Design of A Software-defined Underwater Acoustic Modem with Real-time Physical Layer Adaptation Capabilities Emrecan Demirors,∗ George Sklivanitis,y G. Enrico Santagati,∗ Tommaso Melodia,∗ and Stella N. Batalama,y ∗Department of Electrical and Computer Engineering, Northeastern University, Boston, MA Email: {edemirors, santagati, melodia}@ece.neu.edu yDepartment of Electrical Engineering, The State University of New York at Buffalo, Buffalo, NY Email: {gsklivan, batalama}@buffalo.edu ABSTRACT 1. INTRODUCTION This article describes the design of a custom software-defined Underwater acoustic (UWA) communications in combi- modem with adaptive physical layer for underwater acoustic nation with software-defined radio (SDR) techniques have (UWA) communications. The modem consists of a commer- lately attracted considerable attention, mainly because of cial software-defined radio (SDR) interfaced with a wide- the flexibility that SDRs may offer in rapidly time varying band acoustic transducer through amplifying circuitry. With environments. Typical challenges in underwater communi- this custom-built platform, we focus on the unique physi- cations include high path loss, severe multipath, Doppler cal layer challenges of the underwater acoustic channel to spread, limited bandwidth, and fast time-varying channel demonstrate the benefits of real-time adaptation in such [1]. On the other hand, SDR techniques have been envi- rapidly varying environments. We first focus on an Orthog- sioned as a powerful tool for designing cognitive, intelli- onal-Frequency-Division-Multiplexing (OFDM) transmission gently adaptive radio links in radio frequency (RF) com- scheme. In particular, for the forward link, we consider munications [2{4], enabling this way efficient spectrum us- and implement a high-data rate Zero-Padded OFDM (ZP{ age. The work in [5, 6] outlines potential benefits of using OFDM) physical layer with a superimposed convolutional SDRs in UWA communication, and discusses how \cogni- error-correction coding scheme. ZP{OFDM offers high re- tive" and adaptive signal processing techniques may be use- configurability in terms of number of OFDM subcarriers, ful paradigms to shift away from hardware-based modems. modulation type (e.g., BPSK, QPSK), and error-correction Because of the severe multipath channel variations and coding rate. Real-time adaptation at the transmitter is the spatially and temporally variable bandwidth [1] in the achieved through a robust feedback link based on a binary UWA, an adaptive software controlled, easily reconfigurable chirp spread-spectrum modulation (B-CSS). We demonstrate transceiver is highly desirable. Commercially available acous- that joint real-time adaptation of system parameters such tic modems offer limited customization capabilities, while as modulation constellation and channel coding rate leads adding functionalities to proprietary systems is usually dif- to significant data rate increase under preset bit-error-rate ficult and expensive. Conversely, working with a platform (BER) constraints. Moreover, in the same context, we present where most of the functionalities are controlled in software for the first time a seamless switch of our SDR transmitter allows researchers to freely investigate designs where differ- between different signaling technologies such as OFDM and ent parameters can be reconfigured at runtime (e.g., modu- direct-sequence spread-spectrum (DS-SS). lation, carrier frequency, coding). Prior work [7{9] on adap- tive PHY schemes considers UWA modems with limited number of operating modes that offer data recording ca- Keywords pabilities while they consider adaptation to environmental Underwater modem, underwater networks, software-defined variations in the form of extensive offline simulations. The radio, software-defined networks. works in [10, 11] discuss in detail design considerations and present experimental results of a real-time adaptive Orthog- onal-Frequency-Division-Multiplexing (OFDM) technology. However, the fixed hardware of these off-the-shelf acoustic modems supports only limited reconfigurations. Therefore, Acknowledgement: This work is based upon work sup- the need for a modem that supports full reconfigurability to ported in part by the US National Science Foundation under enable cognitive, runtime-adaptive communication systems grants CNS-1422874 and CNS-1126357. is evident [12,13]. Moreover, such SDR-based modem would Permission to make digital or hard copies of all or part of this work for offer the ability to incorporate further algorithmic develop- personal or classroom use is granted without fee provided that copies are not ments to improve its performance through software updates. made or distributed for profit or commercial