Physically Large Antenna Arrays: When the Near-Field Becomes Far-Reaching

[@Soflotony]

Excellent presentation ! clear and concise ! Thank you.

[@kevenzhong871]

It is really amazing! According to the conclusion, after the time, frequency, code and spatial angular, this four dimension, a totally new dimension is born! If possible, we can call it the "near field focus dimension"!

[@JWang-co2vj]

Stepping out of the bubble of telecommunication to see what happens in broader sensing fields, for example, radar (microwave imaging), radio-astronomy/radio telescope, optics for microscope and EM field, it can be found that the concept has been extensively investigated for decades. By the way, the spot area after beamforming in the near field of the array is called point spread function in microwave imaging, optical imaging for microscope, etc.

[@Theepakshoundrabalan]

Mind Blowing work.

[@Julia-hu4xe]

I learned that in the nearfield the energy is somehow swashing back and forth. So to my mind, the power at a point is time dependent. But in your explanation that does not to be of interest, because we point at a special local point all the time. Is that a problem in near field beamforming? Thanks a lot.

[@WirelessFuture]

The electromagnetic waves are time-varying and oscillate according to their frequency content, so that phenomenon is there both in the near-field and far-field. If the transmission has a stationary behavior then the waves will reach a steady-state behavior when it comes to the oscillations. I think what you are asking about is connected to the reactive near-field, where some of the field components are not propagating away from the transmitter but moves around near the transmitter. This is the situation that many people associate with the term “near-field”. However, this presentation and the recent research on large antenna arrays consider the _radiative_ near-field, which lies in between the reactive near-field and far-field. With physically large array, we radiative near-field might cover all the distance relevant for wireless communications (e.g., from 20 to 2000 m). All the field components are propagating away from the transmitter but has not yet taken the form of planar waves.

[@sunshinecao]

Thank you. It's an interesting lecture. I have a question that in near-filed and far field regions(communication perspective) slide, around 13min, about the Amplitude difference, how to get the condition d>=1.2D, the amplitude difference is less than cos(pi/8)?

[@WirelessFuture]

The value was selected a long time ago and somewhat arbitrarily. You compute the amplitude difference between the wave reaching the center and the edge of the array. At the distance d=1.2D, it becomes 1.2/sqrt(1.2^2+1/4) ≈ cos(pi/8). We explain this on page 2 in the following paper: arxiv.org/pdf/2110.06661.pdf

[@sunshinecao]

@@WirelessFuture Thank you very much for the reply.

[@lucidasser7153]

Just a short question to slide 13. What ist the Far field channel gain? This depends on the distance. Does it refers to the gain at d_FA? I conclude that the gain in the far field is always greater than a near field gain - is that correct? Because you use it as the normalization value. Thanks.

[@WirelessFuture]

The far-field channel gain is stated on slide 11. It is the conventional Friis propagation formula in this line-of-sight scenario. Yes, the near-field gain is smaller due to the three phenomena mentioned on slide 12. The gain that we consider here is not a specific number but formula.

[@lucidasser7153]

@@WirelessFuture thanks for the quick answer, and setting the desired focal distance means that we set the phases and amplitudes from the elements so that we get coherent interference at this point? In far field beamforming just the angle matters, now we must give the system angle plus distance.

[@WirelessFuture]

@@lucidasser7153 Correct

[@Kabodtubeofficial]

Hello professor. Can you please tell me what Weighted sum rate is in wireless communication. for eg in IRS aided massive MiMO communication there are papers thet try to maximize the weighted sum rate.

[@WirelessFuture]

The data rate of a user is measured in bit/s (or sometimes bit/s/Hz). The sum rate is the summation of the data rates of all the users in the system. When computing the weighted sum rate, we multiply the rates of some users with weights before adding them together. Maximization of a weighted sum will prioritize the users with large weights. I recommend you to read: hal.archives-ouvertes.fr/hal-01098893/document

[@Kabodtubeofficial]

@@WirelessFuture thank you

[@James_Knott]

With multiple beams, can the power level in each beam be independent of the other beams? That is, can the beam for a nearby user provide less power than one for a user that's further away?

[@WirelessFuture]

Yes, the power levels can be controlled independently. For each signal, you assign a "beam" that determines the spatial pattern the signal will get and then you assign a transmit power to it, which determines the amplification of the beam pattern. You do that for all the different beams and then superimpose the signals (i.e., transmit them simultaneously). The only connection between the beams is that the total transmit power is limited in the base station, so if we assign more power to one beam, we need to reduce the power to other beams.

[@stellatauer761]

Thanks. We get this gain curve where the gain increases at about 10 times the FRAUNHOFER distance. From that point closer to the transmitter, we have very low gain. The reason for that low gain are these 3 near-field phenomenons, is that right?

