1-@28:00 how would the spherical beamforming vectors mathematically differ from the normal steering vectors used in the far field beamforming ? 2-why does the power decrease if we are at a distance less than the focal point (it is already intuitive why is it as such if we are far ) but why if we are closer to the antenna ? 3-given a focal requirement of x meter for ex then how to design the beamforming vectors ?
1. The channel vectors a or b would then have to be modeled to take the spherical wavefronts into account. In a LOS scenario, you would get a steering vector in the far-field, while you need to consider the exact delays (phase-shifts) in the radiative near-field. You can see (24) and (25) in arxiv.org/pdf/2002.04960v4.pdf 2. It is the beamforming gain that reduces at closer distances. This is because the signals from the different antennas are not adding constructively at those locations. You can think about the phenomena that happens when you focus a camera on a person. Objects that are closer and that are much further away will be blurry. I recommend the following paper: arxiv.org/abs/2110.06661 3. It is not more difficult than estimating the channel and selecting the beamforming vector to maximize the SNR at the desired receiver (focal point). What changes is the shape of the "beam" and that shouldn't use channel estimation schemes that make use of steering vectors (e.g., avoid codebook-based beamforming, while reciprocity-based beamforming works directly).
The basic system model is found in the light red box after 31 min. As you can see it is sum_n a_n b_n exp(-j theta_n). One can express this using vectors or matrices in different ways. We present one such expression in the box, but some people like to write it as a^T D b, where D is a diagonal matrix with the theta phase shifts. The fact that it is diagonal is just a notational thing and has no other meaning than representing the fact that the signal arriving to an element is only reflected from that element, and not leaked into other ones. There it recent work on “beyond diagonal RIS” where the hardware contains a network of interconnections between the elements so signals that arrive to one element can be reflected from other elements. This is an ongoing research area where people try to identify what kind of benefits one might get from using such a significantly more complex hardware setup.
Most of my papers and presentations on IRS are frequency-independent, because the basic principles of wave propagation are the same. Is there anything particular terahertz-related that you think we should cover?
Hello professor. In order to plot the channel at LOS explained, I plotted it as shown in the picture shown in @1:09:40 using the rician channel vector and array response. The question is, why is the beam angle of the LOS path formed at 0 degrees? Could it have something to do with rician channel modeling? Thank you for always great lectures.
Yes, we assumed that the LOS component in the Rician model is the array response vector for the zero-angles (i.e., a vector with 1 1 1…). If you change it to another angle, then the preferred beam angle will move accordingly.
I am working on an IRS design. I have done unit cell simulation but for array simulation I am getting some problems. So don't know whether I am going in right direction or wrong that's why asked you to make a video.
Thank you so much for the tutorial Professor. I have a query regarding the phase shift matrix. What is the reason of taking diagonal of phase shift matrix.
Hi! In this tutorial, we are collecting the phase shifts in a vector that we call \omega. I don’t think we use a diagonal phase shift matrix. But there are other presentations that use a matrix description instead. The key point is that one should write the channel using the summation formula in the light red box on Slide 15. One can then write it using vectors and matrices in different ways, and different authors prefer different representations. But they are all equivalent.
Thank you for the lecture and the additional slides. At the first glance, the voltage values for the meta materials (up to 19 V). Hopefully there won't be any problems with the energy consumption. In slide 18 on the right image, perhaps the letters are wrong (h_2,g_2) - I expected a_2,b_2. In slide 15: the unknown parameter is theta_n, and we need to determine that parameter. Is that right? We want to adjust the phase of the surface. Because in the last equation the sqaure of the norm of the element wise product is independent of the phase theta_n.
Slide 18: Yes, you are right. Slide 15: Yes, the design variable it theta_n. If we select it as stated in the yellow box, you can achieve the upper bound where all the phases are gone. What we are doing more precisely is to make sure that all terms in the sum get zero phase, so we can just add up the amplitudes.
hello ,sir.I want to ask you a question，in the video 1:14:10 The Time-domain end to end channel matrix V. If I want to use matlab to express this matrix, how should I express it.
