Evaluation of Communication Link Performance and Charging Speed in Self-Powered Internet of Underwater Things Devices

The energy consumption of the Internet of underwater things (IoUT) nodes is a capital aspect that dramatically affects the applicability of wireless optical technologies in several scenarios, such as ocean monitoring or underwater sensor networks. Simultaneous lightwave information and power transfer (SLIPT) is a cost-effective and energy-efficient solution for energy-constrained wireless systems. Nonetheless, the reported battery-charging times for underwater operations are high, which should be improved to make this technology attractive enough to be considered for actual implementations. This paper provides a new SLIPT strategy, introducing a specific optical signal transmission scheme by controlling the transmitted direct current (DC) level component. The scheme is based on a DC-On Off Keying (OOK) modulation with an adjustable signal range (SR) to improve the energy-harvesting process and battery-charging time for underwater operations. The results reveal that the system provides a signal-to-noise ratio higher than 28 dB and the bit-error rate of less than 10−10 which is below the forward error correction limit with improved charging time around 30 minutes and 15 seconds for 5 F and 9.4 mF, respectively over 20 cm, and 63 seconds for 9.4 mF in 50 cm link distance.

solar cell for EH and an a-Si solar cell for ID. In addition, 81 a hardware pre-equalizer to improve the 3-dB bandwidth was 82 used. However, the mentioned system focused on communi-83 cation performance, and the Si solar panel for EH was not 84 connected to the system to operate underwater. The results 85 showed the system was capable of achieving orthogonal fre-86 quency division multiplexing (OFDM) signals at the data rate 87 of 1.2 Mbps over a transmission distance of 2 m under turbid 88 water [18]. 89 The studies mentioned above are mainly focused on 90 communications performance. Therefore, solar panel per-91 formance in the case of EH needs to be investigated and 92 optimized. In [19], the authors proposed a time-splitting 93 SLIPT system using a 430 nm blue laser and 5.5×7 cm 2 solar 94 cell to charge a battery with 840 mW of power capacity. The 95 observed charging time of the battery was 124 minutes. Once 96 fully charged, this hardware could activate a temperature sen-97 sor in a water tank for more than 2 hours. The sensor mea-98 sured the temperature when the battery's voltage was higher 99 than the threshold voltage of 3.6 V. Otherwise, it entered 100 sleep mode to recharge the battery. Furthermore, a data rate 101 of 500 Kb/s over a 1.5 m link range was achieved. In addition, 102 De et al. used a 4.8 W LED as a transmitter for charging 103 a 5 F capacitor to power an Internet of underwater things 104 (IoUT) device, which took 90 minutes over a 30 cm link 105 distance. Once the capacitor was fully charged, the device 106 sent a real-time video streaming for 1 minute. Although the 107 charging time for EH was considered in the mentioned work, 108 no approach was proposed to decrease it. However, in [11] the 109 optimization of the splitting factor (either in time or power) 110 in terms of harvested energy while satisfying BER and spec-111 tral efficiency constraints was analyzed. In addition, closed-112 form equations were derived for the average harvested energy, 113 BER, and spectral efficiency under the presence of under-114 water turbulence, which was modeled using a log-normal 115 distribution. Uysal et al. improved the charging time by an 116 optimization spilliting factor. However, to the best of the 117 author's knowledge, the impact of the transmitted waveform 118 on EH and communication performance in SLIPT UWOC 119 systems has not been analyzed. 120 In this work, a new power-splitting SLIPT scheme 121 is proposed to improve the energy harvesting rate for 122 VOLUME 10, 2022 IoUT applications. The proposed system comprises a white-light LED (WLED) source and a solar cell as a receiver 124 for both EH and ID. White-light laser sources are more favor-125 able in increasing the energy conversion efficiency of the 126 solar cells than red/blue-light lasers [14], [20]. However, LDs 127 require better alignment compared to LEDs, and their lifetime 128 is limited. Furthermore, non-coherent light performs better in 129 turbulent water [12], [21]. The main objective of this contri-   The rest of this work is organized as follows. In section II, 145 the system architecture is represented. Then, in Section III, 146 the simulation and the experimental setups are described.

