1. Introduction

With the advancement of science and technology and the growing demand for resources, mankind has already moved footsteps toward oceans, and yet, has laid out many network nodes on seafloor of different areas. The data collected by those nodes are transmitted toward underwater platforms, water surface or shore through wired cables, or underwater wireless methods like acoustic communication or optical communication. The usage of underwater wireless communications is easier and more convenient in maintenance, with great flexibility and minimum constrains on the movement of underwater devices during data transmission. Underwater acoustic communication (UAC) is competent for tens of kilometers of long distance communications but subjected to some intractable problems such as limited bandwidth, large transmission delay and multipath effects. On the other hand, underwater wireless optical communication (UWOC) is restricted within hundreds of meters nowadays but benefits from rich bandwidth resource, low latency, high security and small device footprints [1–9]. Theoretically, C. Wang et al. predicted that the UWOC is feasible for more than 100-m transmission in clean ocean and 300 m in pure sea water [1]. And practically, researchers in recent years have achieved UWOC with tens of meters’ distance with high data rate. In 2017, Liu et al. summarized latest UWOC milestones and demonstrated a 34.5-m, 2.7-Gbps UWOC system in tap water based on on-off keying (OOK) modulation and a 19.4-mW, 520-nm laser [2]. In the same year, Y. Chen et al. achieved a 5.5-Gbps UWOC with a green laser and orthogonal frequency division multiplexing (OFDM) through a hybrid air-water channel (5 m in air and 21 m in tap water) [3]. Others like C. Shen [4] and C. Fei et al. [5], achieved UWOC of 20 m and 15 m in tap water channels, using OOK and OFDM respectively, both with high data rate as well. Till now, most of the UWOC experiments were conducted in tap water channels, using OFDM or OOK as modulation formats [2,4] while using APDs for signal detection. As a multi-carrier transmission scheme, OFDM features high spectral efficiency and strong resistance to multipath effects, and has been widely used in UWOC to acquire high data rate. However, it might be too complex for some applications that are sensitive to power consumption and merely require moderate data rate. As for OOK, it is a simple modulation scheme, especially suitable for UWOC systems where multipath effect is not obvious. Those experiments based on OFDM and OOK did obtain promising results, especially for high data rates. However, they may be not so attractive for some practical scenarios with very tight optical power budget yet still requiring long distance transmission, such as the internal communications among a swarm of autonomous underwater vehicles (AUVs) or underwater sensor nodes.

In this paper, we propose a UWOC scheme based on pulse position modulation (PPM) and multi-pixel photon counter (MPPC), enabling a 46-m underwater transmission with ultra-low transmitting optical power. Before our work, C. Gabriel [10] and S. Meihong [11] et al. had discussed the suitability of PPM format in UWOC. D. Anguita et al. [12] designed a 10-15 m UWOC system for underwater wireless sensor networks based on 2-Mbps 4-PPM and 1-Mbps 16-PPM. Compared with the aforementioned modulation formats, PPM features higher energy efficiency and lower requirement on signal-to-noise ratio (SNR) for a given bit error rate [13,14], making it competitive for photon-starved long-reach UWOC systems. Meanwhile, for underwater optical signal detection, high sensitivity is a primary requirement due to the heavy attenuation of water on light. “Geiger mode” APD-based single photon avalanche detector (SPAD) can detect low energy signal through multiplying initial photo-generated carriers. As those carriers hit the lattice in avalanche gain region under high reverse bias, they produce more carriers. All these carriers will eventually reach a high or even saturated output current. However, no matter the avalanche process is triggered by how many photons, once the generated current is saturated, it will not decrease until a quenching circuit helps diminish it. The time span from receiving photons to the quenching of current is called dead time, during which, SPAD cannot detect any incident photon. It means SPAD can only qualitatively tell whether there is incident light or not while cannot distinguish the light intensity. To overcome this problem and keep the high sensitivity at the same time, multiple SPAD units can be arrayed in parallel like pixels, such that light intensity could be quantified through the number of SPAD units that receive photons and cause avalanches. Such a pixel array is called an MPPC, which also supports lens-free light detection due to its relatively large active area.

2. Operation principles

In PPM, information is encoded with short pulses with the same width and amplitude, implying high peak power for a given average power. For PPM, the information of m bits is encoded by transmitting a single pulse in one of the L = 2^m possible time slots. If we note the m-bits binary data as k = (a₁, a₂… a_m), then the pulse position in the 2^m time slots is $S_k=1+{\displaystyle \sum {}_{i=1}^m}{2}^{m-i}{a}_i$. In this paper, the frame structure of the PPM is shown in Fig. 1(a). An entire frame consists of L symbols, with each symbol duration equals to the slot width as well as the bright pulse width, and there is no guard interval between frames. The major problem of PPM is that the relatively high energy efficiency and pulse power comes at the cost of reduced spectral efficiency, thus it is less competitive in bandwidth-starved systems, such as highspeed wired communications. Fortunately, it is not a severe problem thanks to the relatively rich channel bandwidth in UWOC, especially when considering the practical requirement on data rate for a majority of underwater applications.

