Regarding UWB positioning, the determination of LOS (Line of Sight) and NLOS (Non-Line of Sight) conditions can affect location accuracy. I have several questions to ask. 1. For LOS signals, when encountering signals with high attenuation, they are mostly reflected NLOS signals, and those with less attenuation have the first path NLOS signals. In these three scenarios, is the first path detected by the UWB’s built-in LED algorithm the same? Is the LED a value that is almost fixed? What are the differences in LED detection in these three situations? 2. I know that the fp_index is roughly between 740-750ns in the CIR. So, in these three scenarios, how is the fp_index determined? It seems similar to the first question, haha. Thank you all for your replies!
As far as I know Qorvo/Decawave don’t publish details of exactly how their leading edge detection works. Based on a combination of what is in the documentation and some reasonable sounding assumptions it is roughly this (note this is pure speculation and guesswork):
Based on the preamble detection the chip makes an estimate of where the start of the frame is going to be. The CIR data buffer is aligned to the incoming signal based on this initial estimate. Since this first guess is more likely to be too late rather than too early due to multipath effect this initial guess is aligned so that it will be around 3/4 of the way along the CIR memory. This is why the LED is normally in the 740 region.
Once the CIR accumulator has been filled the average noise level is calculated and a background noise estimate made. Based on this noise level and a couple register values a threshold level is set, exactly how this calculation is done is unclear. The system then looks for the first peak in the CIR data that is above this threshold. Finally it applies some interpolation like algorithm to the received data to try and refine the receive time estimate to give a more accurate number.
The LoS signal is always before a NLoS signal so as long as it’s over the threshold it will be picked up even if the NLoS is significantly stronger. If however the LoS is too weak to cross the threshold then it may be missed. Playing with the noise level sensitivity numbers to lower the threshold may improve the detection of weak LoS signals but at the risk of false detects or noisier results.
If the NLoS is both significantly stronger and later than the LoS signal then you may get a leading edge index that is earlier than the normal ~740 value. There are also some unusual reflection conditions (generally only in very large areas with good reflectors at the ends) that can result in false edges being detected at times significantly earlier than the true signal.