advantage and that copies bear Based on these considerations, we present our design of a this notice and the full citation on the first page. Copyrights for components custom, high-data-rate, highly reconfigurable software-defi- of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to ned underwater modem and demonstrate the performance redistribute to lists, requires prior specific permission and/or a fee. Request benefits of reconfiguration, real-time adaptation, and phys- permissions from [email protected]. ical layer agility in rapidly varying environments such as WUWNET ’14, November 12 - 14 2014, Rome, Italy the UWA channel. The software-defined underwater modem Copyright 2014 ACM 978-1-4503-3277-4/14/11 ...$15.00 consists of a Universal Software Radio Peripheral (USRP http://dx.doi.org/10.1145/2671490.2674473 N210) interfaced with a Teledyne RESON TC4013 minia- ture wideband hydrophone that is used both for project- USRP N-210 ing and receiving sound in a time division fashion. The FPGA-Xilinx Spartan 3A-DSP software-defined functionalities are implemented mainly in SWITCH the GNU Radio framework that runs on an external host- Acoustic Decim DDC ADC LFRX PreA Transducer PC. For the forward link, we implement a high-data rate UHD Network ZP-OFDM physical layer with a superimposed convolutional Driver Command error-correction coding scheme. The modem offers higher or & Control similar data rates compared to commercial and experimen- Data Streaming Inter DUC DAC LFTX PA tal acoustic modems [1] and other custom built modems [13] CIC offering high reconfigurability in terms of the number of OFDM subcarriers, modulation type (e.g., BPSK, QPSK), and error-correction coding rate. To enable efficient real- Gigabit Ethernet time adaptation both at the transmitter and receiver, we HOST also design a new robust feedback link based on binary chirp spread-spectrum modulation (B-CSS). Finally, we demon- GNURADIO MATLAB strate seamless switch between different communication tech- nologies to highlight the flexibility of the proposed modem in physical layer adaptation. The custom built underwater modems are deployed and tested both in a water test tank and in a shallow outdoor Figure 1: Hardware architecture of the software- lake. First, we evaluate the proposed system setup in terms defined acoustic modem (SDAM). of bit-error-rate (BER) as a function of signal-to-interference- plus-noise-ratio (SINR) at the receiver, for various modula- tion schemes and error-correction coding rates. Second, we to 170 kHz. The USRP's motherboard is equipped with an demonstrate real-time data rate performance versus power analog-to-digital converter (ADC) and a digital-to-analog- consumption tradeoff points under predefined BER reliabil- converter (DAC) that are both controlled by a 100 MHz ity constraints. Finally, we demonstrate for the first time a master clock. The sample rate of incoming digital sam- seamless interchange between ZP-OFDM and direct-sequence ples (from ADC) and outgoing samples (to DAC) is fixed at spread-spectrum (DS-SS) communication technologies at run- 100 Msample=s, while the FPGA digitally interpolates/deci- time. mates to match the hardware sample rate to the rate re- To sum up, the major contributions of this paper are: quested by the user. High rate baseband signal processing • Design of a high-data-rate, highly reconfigurable soft- can be conducted either in the FPGA (Xilinx Spartan 3A- ware-defined modem based on USRP; DSP3400) or in the host-PC, which is connected to the radio through a Gigabit Ethernet (GigE) connection. • design of a new robust chirp-based feedback channel; To implement transmitter and receiver algorithms, we used • design of mechanisms for flexible adaptation of PHY a free and open-source software framework called GNU Ra- parameters (e.g., modulation, error-correction coding dio that is commonly used (i) to drive the USRP operations rate) and seamless switch between different communi- from the host-PC, (ii) as well as implementing signal pro- cation technologies. cessing operations in combination with MATLAB scripting language. GNU Radio offers a broad set of signal processing The rest of the paper is organized as follows. In Section blocks, implemented in C++ that can be used to develop a 2, we describe the modem architecture and discuss the de- large variety of wireless communications applications. These sign and synthesis of a software-defined underwater modem basic C++ blocks are usually wrapped into Python classes from first principles. Section 3 analyzes the physical layer to be instantiated from Python scripts, or used as building schemes for the forward and feedback communication links. blocks of a communication flowgraph from a graphic user Section 4 presents