No, the expression on the slide is correct. You can find the exact derivation in this paper: arxiv.org/abs/2102.00742 I guess you have seen the 1/2 scaling factor in some textbooks, which want to use the same transformation from passband to baseband for the signals and the impulse response. But this messes up the expressions by requiring these extra scaling factors, which are not given by nature but is a consequence of the choice of notation. We use a different notation so that a convolution between impulse responses in the passband becomes a convolution of impulse responses in the baseband, without artificial scaling factors.
The range of radio waves is quite short in water, but can be reflected off surfaces. Hence, I believe one can build an IRS underwater if one can find material with configurable reflection coefficient.
Dear sir did you actually simulated 1100 element array in this work :"RIS-Aided Wireless Communications: Prototyping, Adaptive Beamforming, and Indoor/Outdoor Field Trials" of yours or only simulated a small array and followed the results for the larger one?
We simulated the behavior of a single element (Figure 3). This is the only simulation in the paper, the rest are measurements. The 1100 element array shown in Figure 4 was manufactured and used in the field trials to obtain measurements. There is no extrapolation from a small to a large array.
Yes but not in a particularly efficient manner. The distinguishing factor between an IRS and a traditional relay/repeater is that the IRS doesn’t amplify the signal, only reflects it. Hence, we can only configure the IRS to lower the amplitude not increase it. This feature can be utilized to further control the shape of the beam pattern of the reflected signal, but I don’t think it as useful as changing the phases to control the coherent/destructive interference pattern.
Thanks for the tutorial. Are there any practical tutorial examples beyond mathematical descriptions on how to model the channel directly from measurement?
As far as I know, there is rather little experimental channel modeling. One example that comes to my mind is "Wireless communications with reconfigurable intelligent surface: Path loss modeling and experimental measurement" arxiv.org/abs/1911.05326
Dear sir excellent video. Can you make a video discussing the how to do simulation of IRS designs in the simulators like HFSS or CST? It will be of great help. Thanks a lot for giving such a clear understanding on IRS.
I believe most people in the communications community is modeling a single element and then consider an array of them, while ignoring mutual coupling and other effects that would change the element properties when there are multiple ones. This is what we considered in this video and the full-scale simulations are important future work. There might be effective solutions to simulate an entire RIS array using CST, if you are looking around in journals such as IEEE Transactions on Antennas and Propagation.
Yes, all the analysis is done in the complex baseband. Here is a paper that shows the connection between phase-shifts in the passband and complex baseband. arxiv.org/abs/2102.00742
@@WirelessFuture Thank you, and is there always a minus before the j*theta? The EULER representation is without the minus (r*e^(j*phi)), but there is always the minus.
@@jasminnadic2103 Both representations can be utilized. The reason for having a minus sign in this context is that phase-shifts are related to delays, and a signal s(t) that is delayed by tau seconds becomes s(t-tau). The minus sign appears naturally...
Thank you so much for your effort and time, Sir. I want to know how to apply RIS in LoRa-based Networks to scale interference prevention in dense network , in large scale IoT, for example ?
I haven't looked in to that scenario so I cannot give you a straight answer. An RIS is generally good at creating one good propagation path. It is less good at rejecting interference, even if one can for sure try to reflect signals to avoid interference.
The research is still in the exploratory phase, but there are some works on implementation aspects. In addition to the one mentioned in the presentation, you can have a look at: ieeexplore.ieee.org/iel7/6287639/9668973/09668918.pdf
nice tutorial, I tanks. can I get detail mathematical analysis of IRS assisted MIMO system with hybrid beamforming for NOMA and OFMA systems, tanks for your help and kindness.