[@WirelessFuture]

I'm not sure which gain curve you are referring to. If you focus a beam at the far-field, the gain will be low when you are in the right direction but closer than the Fraunhofer distance/10. One can compensate for that by modifying the beamforming, taking the spherical wavefronts into account. There is then an interval in the radiative near-field where you can get the full gain. However, when the propagation distance becomes comparable to the Björnson distance, then the gain will always be reduced. This is when the three near-field phenomena are important. Figure 1.15 in the following book chapter shows this: arxiv.org/pdf/2209.03082.pdf

[@이재현-t6i]

The professor misunderstood what the near-field is. The difference between near-field and far-field comes from the distance-dependency of the induced electric field radiating from the antenna. For example, the near-field E-field is inversely proportional to the distance-square or distance-cubic, whereas the far-field E-field is inversely proportional to the distance. If following his idea, the massive MIMO array antenna which has 16x16 or 32x32 antenna elements (commercially utilized for 5G BS) always suffers from near-field effect, which is not true. The array distance is already considered when we use MIMO, the far-field assumption, which assumes that the radiated E-field is inversely proportional to the distance between the observation point to the antenna element, as explained above.

[@WirelessFuture]

The point that is made in the video is that there isn’t a single near-field definition that is applicable in every context; the most suitable definition depends on whether one studies the radiated signal from an antenna or from an array (or the counterparts in terms of reception, which are equally important in communications but isn’t considered in EM literature) and whether one cares about the amplitude, phase, or beam shape of the radiated signal. First of all, the near-field can be divided into the reactive and radiative parts. In the reactive near-field, the E-field is inversely proportional to the distance-square or distance-cubic, as you are pointing out. In the radiative near-field (Fresnel region), the E-field is inversely proportional to the distance but one cannot use the plane-wave approximations that are applicable in the far-field. Moreover, when one considers an array of many antennas, the far-field distance of a single element is much different from the far-field distance of the entire array. This is further enhancing the effect that E-field of every element is inversely proportional to the distance, but yet one cannot apply the far-field assumption when studying the jointly radiated waveform in communications (in terms of physical beam-shape, array response vectors, and SNR computation). The radiative near-field region of conventional arrays is so short that it can be neglected but this won’t be the case when using physically large arrays and short wavelengths. This is where finite-depth beamforming becomes possible. The following paper discusses these things in further detail: arxiv.org/abs/2110.06661

[@pitmaler4439]

As always it is very interesting. By the way: In slide 16 you spoke about the focusing on different distances. Later you also mentioned the multiplexing in the near-field. Perhaps you can address anytime how it is realized. I assume that one work with the normal digital or analog beamforming. But it is somehow difficult to imagine, how to multiplex targeted in the near-field. Thank You.

[@WirelessFuture]

Hi! It isn't hard to implement it in practice using "normal" digital beamforming. One estimates the channel matrix, notice that it has a higher rank than it would have had in the far-field, and then apply standard SVD precoding + waterfilling power allocation (as Telstar described in the seminal paper "Capacity of multi-antenna gaussian channels"). It is only the modeling of the channel matrix that changes. How will the transmissions look like, geometrically? With two uniform linear arrays, you will basically transmit one beam towards the middle of the receiver (strongest singular value), then transmit beams to the different sides of the array - as sketched at the bottom of Slide 21. With uniform circular arrays, one can sometimes create "beams" with different angular orbital momentum (OAM).

[@pitmaler4439]

I just read something that in the near field E and H decrease way faster than in the farfield (≈ 1/r^3 or 1/r^2 or so). Can you really give such a curve term for the near field? Because like you said, you must focus the beam at a point and the gain is not Independent from the distance. That seams to be a bit odd to me. Thank you for all the videos.

[@WirelessFuture]

The ”extra” terms that you refer to are the ones mentioned in the red box at @7:00. These effects appear in the reactive near field. This presentation is mainly about the radiative near field, which exists in between the reactive near-field and the far-field. It is normally a negligibly small region but it can be very large when using large arrays and carrier frequencies. The graphs that shown this video are only considering the focusing effect (beamforming gain, array gain or whatever we want to call it). On top of that, you will also get a distance dependent pathloss similar to that in the conventional farfield.

[@jasminnadic2103]

Do I understand right. Within the radiative near field, the gain, and thereby received power is dependent on the distance from the transmitting array? In other words, in this region a power-plot in a plane is a hilly environement. You have mountains and valleys which depend on the angle and the distance. Thank you.

[@WirelessFuture]

Yes, that is correct.

[@marcinrybakowski8926]

Thank you professor for great explanation of this finite depth near field beam-forming. When we look to EMF exposure aspects for this type of extremely large arrays we can say that the persons who are in the outside areas of this defined focus spot will have much lower exposure to EM field? This means that persons could be very close to this very large arrays because it does not mean that the antenna with this huge gain potential will radiate huge EM field toward these persons?

[@WirelessFuture]

Yes, what you are saying makes good sense. The maximum array gain is proportional to the array area and is huge for large arrays, which means that the signal is strong at the focus spot and then quickly reduces when moving away from that spot. Since a larger fraction of the transmitted power is at the focus spot there must be less elsewhere since the total power is fixed. As far as I known, the research community has not analyzed the EM exposure in detail in these scenarios.