147
Next, the results are explained and discussed in Section IV.

148
Finally, the concluding remarks are provided in Section V. Figure 1 shows a block diagram of a power-splitting UWOC

151
SLIPT system. A non-return-zero (NRZ) On-Off Keying 152 (OOK) data signal is generated and is sent to a controller 153 module which introduces different DC levels (.e.g., Level 0, 154 Level 1, Level 2), which are provided by V g1 or V g2 with data 155 signal of V g0 to the WLED's driver (the scheme and perfor-156 mance of the driver will be detailed in this paper). Moreover, 157 a power supply is used to provide the DC bias (V cc ) to supply 158 the WLED transmitter. The modulated light beam (P tx ) is 159 transmitted through the underwater channel to the receiver, 160 where the incoming optical light (P rx ) is collected by the solar 161 cell and converted to an electrical signal (P s ). The received 162 power is split into α P s and (1 -α) P s quantities for EH and ID 163 by a splitter, respectively. The receiver can detect and decode 164 the data signal when there is enough harvested energy (P h ). 165 In this work, the performance of a UWOC system with 166 SLIPT capability using the power splitting technique for both 167 EH and ID is analyzed. Nevertheless, it mainly focuses on 168 improving the energy harvesting rate, which determines the The main problem for EH in underwater environments 175 is the long time needed to recharge the capacitor or the 176 battery due to the high attenuation of the transmitted power 177 and lack of enough ambient light. Therefore, improving the 178 charging speed is essential for optimizing different system 179 features, such as response time or data-logging periods. Once 180 the capacitor is recharged to the threshold voltage, the har-181 vested energy can be applied to supply IoUT nodes [22]. 182 The recharging time depends on several factors, such as aver-183 age transmitted power, link distance, solar cell size, power 184 conversion efficiency, battery capacity, etc. In addition, the 185 average transmitted power can be controlled by adjusting the 186 signal DC level, affecting the energy harvesting performance. 187 In this study, different discrete DC levels have been defined, 188 and the best-performing ones can be selected depending on 189 both the channel and device state conditions. The variation 190 in the DC level component introduces changes in the SR 191 of the AC (data signal) component since the higher the DC 192 level, the lower the SR. Hence, a trade-off between EH and 193 communication performances can be considered for selecting 194 the appropriate DC level. Figure 2 shows the proposed trans-195 mission circuit scheme. In this circuit, three MOSFET tran-196 sistors have been considered. M 0 controls an OOK modulated 197 signal. Moreover, by switching on M 1 or M 2 , the low level of 198 the OOK signal rises. As a result, the DC level of the sig-199 nal increases, thus increasing the average transmitted power. 200 However, the SR of the AC signal decreases. A generaliza-201 tion of this scheme includes several MOSFETs to provide a 202 range of discrete DC levels based on the value of resistors 203 (R 1 , R 2 , . . . , R i , i ≥ 1) in the MOSFET drains. In this work, 204 the proposed system provides four possible DC levels by 205 setting the different MOSFET in ON-OFF states. As a gen-206 eral rule, N transistors can provide 2 N−1 different DC levels. 207 Therefore, depending on channel and device state conditions, 208 the DC level can be adapted to increase harvested energy or 209 SR. For instance, higher DC levels reduce the battery charg-210 ing time. Conversely, lower DC levels provide higher SRs and 211 improve SNRs. The output optical power is given by: where P tx is the output optical power waveform, P max is the 214 maximum output power controlled by R 0 (shown in Fig. 2), 215 and s(t) is the transmission OOK waveform (i.e., 0 or 1).

217
As shown in Fig. 3, the obtained signal from the solar 218 cell is divided into two branches for EH and ID, and the 219 photo-generated current I s in an aquatic environment is 220 governed by:

238
where I D s , I D sh and I F are the current of diode D s , the EH and 239 ID branches current, respectively.

240
The current of each stage from Fig. 3 can be expressed as: 241 243 and where I s 0,1 is the reverse saturation current, q is the elemen-245 tary charge, k is the Boltzmann constant, T is the absolute 246 temperature, n 0,1 is the diode ideality factor, V T is the thermal 247 voltage, and Z F is the equivalent impedance of the ID stage, 248 including the load impedance of the amplification front-end. 249 Z F is obtained as: In addition, V D s and V D sh are the voltage across the D s and 252 D sh , respectively as given by: and the voltage across the capacitor is defined as: By substituting 4, 5 and, 6 in 3, we have: This equation is nonlinear. Therefore, to simplify the 259 receiver circuit model, two stages can be considered. (i) At the 260 beginning of the process, the LED starts the energy emission, 261 and C in begins charging from the constant current provided by 262 the solar cell (I D sh ), and the rest of the circuit can be neglected. 263 In this state, the equivalent impedance in the EH branch is 264 less than in the data detection branch. Thus, the received 265 signal passes to the EH branch more than the ID branch. 266 Accordingly, the bandpass filter output (V sig ) is minimum. 267 In this configuration, the diode D sh is in ON mode with a 268 short circuit behavior (Z D sh ≈ 0). This state is maintained 269 until V in reaches the threshold voltage of the boost converter 270 (in t = t s ), which enables it to supply the output voltage of 271 3.3 V. Then, the rest of the circuit begins operating and con-272 suming the energy stored in the input capacitor through the 273 boost converter. In the case that the transmitter can provide 274 enough optical power to the system, the input capacitor can 275 provide the necessary power to maintain the circuits work-276 ing and continue charging. Following the diode I-V curve in 277 Fig. 4, as V in increases, the diode voltage reduces, and its 278 impedance increases. (ii) When V in reaches a voltage higher 279 than (V s − V D sh ), the diode enters OFF mode, and the equiv-280 alent circuit consists of the bandpass filter only. Considering 281 the voltage divider, V sig yields the following expression:  voltage variation becomes smaller as the current increases, 305 with the possibility of saturating in environments includ-306 ing the presence of other light sources. However, the pro-307 posed system works perfectly in a deep, low-light marine 308 environment.