Hi! The combination OFDM + IRS is very important, and also covered in the tutorial. You can find a deeper mathematical analysis here: arxiv.org/pdf/2102.00742.pdf You can probably find papers on hybrid beamforming and NOMA as well, but these are two topics that I don’t work on myself. Hybrid beamforming is just a temporary solution: ma-mimo.ellintech.se/2019/05/02/when-will-hybrid-beamforming-disappear/ And NOMA cannot provide much practical gains: ma-mimo.ellintech.se/2021/04/15/is-there-a-future-for-noma/
With respect to slides 27 and 28: the connection between correlation values and the Eigenvalues is about inverse proportional, is that right? One wants many high-valued Eigenvalues, and that corresponds to low correlation numbers on slide 27. I believe to remember that with many high-valued Eigenvalues one can say: the energy is spread widely than with only a few high-valued Eigenvalues.
@@WirelessFuture Thank you for your quick answer. I just have a little unclarity. On slide 13, there is the equation for the End-to-end impulse response. It is in the time domain, clear. On slide 15 this equation for the End-to-end channel, is that in the time- or frequency domain? Thanks.
@@nelebauer7864 Slide 13 provides the continuous-time impulse response. For a narrowband channel, one can write each impulse response in the complex baseband as a delta function times a complex scalar. This is what we present on Slide 15. It is in the time-domain. You can find all the intermediate steps in the following tutorial paper: arxiv.org/pdf/2102.00742
Hi professor! Are you aware if OTFS is considered a good choice for IRS-assisted communication systems in presence of high Dopplers? Or is it not worth doing it using OTFS as there are no papers yet on IRS using OTFS?
Hi! I don’t think think there is a deep understanding of the combination of OTFS and IRS so far. I think the current consensus regarding OTFS is that it provides gains over OFDM in scenarios with very high mobility, but that the gains are not sufficient to motivate a complete redesign of the waveforms. Generally speaking, I don’t think that IRS-assisted communication is a good fit for scenarios with high mobility, because of the resource-intensive configuration procedure. When the desired reflection angle changes rapidly due to mobility, you will have to reconfigure the surface rapidly to point the reflected beam in the right direction, even if the data isn’t encoded into the emitted signal using frequency domain subcarriers (OFDM) but the delay/Doppler domain (OTFS). But now I’m just guessing so this could indeed merit further research!
@@WirelessFuture Thank you professor for the reply. I agree with your points. For now, I feel too that it would be too complex to blend in IRS with OTFS. But what about in an IRS-assisted LEO satcom system where the LEO satellites circle the orbit in speeds in the range of 7-8km/s. Do you think in such scenarios, it might be worth to replace OFDM with OTFS? Btw, just to let you know, I am in my second year of PhD at Monash and I get super inspired by your teaching. Once I feel confident enough in my field, I hope to apply for an internship under you, maybe next year!
@@amitbora6336 The LEO satellite and OTFS could indeed be a scenario worth studying. It is high by predictable mobility which should make it convenient to also predict where the signals will be in the delay/Doppler domain.
![Ep 23. Wireless Localization and Sensing (With Henk Wymeersch) [Wireless Future Podcast]](http://i.ytimg.com/vi/nYxFiNosm0g/mqdefault.jpg)
![Ep 23. Wireless Localization and Sensing (With Henk Wymeersch) [Wireless Future Podcast]](/img/tr.png)
1-@28:00 how would the spherical beamforming vectors mathematically differ from the normal steering vectors used in the far field beamforming ?
2-why does the power decrease if we are at a distance less than the focal point (it is already intuitive why is it as such if we are far ) but why if we are closer to the antenna ?
3-given a focal requirement of x meter for ex then how to design the beamforming vectors ?