 

Fig. 1 (a) Frame structure of the PPM signals. (b) Laser driving voltage vs laser current (V-I curve) and optical power versus laser current (P-I curve). (c) The 4-PPM signal captured by the MPPC with a slot width of 200 ns.

Download Full Size | PDF

Since the length of the underwater channel in the experiment is limited to 46 m, we have to reduce the transmitting optical power to investigate the high sensitivity of the MPPC. Even when the laser works under spontaneous emission state, the MPPC may still guarantee a 46-m UWOC. In our experiment, a 450-nm blue laser is used, and Fig. 1(b) shows the curves of laser’s operation voltage/optical power versus driving current. We chose the laser due to its wide range of spontaneous emission state, which could help us study the UWOC performance when the laser works under both stimulated emission and spontaneous emission states. The 3-dB bandwidths of the laser and the MPPC in the following experiment are around 500 MHz and 4 MHz, respectively, so the bandwidth of the UWOC system is mainly restricted by the MPPC, implying a system bandwidth of around 4 MHz. As an example, when the laser is driven by a 4-PPM signal with a slot width of 200 ns, the captured waveform by the MPPC is well recovered, as shown in Fig. 1(c). Note that the power consumption of the MPPC is around 100 mW.

3. Experimental setup

In the experiment, PPM was employed as the modulation format and a MPPC was utilized as the receiver, with the experimental setup being shown in Fig. 2(a). A 46-m white Polyvinyl Chloride (PVC) tube that consists of eleven 4-m tubes and twelve tube connectors was filled with tap water to simulate a 46-m underwater channel. The diameter of the tube is 20 cm. Firstly, we generated a string of random binary symbols on personal computer and then encoded it in the PPM format. Then the baseband PPM signal was generated via an arbitrary waveform generator. 5 MHz and 10 MHz time slot frequency was used in the experiment. The amplitude of the PPM signal was further amplified by a power amplifier with a passband ranging from 2.5 kHz to 500 MHz and a fixed gain of 25 dB. A key-press variable electrical attenuator was placed following the power amplifier to adjust the signal amplitude according to the laser driving demand. A DC bias voltage was combined with the adjusted PPM signal via a bias-tee in case the laser needs to work under the stimulated emission state. Due to the low frequency cutoff of the power amplifier, the output PPM signal dropped down by a voltage that equals to the mean amplitude of the input signal, and thus an extra DC voltage was required for proper compensation. Finally, a 450-nm NDB7875 blue laser was driven by the combined signal from the bias-tee. A variable metallic neutral density optical filter was placed after the laser to manually adjust the optical power that fed into the underwater channel. The total time slots of the generated PPM signal were more than 10⁵ to ensure the reliability of the experimental results.

 

Fig. 2 (a) Experiment setup of the proposed 46-m UWOC system using an MPPC receiver. AWG, arbitrary waveform generator; PA, power amplifier; VEA, variable electrical attenuator; DC, direct current; BT, bias-tee; LD, laser diode; OF1, variable metallic neutral density optical filter; OF2, 450 nm optical filter; OSC, oscilloscope. (b) The 46-m PVC tube filled with tap water to simulate a 46-m underwater channel.

Download Full Size | PDF

The 46-meter-long underwater channel placed strict requirements on the alignment between transmitting and receiving sides. We adjusted the pitch angle of the laser through a pitch adjustment pedestal and adjusted the laser horizontal deflection angle manually to assure both sides align. At the receiving side, the optical signal was filtered by a 450-nm optical filter before being captured by an MPPC. The output of the MPPC was then recorded and displayed on a Tektronix DPO71254C mixed signal oscilloscope (OSC). The recorded signal was sent to a computer through a local area network, for offline decision, slot error rate (SER) calculation and demodulation. Note that the MPPC was covered with black bags during the experiment to reduce background noise. We assume the practical deployment scenarios of the proposed scheme will be in the deep sea where the background light is extremely weak.

