Questions about the parameters corresponding to the experimental datasets and the shape of Pmusic

[ClarenceZSK]

@egaebel Thank you so much for your code! It helps me understand and implement SpotFi very much. But I have optimized your code to run more efficiently.

I am trying to use your experimental data to verify the correctness of my code. In your code, the antenna distance is set to be 0.1. To my understanding, if the antenna distance is larger than half length of the magnetic wave, it will course ambiguities of AoA detection in a linear antenna array. Thus when I run your original code, a time period of the Pmusic is:

This figure shows three paths in terms of ToFs. Each path appears multiple peaks. This confuses me a lot. Is it due to the 0.1 antenna distance coursing an ambiguity? Or am I missing the correct parameter setting?

Moreover, from your outputs, "True number of computed paths: 6". But the figure only shows three paths. Could you please help me explain it? (I had tuned the tau value to be 0-2000ns. It shows the shape of the figure above periodically. )

[egaebel]

SpotFi relies on multiple linear arrays of antennas, in my work I only used a single linear array which significantly contributes to the ambiguities you're seeing.

There are multiple ToF peaks because the peaks occur cyclically an 2pi intervals. The value for tau that I iterate over in my code is larger than it needs to be because it's better to be too large than too small.

Hope that helps!

[egaebel]

Oh one more thing about the discrepancy between the "True number of computed paths" and what the figure shows. The number of computed paths is obtained from the number of non-zero eigenvalues. This is nice in theory, but in practice they aren't quite 0 because of noise so I very roughly find the biggest drop in eigenvalues and declare this to be where the "zeros" start. There are probably better methods to do this, but that's how I did it at the time.

[ClarenceZSK]

Thank you very much for your prompt reply.

The figure above actually only shows the spectrum within the time from 650ns to 850ns, which exhibits multiple ToF peaks that are not due to the 2pi interval. The spectrum of a larger tau setting that exhibits ToF peaks occurring by the 2pi interval is shown as follows:

So, the three ToF peaks in the first figure probably represent three paths.

"SpotFi relies on multiple linear arrays of antennas", to my understanding, in the system, each device (no matter a sender or a receiver) only equipped a linear antenna array that consists of three antennas. But the system relies on multiple APs. Thus, the multiple APs' measurements will help remove the ambiguities. Is it right?

Actually, when I position the antenna array whose distance interval is less than half of the wavelength (2.61cm), the spectrum only show one peak in the AoA dimension per ToF value. The figure is as follows:

Unfortunately, it only exhibits one ToF peak in a 2pi interval. The signal subspace also only has two eigenvectors using your "biggest drop" method:). When I deliberately add two more eigenvectors into the signal subspace, the spectrum shows more ToF peaks as shown below.

Does it imply that the ambiguities in each ToF value in the first figure are due to the large antenna interval (larger than half of the wavelength)?

[whubaichuan]

Hi， ClarenceZSK。 How is it going now??? Are you succeed?

[whubaichuan]

@ClarenceZSK Hi， ClarenceZSK。 How is it going now??? Are you succeed?

[zwlovefish]

hi @egaebel thank you for your code, when i use your code to run my own code,but it Come over，and has two peek. could you help me ? please