[color=#333333]Hi all guys![/color]
[color=#333333]In NLOS scenario (so with important ranging errors), when I extract CIR, the reported first path is the “real” first path (the one that I would have in LOS case) or the one that is considered wrongly as first path by the chip and that leads to ranging error? You can find attached NLOS/LOS CIR with red line identifying FirstPath index and green line [color=#333333]identifying PeakPath index.[/color][/color]
[color=#333333]If the second option is true (it seems the most logical one), why the wrong detected first path is not also peak path? Theorically, wrong first path should be the path which has suffered less enviromental reflections (cause it arrived before others) and so the one with the biggest amplitude.[/color]
[color=#333333][color=#333333]Moreover, why does leading edge algorithm always seem to output a value of ~740-750ns (index), no matter what is the situation (LOS or NLOS) between the transmitter and receiver? [/color][/color][color=#333333]Theoric[/color][color=#333333]ally, in NLOS scenario RX timestamp should be bigger cause it will lead to a bigger ToF and so to a bigger extimated range.[/color]
[color=#333333]I’m building a classifier based on same parameters retrieved from CIR features, like that in NLOS cases FirstPath index is far away from PeakPath index and that FirstPath amplitude is much lower that PeakPath one. I have reached very good accuracy, but I can’t explain why first path behaviour is like that.[/color]
[color=#333333][color=#333333]I really can’t get an explanation.[/color] [/color]
[color=#333333]Thanks,[/color]
[color=#333333]Alessio[/color]
The reported first path is always the one that the chip thinks is the LOS first path. It reports the point when the signal crosses the threshold as being the time of the first path signal. Whether the chip gets this wrong or not doesn’t matter to the chip, it doesn’t know if it is LOS or NLOS, it doesn’t know if it is right or wrong. All it knows is that the signal crossed the threshold at that time.
Since the peak must be higher than the threshold (or it wouldn’t be counted as a packet reception) the peak will always be after the first path, how much after will depend on the signal and environment.
The index will normally be around and index of 740, that is to do with the way that the chip lines things up in it’s accumulators. It starts off by making a guess as to where the first path will be based on a non-accumulated signal peak. Since it is unlikely that the true first path will be significantly after this first guess but quite possible that it will be earlier it aligns the accumulator such that the initial guess towards is about 3/4 of the way through the accumulator structure.
Hi AndyA, thank you for the answer!
-
I can’t really get the physics that leads to the different behaviours between LOS/NLOS in terms of paths amplitude and paths location. Why in LOS peak path is approximatly also the first path (that makes sense) and in NLOS peak path is far away from first path and also with a much bigger amplitude?
The only explanation I gave myself is that, in NLOS, first path is always the direct path after being attenuated and slowed down by the obstacle while peak path is the reflected path that comes later cause travels more in the air (I can’t get why delay introduced by air is always much more respect to the one introduced by obstacle…) and with a bigger amplitude cause it is not attenuated but only reflected (even if reflections should also attenuate the signal…).
I have always thought that in NLOS case direct path was under noise treshold. But, if it is so and NLOS first path is a reflected path, I can’t get why it is always different from peak path, that is simply another reflected path. -
Are you referring to the fact that, to accurately determine RX timestamp, the DW1000 employs an internal algorithm (LDE on the CIR) to adjust the RMARKER receive time (what you named as “first guess”)? If it is so, can I conclude that CIR First path index will be always approximatly the same both in LOS and NLOS cause it is not influenced by what happened to the signal before being received?
Thanks,
Alessio
First off I should point out that my answers are my understanding of how the internals work based on experience and some (hopefully) logical deductions rather than any inside knowledge. I think I’m getting this right but if anyone from decawave wants to confirm it that would be good.
In a LOS situation the direct path will always be first, in an open field it will also always be the strongest signal and so first path index and peak path index will be separated by only 1 or 2 counts, the peak will be in the middle of the pulse while the first path will be the start of it.
However even with LOS the peak could be significantly later and stronger than the direct signal. Imagine a situation where the transmitter was in the middle of a parabolic dish. The reflected signal will be delayed by twice the dish radius but could easily be 10-20 dB stronger than the direct signal. That’s an extreme example but in a building with lots of metal you could easily get multiple reflection paths that combine to give you a stronger signal.
in NLOS the direct signal will be attenuated and delayed by the obstruction. How much it is delayed and how much it is attenuated will depend on the obstruction. However generally any obstruction that is thin enough to not completely attenuate the direct path will only add enough delay to add ~10 cm or less to the path length, often a lot less. Very few reflections will have a path length that is less than 10 cm longer than the direct path. This means that in a NLOS situation the direct path will still be the first signal you see if you can see it at all, whether the direct path is detected will depend on the obstruction and threshold used. If the first path is completely blocked the shortest reflection of significant strength (which could be the sum of multiple reflections of similar path lengths) will be reported as the first path.
As with the LOS situation the strengths of the reflections will depend a lot on the environment.
Generally CIR first path index will always be around 740 except when you have a heavily attenuated direct signal or very strong reflections.
If you look at my past posts you will see a CIR plot of a situation I had where the first path had an index of ~400. This was actually an error, the first path was around 740 (I had LOS). A strong reflection of the previous pulse was getting interpreted as a highly attenuated first path. 600 sq m metal walls at each end of the building will give you lots of interesting reflections. But a similar thing would be seen if you had just the right amount of signal blockage.