gr-lora.zip

# gr-lora This is an open-source implementation of the LoRa CSS PHY, based on the blind signal analysis conducted by @matt-knight. The original research that guided this implementation may be found at https://github.com/matt-knight/research ## LoRa LoRa is a wireless LPWAN PHY that is developed and maintained by Semtech. It is designed to provide long range, low data rate connectivity to IoT-focused devices. A reasonable analogy is that LoRa is like cellular data service, but optimized for embedded devices. ## Implementation for Gnuradio Since LoRa is closed-source technology, all open source implementations may differ from each other. Currently, there exist two versions of lora implementation for gnuradio simulation and experiments on SDR devices, e.g. USRP. We call them version [rpp0](https://github.com/rpp0/gr-lora.git) and version [Bastille](https://github.com/BastilleResearch/gr-lora.git). - **Version rpp0** is implemented on gnuradio 3.8. It only provides receiver module including demodulation and decode part. The receiver can demodulate and decode messages emitted by RN2483 nodes with a USRP source. This implementation is relatively more robust than the other one. (This is my own feeling.) - **Version Bastille** is implemented on gnuradio 3.7. It provides both transmitter and receiver, i.e. encoder, modulator, demodulator can decoder. The repository is not updated for several years. Still there are some problems in the orginal version: 1) The encoder only implements situations for SF = 8. 2) Gnuradio 3.7 is based on python 2.7, which is out of date. 3) The transmitting-receiving process only works for low data rate mode. 4) The performance is highly affected by ```fft_size_factor```, which is confusing. A larger fft_size_factor leads to worse performance. 5) Continous check is not robust for ```preamble detect``` and ```SFD found```. Note that 0 and 2^SF-1 is neighbors. 6) Both transmitter and receiver keep a sampling rate of bandwidth without oversampling. This can be not stable. ## Lora modulation-demodulation In our repo, we mainly focus on modulation-demodulation. Lora is modulated by linear chirps and demoluated by fft. Consider parameters: - ```SF```: spreading factor - ```BW```: bandwidth - ```Fs```: sampling rate Then a symbol is within interval [0, 2^SF - 1]. Assume that value of some symbol is ```n```. We use ```N``` to refer to 2 ^ SF, i.e. . Each symbol occupies a time interval . Then a standard upchirp is =exp(2{\pi}jt(-\frac{1}{2}BW+\frac{BW}{2T}t+\frac{n}{T}))), where ```0 <= t < T``` The downchirp can be derivated in the same way. Suppose , ```K``` is usually an integer. Then we have the discrete signal after samling: =s(\frac{i}{KBW})=exp(2{\pi}ji(-\frac{1}{2K}+\frac{i}{2K^2N}+\frac{n}{KN}))), ```0 <= i < KN``` ## Improvement Therefore, we have refined implementation based on version Bastille. We have addressed problems ```iv, v, vi``` mentioned above. ## Configuration - Spreading Factor: Number of bits per symbol, typically [6:12]. - Code Rate / # Parity Bits: Order of the Hamming FEC used. The number of data bits per codeword is always 4, but the number of parity bits can range from [1:4]. - Header: Whether frames produced/consumed by the frame will contain the 8-symbol header. - FFT Window Beta: Controls the shape of the Kaiser windowing curve that is applied to the FFT input IQ. - FFT Size Factor: Multiplier applied to the width/number of bins of the FFT. A multiplier of 1 yields 2\*\*spreading_factor bins, the minimum number required by the modulation. Received symbols are divided down to map within the valid range of [0:(2\*\*sf)-1]. Initial experimentation reveals 2 yields good performance. ## Installation ``` git clone git://github.com/BastilleResearch/gr-lora.git cd gr-lora mkdir build cd build cmake ../ make sudo make install sudo ldconfig ``` ## Usage Example flowgraphs are provided in the examples/ directory. Socket PDUs are used as the primary input interface. Socket PDUs configured as UDP Servers may be connected to via: ```$ nc -u localhost 52001``` ## Limitations and TODOs - there are a lot of todos. This repo is still updating for research. ## Acknowledgements - Balint Seeber and the Bastille Threat Research Team - IQ Donors Thomas Telkamp and Chuck Swiger