Optimized Bit-Packing for Bit-Wise Software-Defined GNSS Radio

One way that software-defined radio (SDR) receivers for Global Navigation Satellite Systems (GNSS) differ from traditional GNSS receivers is that radio-frequency (RF) data serialization becomes more than a mere implementation detail deep within some proprietary integrated circuit. GNSS-SDR receivers obtain positioning, navigation, and timing (PNT) solutions in software on a general-purpose processor, enabling rapid development of new PNT technologies. Data serialization and deserialization is the lifeblood of GNSS SDR, and a fundamental tool for cooperation among researchers. The existing ION Metadata Standard facilitates the exchange of RF data with a formal notation (schema) for downconversion, bit ordering, byte ordering, and stream multiplexing. But clear notation is only a first step: serialization directly impacts both computational cost and tracking performance in GNSS-SDR receivers. This is particularly true for "bit-wise" receivers, which re-order computation to loop over bit positions of varying magnitude rather than over samples of varying index. Such receivers prefer input data orderings that separate e.g. runs of high-order bits from runs of low-order bits. This paper addresses the Standard's current inability to describe such a format. However, because it is not always possible to re-order data in hardware to perfectly match the needs of the software-and there are indeed good reasons not to do so, such as version skew-this paper also considers bit-packing for bit-wise SDRs more generally. Taking into account resource costs both in hardware and in software, this paper contributes tools for the development of efficient new bit-packing schemes for next-generation SDR receivers that may feature heterogeneous sample rates, quantization encodings, channels, or even antennas. These tools assist in the development of efficient hardware and software for bit-wise GNSS SDR by optimizing "bit-packing schedules" and producing both VHDL code for packing and C code for unpacking RF data. An evaluation of the resulting code is given in terms of operations per sample. Finally, this paper presents "first light" experimental results for a new low-cost dualantenna, tri-band GNSS-SDR front end developed in-house at the Radionavigation Laboratory-the proximate motivation for the entire work.