossRef] - Hraimel, B.; Zhang, X. Low-cost broadband predistortion-linearized single drive x-cut Mach-Zehnder modulator for radio-over-fiber systems. Photonics Technol. Lett.
**2012**, 24, 1571–1573. [Google Scholar] [CrossRef] - Habib, U.; Aighobahi, A.E.; Quinlan, T.; Walker, S.D.; Gomes, N.J. Analog Radio-Over-Fiber Supported Increased RAU Spacing for 60 GHz Distributed MIMO Employing Spatial Diversity and Multiplexing. J. Lightwave Technol.
**2018**, 36, 4354–4360. [Google Scholar] [CrossRef] - Jiang, D.; Liu, G. An Overview of 5G Requirements. In 5G Mobile Communications; Xiang, W., Zheng, K., Shen, X., Eds.; Springer: Cham, Switzerland, 2017; pp. 3–26. [Google Scholar]
- He, J.; Lee, J.; Kandeepan, S.; Wang, K. Machine Learning Techniques in Radio-over-Fiber Systems and Networks. Photonics
**2020**, 7, 105. [Google Scholar] [CrossRef] - Yang, Y.-S.; Huang, W.-C.; Li, C.-P.; Li, H.-J.; Stüber, G.L. A Low-Complexity Transceiver Structure with Multiple CFOs Compensation for OFDM-Based Coordinated MultiPoint Systems. IEEE Trans. Commun.
**2015**, 63, 2658–2670. [Google Scholar] [CrossRef] - Aslam, N.; Xia, K.; Hadi, M.U. Optimal Wireless Charging Inclusive of Intellectual Routing Based on SARSA Learning in Renewable Wireless Sensor Networks. IEEE Sens. J.
**2019**, 19, 8340–8351. [Google Scholar] [CrossRef] - LTE. Evolved Universal Terrestrial Radio Access (E-UTRA); ETSI, TS 136 104 V8.2.0; Base Station (BS) Radio Transmission and Reception; European Telecommunications Standards Institute: Valbonne, France, 2018. [Google Scholar]

**Figure 1.**Schema of RAN 5G covering the back and fronthaul with application scenarios such as houses, sports fields and transportation, etc.

**Figure 2.**RoF system block diagram with Radio over Fiber and CNN based DPD system. Transmission of I/P and O/P across the general RoF link yields the RoF NN model $\widehat{I}$. After that, we backpropagate error through I to train ${I}^{-1}$. The DPD-RoF model is then linearized by linking it to an RoF link. DPD is done in the digital baseband, which eliminates the need for DACs and ADCs The link that how CNN is made and trained is shown in Algorithm 1.

**Figure 3.**N number of hidden layers has K neurons per hidden layer, as shown in the schematic picture of the feedforward fully connected NN structure utilised.

**Figure 4.**Experimental block diagram for analog multiband 5G NR system. The functions: A: Choose between VSA 1 and VSA 2. B: Choose between a performance post-processing block and a synchronisation block. C: Synch is connected. The synchronisation block is followed by training. D: DPD disables the training DPD disables the training DPD disables the training DPD disables the training DPD disable E: Time synchronisation (TS) technique is required. F: Required for validating DPD inputs before sending them to the VSGs. For the DPD training CNN has been chosen.

**Figure 5.**EVM performance comparison (

**a**) for CNN and No DPD case for flexible 5G NR waveforms. (

**b**) 5G NR performance DPD efficacy in terms of EVM for proposed CNN DPD method vs GMP method and without DPD for varying RF input power.

**Table 1.**Nonlinearity mitigation methods employed (Dig = Digital, Elec = Electrical, Opt = Optical, ML = Machine Learning, ACLR = Adjacent Channel Leakage Ratio, EVM = Error Vector Magnitude are used to abbreviate in the table).