310
This section presents the performed analysis to validate the 311 features of the proposed scheme described in the previous 312 sections. The system is evaluated analytically through simula-313 tion and experimentally under a laboratory testbed condition, 314 which will be described later in the paper. In section III-A, 315 the system performance of the proposed transmission and 316 reception schemes with their results are described, and in 317 section III-B, the proposed solar cell-based UWOC system 318 is experimentally evaluated in a water tank filled with tap 319 water.   switched on while M 0 is controlled by a 100 Hz NRZ-OOK 371 modulation since it is limited by the 3-dB modulation band-372 width of the used solar cell. After transmitting through a water 373 tank filled with tap water in 20 cm and 50 cm link distances, 374 a light spot is formed on a 5 × 7 cm 2 solar cell. The proposed 375 system is based on the power splitting technique. Therefore, 376 the received power is divided into two streams for detecting 377 the signal and harvesting the energy as depicted in Fig. 3. 378 The water tank side walls are covered by a black material to 379 avoid reflections. The 3-dB system bandwidth is limited by 380     9.4 mF) and three different DC levels in the transmitted signal 400 over a transmission distance of 20 cm. The same metrics 401 using a 9.4 mF input capacitor are also evaluated in a 50 cm 402 link range and presented in Fig. 11. Due to the long charging 403 time of the 5 F capacitor caused by the low average received 404 power in the 50 cm link range due to high attenuation, the 405 obtained results are not applicable for practical implemen-406 tation. Therefore, to account for this in Tabs. 3, 4, the label 407 (Not-measured.) is mentioned. As illustrated, the SR varies 408 during the charging time of the input capacitor, reaching its 409 maximum value when the capacitor is fully charged. The 410 charging time (t fc ) of the input capacitor for 5 F and 9.4 mF 411 in different DC levels of the transmitted signal and LED 412 illuminances are provided in Tab.3. Moreover, the maximum 413 current (I out ) that the boost converter can provide while keep-414 ing the output voltage constant at 3.3 V for 5 F and 9.4 mF in 415 20 cm link distance (d) is measured. However, the obtained 416 current for the 50 cm link range is negligible to mention in 417 the table and is displayed as a (Negligible) label. Table 4 418 demonstrates the system SNR over a transmission distance of 419 20 cm and 50 cm in two cases: once the input capacitor is fully 420 charged (SNR fc ) and once the boost converter starts working 421 (SNR s ). As shown, the obtained SNR in both starting the 422 boost converter and when the input capacitor is fully charged 423 is significantly promising. As depicted, the SNR decreases 424 by increasing the DC level where the dominant effect for 425 decreasing the SNR is reducing the SR, not increasing the 426 noise due to the low variation of the standard deviation by 427 increasing the transmitted power based on the solar cell I-V 428 curve. The estimated BER for an NRZ-OOK modulation for 429 each measured system SNR is given [23]: where, erfc(x) is the complementary error function. Accord-432 ing to (12), the maximum BER is lower than 10 −10 for the 433 obtained SNR of 28.42 dB, which is below the forward error 434 correction (FEC) limit.    20 cm link distance, and 12 seconds, 63 seconds for the 50 cm 446 link distance, respectively. In all the cases, the system SNR 447 is higher than 28 dB, and the BER is below the FEC limit;  he works on high-performance coding and modu-616 lation schemes and channel estimation techniques. 617 He has been a Researcher in different national and international projects 618 financed by national and European administrations and companies. He is 619 also the author of three book chapters, more than 30 articles in international 620 journals, and more than 90 conference papers.