Old and new concepts in undersea networks

Abstract

The past decades of development of underwater acoustic communications (acomms) have produced many advances in physical layer algorithms, in hardware and firmware capabilities tailored to the underwater environment, and in concepts for networked sensors. In general, most deployed systems and most conceptual networks have assumed a strictly half duplex channel wherein all links must be shared over a bandwidth-constrained channel. This has led to an almost universal insistence on avoiding collisions among receptions, which might be achieved by TDMA schemes or by adapting other network concepts derived from (RF) mobile communications technology. For long duration deployment of underwater systems, any requirement for tight TDMA access forces the modem developer to use power-hungry high accuracy clocks, thus quickly draining precious battery resources. While networked acknowledgements at the speed of light make perfect sense, the latency of the acoustic channel demands excessive consumption of time to enable the network simply to know that a message has ultimately been received. There are many studies which attempt to devise minimally intrusive ARQ approaches, but ultimately acknowledgements in some form are necessary. In this paper we revisit some older technology, combine it with new capabilities in hardware, and add some signal processing insights to support discussion of a back channel providing asynchronous, multi-access command and control of a network, with supervised use of a second acoustic channel for high data rate, priority, and/or large data volume transfer. The two contending modulation schemes for asynchronous signalling have long been frequency hopping and direct sequence (DS), both of which are (typically) non-coherently processed spread spectrum methods. For a given bandwidth and data rate in additive white Gaussian (AWGN) the performance of the two is virtually identically. However, in a dynamic, multipath-rich environment, the well-recognized requirement for power control for DS signalling, given the constraint of channel latency, makes FH signalling the preferred choice. Justification for this is found throughout the development of the (NATO) JANUS underwater protocol, which itself is a frequency hopping scheme. In the context of power control, it is possible to transfer the burden of interference mitigation from the multiple transmitters to the receiver at a single modem by judicious spectral whiting and clipping of too loud interference. This not only compensates for variable strength transmissions, but for external sources of interference. Our anecdotal experience shows that a single modem operating this way in an 8 kHz band at a center frequency of 39 kHz can readily accommodate asynchronous reception of at least five messages from irregularly-spaced transmitters. Hardware is now available to support integration of two or more modems into a single system, especially to provide dual frequency acomms. From the hardware/system perspective the main constraint is providing sufficient spectral isolation between the two channels. In our experience, this requires physical separation and spectral filtering between the transducers to overcome high power sidelobe interference which is at best 50 dB below the level of the primary channel. However, given that these constraints are accommodated, we are positioned to consider development of media access control (MAC) and networking systems that benefit from this full duplex approach. We describe an FH signal as defined for the JANUS application and show how it can be used in multi-access applications. We provide an overview of the physical infrastructure we deploy to accommodate the two frequency bands, and we describe a possible network infrastructure utilizing these capabilities. Our goal is to promote discussion of alternatives to conventional half duplex TDMA and conflict avoidance acomms schemes.