No. | Author | Type | Subcategory | Parameter | Linearization | Advantages | Disadvantages |
---|---|---|---|---|---|---|---|

1 | Wang et al. [14] | KNN | ML | BER | Fiber nonlinearity | 0.6 dB improvement | Large training data required |

2 | Cui et al. [15] | SVM | Deep Learning | BER | Modulation Impairments | 1.3 dB improvement | High Complexity |

3 | Li et al. [16] | SVM | ML | EVM | Fiber nonlinearity | 1.5 dB improvement | High data training and complexity |

4 | Gonzalez et al. [17] | ML | AI | BER, OSNR | Cross Modulation Detection | N/A | N/A |

5 | Fernandez et al. [18] | ML | ML | OSNR | Phase Modulation Impairments | 1.4 dB improvement | N/A |

6 | Hadi et al., Liu et al. [19,20,21,22,23,24] | ML | ML | EVM, ACLR | Laser Chirp | 10 dBs improvement | Limited to small link lengths |

7 | Liu et al., Hadi et al. [25,26,27,28] | Dig | ML-NN based | EVM | Laser | Learns nonlinearities | Limited to LTE framework. |

8 | Safari et al., Xu et al., Hadi et al. [29,30,31,32] | Dig | DNN | ACLR, EVM | Black Box approach | N/A | Limited to 20 MHz bandwidth and 256 QAM modulation |

9 | Draa et al.; Chen et al. [33,34] | Elec | Analog Predistortion | IMD3 | Complete RoF system (Laser, photodiode, LNA) | IMD3 for phase maintenance | Suppression of second order nonlinear distortion for high bandwidth is perplexing. |

10 | Hass et al. [35] | Opt | Mixed Polarization | Second/third order nonlinear distortion | Complete RoF system | Suppression of second and third order nonlinearities | To some extent, linear components are compressed. |

11 | Zhu et al. [36] | Opt | Dual wavelength linearization (DWL) | Second/third order nonlinear distortion | Complete RoF system | Suppression of second and third order nonlinearities | Transmission is wavelength dependent, i.e., nonlinear components are suppressed exclusively at anti-phased wavelengths. |

12 | Ghannouchi et al. [37] | Dig | DPD | Third order nonlinearities | Power Amplifier | Wideband improvement possible | The DSP necessary is difficult. |

13 | Duan et al. [38] | Dig | DPD | ACLR, EVM | Laser | Added accuracy with less DSP requirements | The amount of energy consumed is enormous. |

14 | Pei et al. [39] | Dig | DPD | ACLR | Modulator | Higher suppression in ACLR by 15 dB | Feedback complexity. |

15 | Lam et al. [40] | Dig | Digital Post Distortion | ACLR, BER | RoF | All order nonlinear distortion components significantly compressed. | Digitizer with high speed is required.Uplinks are the only ones that apply. |

16 | Hekkala et al. [41] | Dig | DPD | ACLR, EVM | Laser only | Less complexity and over head | Deployment of DPP at the RRH level, so that the prototype price is passed on to the client, adding to the RRH’s complexity. |

17 | Hadi et al. [42] | Dig | DPD | C/HD2, IIP2, IIP3 | Combination of fiber dispersion and laser chirp | Linearizes links up to tens of km | Limited to sinusoidal (single/dual) I/P tones. |

18 | Vieira et al. [43] | Dig | DPD | EVM | Laser | OFDM signal utilzation | RoF link is not generic, contains 10 dB attenuator. |

19 | Hekkala et al. [44] | Dig | DPD | ACLR, BER | Laser | OFDM signal with 12.5 MHz bandwidth | The magnitude (AM/AM) linearization is the only one shown. |

20 | Carlos et al. [45] | Dig | DPD | EVM, ACLR | RoF | LTE 20 MHz signal | Unrealistic feedback. |

21 | Carlos et al. [46] | Dig | DPD | NMSE, ACLR | RoF | LTE 20 MHz with 16 QAM modulation | To test the predistorter’s efficacy, the distributed feedback (DFB) laser was not pushed to greater RF I/P powers. |

22 | Carlos et al. [47] | Dig | DPD | ACLR, EVM | RoF | Ideal and no feedback | The findings are attenuation dependent, meaning that with correct attenuation and different optimization algorithms, results similar to the ideal case can be produced. |

