This is really amazing work. I think your implementation is the first to have all channels all spreading factors listening in with bearable CPU load. Are you able to comment on the SNR capabilities of this demodulator in comparison with the LoRa standard? Can it demodulate an SF12 frame with minus 19 SNR?
DSP is based on this paper: arxiv.org/pdf/2002.08208 AWGN performance is good, but I saw bad performance for longer frames with real nodes. I think, it lacks some tracking of the FFT alignment, which might shift due to sample clock offset. I'm not sure, if there's an implementation that does this.
This is really amazing work. I think your implementation is the first to have all channels all spreading factors listening in with bearable CPU load. Are you able to comment on the SNR capabilities of this demodulator in comparison with the LoRa standard? Can it demodulate an SF12 frame with minus 19 SNR?
DSP is based on this paper: arxiv.org/pdf/2002.08208 AWGN performance is good, but I saw bad performance for longer frames with real nodes. I think, it lacks some tracking of the FFT alignment, which might shift due to sample clock offset. I'm not sure, if there's an implementation that does this.