4. Experimental results

We first investigated the communication performance of the proposed system when the laser worked under a typical condition, i.e. under the stimulated emission state, by properly adjusting its DC bias. Since photon arrivals were discrete and random in nature, fluctuation in the amplitude of the detected bright pulses was observed. And to obtain stable amplitude, the transmitting optical power to the underwater channel was deliberately increased by adjusting metallic neutral density optical filter, such that the output of the MPPC was slightly saturated. As mentioned earlier, the number of time slots for m-bit data is denoted by L. If omitting the optical power caused by DC bias, when L was halved, the average optical power of the PPM signal was doubled, since the occurrence of the bright pulse was doubled, according to the PPM frame. Based on the waveforms recorded by the OSC, eye diagrams for different L-PPM can be plotted, as shown in Fig. 3(a)~3(j). As a benchmark, the back to back eye diagram of the 4-PPM signal with a 5-MHz time slot frequency is shown in Fig. 3(l). For the PPM signal with 5-MHz time slot frequency, data rates for the cases L = 4, 8, 16, 32 and 64, were 2.5 Mbps, 1.875 Mbps, 1.25 Mbps, 0.7813 Mbps and 0.4688 Mbps, respectively. As for the 10-MHz time slot frequency, data rate for each L-PPM was doubled comparing with the case of 5-MHz. All eye diagrams are wide open, and no error slot was observed. Thus, the demodulated signals were completely correct.

 

Fig. 3 (a)~(e) Eye diagrams of 4-PPM, 8-PPM, 16-PPM, 32-PPM and 64-PPM before demodulation, respectively, when the time slot frequency was 5 MHz and the laser worked under stimulated emission state. (f)~(j) Eye diagrams when the time slot frequency was changed to 10 MHz. (k) Enlarged partial eye diagram of (j). (l) Back to back 4-PPM eye diagram.

Download Full Size | PDF

Furthermore, it can be observed from the eye diagrams that all high levels are thin, as the MPPC in the experiment was slightly saturated, whereas the low levels are relatively thick due to the baseline wandering effect on the low-level slots. One way to eliminate those thick bottoms is to remove DC bias such that the voltage of low-level slots is 0 and the laser is driven only by the baseband PPM signal. And to make laser work in spontaneous emission state, the driving voltage was set to be slightly smaller than the threshold voltage of the laser. In this case, laser output power was lower, and the laser worked like a light-emitting diode (LED). Eye diagrams for this case are shown in Fig. 4. Again, no error slot was found in these results. As expected, both the high and low levels of the eye diagrams are thin when the laser working in the spontaneous emission state.

 

Fig. 4 (a)~(e) Eye diagrams of 4-PPM, 8-PPM, 16-PPM, 32-PPM and 64-PPM signals before demodulation, respectively, when the time slot frequency was 5 MHz and the laser worked under spontaneous emission state. (f)~(j) Eye diagrams when the time slot frequency was changed to 10 MHz. (k) Enlarged partial eye diagram of (j).

Download Full Size | PDF

The required transmitting optical powers (in dBm) into water at the transmitting side for different L-PPM signals are shown in Fig. 5. The maximum transmitting optical power among them is −7.6 dBm, for the 10-MHz 4-PPM signal, when the laser worked under stimulated state. Meanwhile, the lowest power even reaches below −27.7 dBm, for the 5-MHz 64-PPM signal, when the laser worked under spontaneous state. The corresponding received powers for both cases at the receiving side, were measured to be −39.2 dBm and −62.8 dBm, respectively.

 

Fig. 5 Transmitting optical power for different L-PPM signals, stimulated: laser worked under stimulated emission state; spontaneous: laser worked under spontaneous emission state.

Download Full Size | PDF

5. Discussion

Although the demonstrated transmission distance of the system is 46 m, we envision that the distance can be further extended by increasing the transmitting power or using higher order PPM. As an alternative to the semiconductor laser used in the experiment, solid state pulse lasers can be used to realize high pulse power. Differential pulse position modulation (DPPM) and digital pulse interval modulation (DPIM) also offer more possible design options for the UWOC system. More importantly, according to the results when the laser worked under spontaneous emission state, the MPPC has demonstrated its potential for designing a long-reach UWOC system even using low-cost light-emitting diodes as the transmitter.

6. Conclusion

In summary, we proposed a 46-m UWOC scheme based on pulse position modulation and multi-pixel photon counter, which could exactly meet the strict requirement on power efficiency in photon-starved long-reach UWOC. We have evaluated the system performance with different PPM signals (time slot frequency: 5/10 MHz), ranging from 4-PPM to 64-PPM, with a 450-nm blue laser that worked under both stimulated and spontaneous emission states. The transmitting optical powers into water were all below −7.6 dBm, with the lowest being merely −27.7 dBm. According to the obtained eye diagrams and SERs, error-free communication was demonstrated even under such low transmitting optical powers.