23 | Roselli et al. [48] | Electrical | Analog | IMD3 | RoF | Fixed phase for IMD3 components | Large-scale manufacture is problematic since each RoF transmitter requires a different predistorter. |

24 | R. B. et al. [49] | Electrical | DPD | IMD3 | RoF | Fixed correction | entanglement between various pathways. |

25 | Veiga et al. [50] | Electrical | DPD | IMD3 | RoF | Phase maintenance is easy | In the frequency domain, to compensate for arbitrary bandwidth limitations. |

26 | MU. Hadi et al. [51] | Dig | Direct DPD | ACLR, IMD | RoF | Only requires transient chirp coefficient, no exhaustive training. | Limited to only few kilometers of length. |

27 | MU. Hadi et al. [52] | Dig | DPD | ACLR | RoF | Feasible closed loop DPD | High complexity. |

28 | MU. Hadi et al. [53] | Dig | DPD | ACLR, EVM | RoF | DVR, GMP, MP | Training is time consuming. |

Parameters | Values |
---|---|

5G NR Waveforms | ${f}_{c}$ = 3 and 10 GHz Flexible O/G/F OFDM waveform Modulation = 256 QAM |

Laser Diode | $\lambda $ = 1310 nm DD-MZM |

Fiber | SSMF $\mathrm{Optical}\mathrm{Fibre}\mathrm{Dispersion}=16\frac{\mathrm{ps}}{\mathrm{nmkm}}$ Length = 10 km $\mathrm{Attenuation}=0.44\frac{\mathrm{dB}}{\mathrm{km}}$ |

Photoreceiver | $\mathcal{R}$ = 0.69 A/W |

Framework | Parameters |
---|---|

Optimiser | ADAM |

Type of activations function | ReLu |

O/P layer type function | Softmax |

Loss type | Mean Square Error (MSE) |

Hidden Layers N | 15 |

Neurons per layer K | 20 |

Regularization method | L2 |

Regularization factor | 0.001 |

Number of epochs | 100 |

Validation split | 0.4 |

Learning rate Batch size | 0.01 16, 32, 64, 128, 256, 512, 1024 |

Training specimens | 300,000 |

Testing specimens | 300,000 |

Involution | |

(N − 1) K^{2} + (4 + N) K + 6 | 5986 |

DPD | Coefficients | # Coefficients | Arithmetic Operations |
---|---|---|---|

GMP | ${K}_{c}\left({Q}_{c}+1\right){R}_{c}+{K}_{b}\left({Q}_{b}+1\right){R}_{b}+{K}_{a}\left({Q}_{a}+1\right)$ | 84 | 244 |

CPWL | (4M + 1) (K + 1) L | 260 | (K + 1) (14M + 2) L = 880 |

MSA | 2(4M + 1) (K + 1) L | 520 | (14M + 2) (K + 1) = 220 |

CNN | (N − 1) K^{2} + (4 + N) K + 6 | 5986 | $4K+4+\left(N-1\right){K}^{2}=5684$ |

Methodology | E.V.M (%) | MSE (dB) |
---|---|---|

No-Digital Pre-distortion | 11 | −27 |

GMP Digital Pre-distortion | 5 | −30 |

CNN Digital Pre-distortion | 2.1 | −39 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Hadi, M.U.
Practical Demonstration of 5G NR Transport Over-Fiber System with Convolutional Neural Network. *Telecom* **2022**, *3*, 103-117.
https://doi.org/10.3390/telecom3010006

**AMA Style**

Hadi MU.
Practical Demonstration of 5G NR Transport Over-Fiber System with Convolutional Neural Network. *Telecom*. 2022; 3(1):103-117.
https://doi.org/10.3390/telecom3010006

**Chicago/Turabian Style**

Hadi, Muhammad Usman.
2022. "Practical Demonstration of 5G NR Transport Over-Fiber System with Convolutional Neural Network" *Telecom* 3, no. 1: 103-117.
https://doi.org/10.3390/telecom3010006