1. The channel vectors a or b would then have to be modeled to take the spherical wavefronts into account. In a LOS scenario, you would get a steering vector in the far-field, while you need to consider the exact delays (phase-shifts) in the radiative near-field. You can see (24) and (25) in arxiv.org/pdf/2002.04960v4.pdf
2. It is the beamforming gain that reduces at closer distances. This is because the signals from the different antennas are not adding constructively at those locations. You can think about the phenomena that happens when you focus a camera on a person. Objects that are closer and that are much further away will be blurry. I recommend the following paper: arxiv.org/abs/2110.06661
3. It is not more difficult than estimating the channel and selecting the beamforming vector to maximize the SNR at the desired receiver (focal point). What changes is the shape of the "beam" and that shouldn't use channel estimation schemes that make use of steering vectors (e.g., avoid codebook-based beamforming, while reciprocity-based beamforming works directly).
hi professor good morning, i have one doubt why we choose only diagonal of element reflection
The basic system model is found in the light red box after 31 min. As you can see it is sum_n a_n b_n exp(-j theta_n). One can express this using vectors or matrices in different ways. We present one such expression in the box, but some people like to write it as a^T D b, where D is a diagonal matrix with the theta phase shifts. The fact that it is diagonal is just a notational thing and has no other meaning than representing the fact that the signal arriving to an element is only reflected from that element, and not leaked into other ones.
There it recent work on “beyond diagonal RIS” where the hardware contains a network of interconnections between the elements so signals that arrive to one element can be reflected from other elements. This is an ongoing research area where people try to identify what kind of benefits one might get from using such a significantly more complex hardware setup.
Could you make a video on use of terahertz freq in IRS for 6G
Most of my papers and presentations on IRS are frequency-independent, because the basic principles of wave propagation are the same. Is there anything particular terahertz-related that you think we should cover?
Hello professor. In order to plot the channel at LOS explained, I plotted it as shown in the picture shown in @1:09:40 using the rician channel vector and array response. The question is, why is the beam angle of the LOS path formed at 0 degrees? Could it have something to do with rician channel modeling? Thank you for always great lectures.
Yes, we assumed that the LOS component in the Rician model is the array response vector for the zero-angles (i.e., a vector with 1 1 1…). If you change it to another angle, then the preferred beam angle will move accordingly.
@@WirelessFuture Thank you! professor!
I am working on an IRS design. I have done unit cell simulation but for array simulation I am getting some problems. So don't know whether I am going in right direction or wrong that's why asked you to make a video.
Dear Sir, Can you please guide how can we implement RIS in MATLAB simulation? is this possible to analyze its effect in 6G ?
Your email address...?
Could you please help with MATLAB Code
Thank you so much for the tutorial Professor. I have a query regarding the phase shift matrix. What is the reason of taking diagonal of phase shift matrix.
Hi! In this tutorial, we are collecting the phase shifts in a vector that we call \omega. I don’t think we use a diagonal phase shift matrix. But there are other presentations that use a matrix description instead. The key point is that one should write the channel using the summation formula in the light red box on Slide 15. One can then write it using vectors and matrices in different ways, and different authors prefer different representations. But they are all equivalent.
Thank you for the lecture and the additional slides. At the first glance, the voltage values for the meta materials (up to 19 V). Hopefully there won't be any problems with the energy consumption.
In slide 18 on the right image, perhaps the letters are wrong (h_2,g_2) - I expected a_2,b_2.
In slide 15: the unknown parameter is theta_n, and we need to determine that parameter. Is that right? We want to adjust the phase of the surface. Because in the last equation the sqaure of the norm of the element wise product is independent of the phase theta_n.
Slide 18: Yes, you are right.
Slide 15: Yes, the design variable it theta_n. If we select it as stated in the yellow box, you can achieve the upper bound where all the phases are gone. What we are doing more precisely is to make sure that all terms in the sum get zero phase, so we can just add up the amplitudes.
hello ,sir.I want to ask you a question，in the video 1:14:10 The Time-domain end to end channel matrix V. If I want to use matlab to express this matrix, how should I express it.
Thanks a lot for the great presentation.
