There is a new version of this course at kzbin.info/aero/PLHGIkY491Cy2LPstPnzMpXj9QEBG5qkJU&si=DLllP1rQnIOTCDRB

@@ProfHenkWymeersch Sincerely thank you for your kindness, I hope you have a great day!!

Yes, you are right. Good catch!

Thank you. You are not the first person to mention this. I will upload a new version with reduced music volume and delete this version.

Hi. I assume you are referring to the matlab code at minute 11. Line 11 generates a normal (Gaussian) random variable with variance sigma^2_psi and then converts it to the log-domain. To convert a variable X to the log-domain, you apply 10*log10(X). Line 14 generates N complex Gaussian channel coefficients. Each channel coefficient is the sum of a real Gaussian random variable with variance 1 and an independent imaginary Gaussian random variable with variance 1. After addition, the variance is 2 (1+1). To get the correct variance Pr, the entire object is multiplied with sort(Pr/2).

Yes. These are the outage probabilities. For instance for M=2, you find that P_out = (1-exp(-5/31.5))^2 = 0.02. Same for the other values of M.

I get it. I asked this question because it writes 0.15,0.01,0.003 on the slide.😂

Hi. Good point. To get the sufficient statistic, it is not needed that the envelope is constant. So the point of the slide is that you can get a compressed observation (namely \beta^H y) based on which you can make an optimal decision. To actually make that optimal decision, you will of course need to compute a maximum likelihood objective function, which has to include the term ||\beta||^2 |s|^2.

I get it. Thank you!

The link is now fixed. Thanks for informing me!

@@ProfHenkWymeersch Hi Professor, the matlab code link seems still pointing to PDF file, could you please fix it? Thanks!

@@skyhighhigh9999 Ah. I only changed it in the playlist, not for this lecture. Please try again. Should be fixed.

@@ProfHenkWymeersch Thank you very much Professor Wymeersch. Your lectures are very enjoyable to watch and learn! Thanks much for putting this up for the community!

No, it considers both the effect of the pathless as well as the shadowing. First you apply the path loss to compute the average power (in dBm) at the specific distance. Then add the normal shadowing (still in the dB domain, log-normal in the linear domain). Combined this gives you one realization of the average power of the large-scale fading at a certain location. The small-scale fading is then generated with a power based on the realization of the large-scale fading.

The thing I'm confusing about is: The path loss is a constant number for a certain distance, so the variance of path loss is always zero. Then the variance is only related to the shadowing, which is a random variable. This is my idea, and I'm not sure where there might be any inaccuracies. Thank you for taking the time to help clarify my doubts.

@@user-ly4dp9pd9c Yes shadowing is random, but after you generate a realization, it would be fixed for a certain small region (on the order of 10s of wavelengths). Within that region, under a fixed path loss and fixed shadowing, the only randomness is due to the small-scale fading. The power of the small scale fading is a function of the deterministic path loss and the specific shadowing realization.

I get it! Thanks for your patience.

Of course. You can download them here: www.dropbox.com/scl/fi/ixdwzee7kecixbw3fmabj/lectureCombined.pdf?rlkey=sy92xeedaqfmzf27w5hv503xt&dl=0

@@ProfHenkWymeersch Thank you very much! Professor.

If you send me an email, I am happy to share the slides.

In this video, we focused on snapshot positioning, but in the papers, you can see we also studied tracking.

Yes definitely! With enough multipath or several base stations, you get estimate the 3D orientation.

Of course. You can find them here: shorturl.at/csHO6

@@ProfHenkWymeersch Thank you professor!

Thanks for your question. You will then need to send uplink pilot signals on all of the user antennas. If the signals are sufficiently long, the BS can create an observation model and obtain a least squares channel estimate of all the channels of all the users.