In this paper we propose an original implementation of a LoRa transceiver using software defined radio. Our goal is to optimize the generation of the waveform on the transmitter side and to propose an efficient implementation of the demodulation algorithms at the receiver. The proposed implementations have been developed to have the smallest memory footprint, both in the software and hardware parts, while ensuring performance levels (latency and throughput) in line with IoT scenarios. In addition, we show that a 1-bit resolution quantization of the complex envelope to be transmitted associated to our receiver introduces only 0.7dB of sensitivity loss compared to the floating point version.