On slide 12, shouldn't we use h(t) = 0.5*h_{pb}(t) e^{-j2\pi f_c t} for the baseband model?
No, the expression on the slide is correct. You can find the exact derivation in this paper: arxiv.org/abs/2102.00742
I guess you have seen the 1/2 scaling factor in some textbooks, which want to use the same transformation from passband to baseband for the signals and the impulse response. But this messes up the expressions by requiring these extra scaling factors, which are not given by nature but is a consequence of the choice of notation. We use a different notation so that a convolution between impulse responses in the passband becomes a convolution of impulse responses in the baseband, without artificial scaling factors.
hello professor! Amazing video but I was curious to know whether it is feasible to use these IRS elements underwater? What is your take on that?
The range of radio waves is quite short in water, but can be reflected off surfaces. Hence, I believe one can build an IRS underwater if one can find material with configurable reflection coefficient.
@@WirelessFuture Thank you for your reply prof!
Dear sir did you actually simulated 1100 element array in this work :"RIS-Aided Wireless Communications: Prototyping, Adaptive Beamforming, and Indoor/Outdoor Field Trials" of yours or only simulated a small array and followed the results for the larger one?
We simulated the behavior of a single element (Figure 3). This is the only simulation in the paper, the rest are measurements. The 1100 element array shown in Figure 4 was manufactured and used in the field trials to obtain measurements. There is no extrapolation from a small to a large array.
@@WirelessFuture thanks a lot sir
Hello.. i have one question ... can IRS change amplitude of reflected signal as well ??
Yes but not in a particularly efficient manner. The distinguishing factor between an IRS and a traditional relay/repeater is that the IRS doesn’t amplify the signal, only reflects it. Hence, we can only configure the IRS to lower the amplitude not increase it. This feature can be utilized to further control the shape of the beam pattern of the reflected signal, but I don’t think it as useful as changing the phases to control the coherent/destructive interference pattern.
Thanks for the tutorial. Are there any practical tutorial examples beyond mathematical descriptions on how to model the channel directly from measurement?
As far as I know, there is rather little experimental channel modeling. One example that comes to my mind is "Wireless communications with reconfigurable intelligent surface: Path loss modeling and experimental measurement" arxiv.org/abs/1911.05326
@@WirelessFuture many thanks for your reply
Dear sir excellent video. Can you make a video discussing the how to do simulation of IRS designs in the simulators like HFSS or CST? It will be of great help. Thanks a lot for giving such a clear understanding on IRS.
I believe most people in the communications community is modeling a single element and then consider an array of them, while ignoring mutual coupling and other effects that would change the element properties when there are multiple ones. This is what we considered in this video and the full-scale simulations are important future work. There might be effective solutions to simulate an entire RIS array using CST, if you are looking around in journals such as IEEE Transactions on Antennas and Propagation.
@@WirelessFuture go it sir. Thanks a lot
What is the future work in this technology (RIS)
Thank you. In slide 15 in omega_theta, this phase-shift is always represented with this complex representation, yes?
Yes, all the analysis is done in the complex baseband. Here is a paper that shows the connection between phase-shifts in the passband and complex baseband. arxiv.org/abs/2102.00742
@@WirelessFuture Thank you, and is there always a minus before the j*theta? The EULER representation is without the minus (r*e^(j*phi)), but there is always the minus.
@@jasminnadic2103 Both representations can be utilized. The reason for having a minus sign in this context is that phase-shifts are related to delays, and a signal s(t) that is delayed by tau seconds becomes s(t-tau). The minus sign appears naturally...
Thank you so much for your effort and time, Sir. I want to know how to apply RIS in LoRa-based Networks to scale interference prevention in dense network , in large scale IoT, for example ?
I haven't looked in to that scenario so I cannot give you a straight answer. An RIS is generally good at creating one good propagation path. It is less good at rejecting interference, even if one can for sure try to reflect signals to avoid interference.
@@WirelessFuture , thank you so much, Sir. I really appreciate your time.
