Implementation of a Coded Modulation for Deep Space Optical CommunicationsMichael K. Cheng, Bruce E. Moision, Jon Hamkins, and Michael A. Nakashima Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, 91109-8099 Email: mkcheng, bmoision, jhamkins, mikenakajpl.nasa.govAbstractWepresentafieldprogrammablegatearray (FPGA) implementation of a turbo-like decoder for a serially concatenated pulse-position modulation (SCPPM) code. NASA developed this coded modulation scheme for deep space com- munications from Mars. Under a nominal mission condition, the SCPPM coded system can operate within a one dB signal energy gap from capacity. The structure of SCPPM makes direct application of theconventional turbo decoding algorithm very inefficient. Here, we describe techniques to increase the throughput and performance of a hardware SCPPM decoder. Using our optimizations, we demonstrate a 6 mega-bits per second (Mbps) decoder realization on a single FPGA. Extension to a higher data rate decoder using multiple FPGAs is readily achievable. Similar codes designed forthe optical channel can benefit from our optimization techniques.I. INTRODUCTIONCommunication over deep-space is difficult. Communica- tions beams spread as the square of the distance between the transmitter and the receiver. For example, geosynchronous Earth orbit (GEO) satellites are about 40,000 kilometers (km) in altitude and the average Mars-Earth distance is 80 million km. Therefore, the extra distance that a communication beam would have to travel from Mars to Earth would make datatransfer 4 million times more difficult than from a GEO satellite to Earth. The signal power required to meet this extra effort and to cover this distance squared loss is greater than 66 dBs! One way to increase the transmission rate from deep-space is through the use of more powerful transmit and receive antennas. However, this comes at a cost in increased antenna sizes which makes realization impractical. Another way is to communicate using frequencies much higher than radio frequency (RF) such as that of optical signals. Beams at higher frequency are more directionally concentrated and this allowsa more efficient reception of the transmit energy 1, Ch. 1. NASAs legacy error-correcting code (ECC) design for RF communication is the concatenation of an inner convolutional code and an outer Reed-Solomon (RS) code 2. Decoding is performed in one pass utilizing hard bit-decisions. The discovery of turbo codes 3 and their suboptimal but effective low-complexity iterative decoding provided NASA a new codefamily with improved coding gains. NASAs first use of turbo codes is on the Messenger spacecraft launched in August of 2004.An efficient ECC design for the deep space optical channel is the serial concatenation of an inner high-order modulationcode and an outer convolutional code, namely serially con- catenated pulse-position modulation or SCPPM. 4. We may approximate true ML decoding while limiting the SCPPM decoder complexity by iteratively decoding the modulation and the ECC. This is in fact the “turbo” principle and more details can be found in 5. This article is a companion to 6 and focuses on design issues critical to hardware implementation that are often not addressed in a high-level design. We also present novel techniques that optimize the SCPPM hardware decoder. The organization is as follows: in Section II we provide a model of the optical communications channel. In Section III, we give an overview of the SCPPM code and its decoding algorithm. In Section IV, we discuss some of the challenges associated with hardware implementation of the SCPPM decoder and describeour efficient approaches in detail. In Section V, we present a fast prototype decoder along with its hardware resource usage and error rate performance.II. SYSTEMDESCRIPTIONWe consider an optical communications system that uses direct photon detection with a high-order pulse-position mod- ulation (PPM) 1, Ch. 1.2. An M-order PPM modulation uses a time interval that is divided into M possible pulse locations, but only a single pulse is placed into one of the possible positions. The position of the pulse is determined by the information to be transmitted. A diagram of the optical communications system in discussion is shown in Fig. 1. The information bits U = (U1,U2,Uk) are independent identically distributed (i.i.d.) binary random variables assumed to take on the values 0 and 1 with equal probability. The vector U is encoded to C = (C1,C2,Cn), a vector of n PPM symbols. At the receiver, light is focused on a detector that responds to individual photons as illustrated in Fig. 2. For each photon sensed, the detector produces a band-limited waveform for input to the demodulator. This waveform is used to estimate the photon count, ki, within each slot i. On the Poisson channel, a nonsignaling slot has average photon count nband a signaling slot has average count ns+ nb