Thank you so much for a nice tutorial. I have a question: how can we implement this IRS including the controller in real-life products?
The research is still in the exploratory phase, but there are some works on implementation aspects. In addition to the one mentioned in the presentation, you can have a look at: ieeexplore.ieee.org/iel7/6287639/9668973/09668918.pdf
@@WirelessFuture This is really fantastic! Thank you! I am gonna reach out to you Prof Emil to discuss a possible opportunity :)
nice tutorial, I tanks. can I get detail mathematical analysis of IRS assisted MIMO system with hybrid beamforming for NOMA and OFMA systems, tanks for your help and kindness.
Hi! The combination OFDM + IRS is very important, and also covered in the tutorial. You can find a deeper mathematical analysis here: arxiv.org/pdf/2102.00742.pdf
You can probably find papers on hybrid beamforming and NOMA as well, but these are two topics that I don’t work on myself.
Hybrid beamforming is just a temporary solution: ma-mimo.ellintech.se/2019/05/02/when-will-hybrid-beamforming-disappear/
And NOMA cannot provide much practical gains: ma-mimo.ellintech.se/2021/04/15/is-there-a-future-for-noma/
@@WirelessFuture thanks, proff. Go on...
With respect to slides 27 and 28: the connection between correlation values and the Eigenvalues is about inverse proportional, is that right?
One wants many high-valued Eigenvalues, and that corresponds to low correlation numbers on slide 27.
I believe to remember that with many high-valued Eigenvalues one can say: the energy is spread widely than with only a few high-valued Eigenvalues.
Yes. The flat dashed line on slide 28 represents no correlation. The fewer high-valued eigenvalues there are, the more correlation there will be.
@@WirelessFuture Thank you for your quick answer. I just have a little unclarity. On slide 13, there is the equation for the End-to-end impulse response. It is in the time domain, clear. On slide 15 this equation for the End-to-end channel, is that in the time- or frequency domain? Thanks.
@@nelebauer7864 Slide 13 provides the continuous-time impulse response. For a narrowband channel, one can write each impulse response in the complex baseband as a delta function times a complex scalar. This is what we present on Slide 15. It is in the time-domain. You can find all the intermediate steps in the following tutorial paper: arxiv.org/pdf/2102.00742
Hi professor! Are you aware if OTFS is considered a good choice for IRS-assisted communication systems in presence of high Dopplers? Or is it not worth doing it using OTFS as there are no papers yet on IRS using OTFS?
Hi! I don’t think think there is a deep understanding of the combination of OTFS and IRS so far. I think the current consensus regarding OTFS is that it provides gains over OFDM in scenarios with very high mobility, but that the gains are not sufficient to motivate a complete redesign of the waveforms. Generally speaking, I don’t think that IRS-assisted communication is a good fit for scenarios with high mobility, because of the resource-intensive configuration procedure. When the desired reflection angle changes rapidly due to mobility, you will have to reconfigure the surface rapidly to point the reflected beam in the right direction, even if the data isn’t encoded into the emitted signal using frequency domain subcarriers (OFDM) but the delay/Doppler domain (OTFS). But now I’m just guessing so this could indeed merit further research!
@@WirelessFuture Thank you professor for the reply. I agree with your points. For now, I feel too that it would be too complex to blend in IRS with OTFS. But what about in an IRS-assisted LEO satcom system where the LEO satellites circle the orbit in speeds in the range of 7-8km/s. Do you think in such scenarios, it might be worth to replace OFDM with OTFS? Btw, just to let you know, I am in my second year of PhD at Monash and I get super inspired by your teaching. Once I feel confident enough in my field, I hope to apply for an internship under you, maybe next year!
@@amitbora6336 The LEO satellite and OTFS could indeed be a scenario worth studying. It is high by predictable mobility which should make it convenient to also predict where the signals will be in the delay/Doppler domain.
💛👍🏾