Thank you Associate professor Emil Björnson for this good presentation about Massive MIMO. You may help us to more about Indoor 3D Massive MIMO deployment and channel modeling.
I don't think that anything particular needs to change when it comes to the signal processing. But yes, the channel modeling will be different and I'm not working on that topic. There are 3GPP models for many kinds of propagation scenarios. I recommend the Quadiga implementation: quadriga-channel-model.de
Thanks for sharing 1- may I know @10:51 how did you calculate the 16 dbi if each element of the 8 is 7 dbi ? 2- what kind of beamforming is done here is it analog beamforming where we merely control the Phase ? 3-@11:40 why can't we beamform in the vertical domain if we have access to the radiating elements per column ? 4-why do we need a pilot sequence of length tau if we can send a single symbol and let the base station measure the channel ? Thanks
1. 10log10(8) = 9 dB is the beamforming gain. Add that to 7 dBi to get 16 dBi. 2. It is a fixed beamforming where all antennas send the same signal. It is like analog beamforming with no phase shift. 3. All the elements in a column must send the same signal. Hence the vertical beam shape is fixed. 4. We can use tau=1 in the case that you describe.
@@WirelessFuture Thanks but for 3-a why in this case horizontal beamforming was achieved 3-b can we apply different weights across columns to achieve vertical beamforming, said in another way what does it take to achieve vertical beamforming ? 4-what are the cases where we need tau >1 5-when we talk about up to eigh beams this means simultaneously or at a time we can have just one beam at a time ? Can I know how in both cases ? If your answer was all simultaneously or one at a time 6-In many papers I see they assume tau k but I don't get it any clue why is that ? 7-when we say we have dual polarized antenna do we have single beam per polarization ? I appreciate you recommended books or reference papers used for teaching those stuff may be used at Linköping or KTH Univs Thanks
3a: Each column is connected to a different RF chain, thus we can vary phase shifts in the horizontal direction and thereby perform horizontal beamforming. 3b: The individual elements in column must be connected to separate RF chains (or at least phase-shifters). This is what is done in Massive MIMO. 4: If we serve K users, then we want tau=K. 5: Simultaneously. 6: We want tau>=K to be able to separate the users during channel estimation. You can think of it like having a seminar where the audience wants to know what K persons have to say. The K persons take turns in talking and therefore you need to divide the lecture into tau=K parts. When one person talks, all the receivers (people in the audience) can listen simultaneously. 7: One can transmit one or multiple beams. I recommend "Fundamentals of Massive MIMO" and "Massive MIMO networks" (massivemimobook.com)
@@WirelessFuture For 3-a and 3-b to better imagine the answer may I know what is the math behind them , I mean what equations govern those answers you provided ? Or what title should l look for 8-do we send independent data stream per polarization ? Thanks
Really nice lecture, enjoyed it a lot. Why can’t the industry take a different approach and devise what an ultimate use case for cellular communication might look like? And derive needed technologies and resources from that?
The ultimate use case would probably be so seldomly used that it isn’t economically feasible to deploy networks that support it. Most of the revenue that it is generated by wirelessly connected devices go to the device manufacturers and software/content providers (e.g., Google, Netflix), not to the telecom operators. That is why I think it is better to design future networks that can improve basic metric such as total capacity, uniformity of data rates, latency, reliability, and energy efficiency without making the infrastructure extremely expensive to deploy. Then the hardware/software companies will build products that make use of it as well as they can.
@@WirelessFuture my guess on what the ultimate case for the cellular communication might be is smthg that fully resembles and recreates in-person face-to-face interaction, capturing and re-creating it verbally and visually. IMHO, a face-to-face meeting is in ultimate act of communication the ICT whole industry is trying to re-create. No wonder why VR/glass free 3D content and tech are emerging. Yet, one can see how big the gap is between current capabilities and the final goal we strive to achieve.
Can't you use a "step up regulator" on the input to your antenna to increase range according to the size of the regulator? Not an electric voltage regulator. But a magnetic regulator. That's basically the same concept to increase the range?
Am I correct in assuming that beamforming is happening within ms to serve different users in different directions? Or does beamforming effectively act as serving static directions and moves over minutes? The amount of processing power to do active beamforming seems to be quite significant. Probably a stupid question but I'm trying to understand 5G physical layer / RU architecture and the benefits of functional splits primarily 7.1 and 7.2.
You are right that the beamforming happens at the ms level, at the frame level. The computational complexity might seem high, but it is actually just basic matrix/vector operations so it isn’t a big deal for a modern DSP or ASIC. The main benefit of functional splits and C-RAN is to gather many computational tasks at one point, and thereby save on processing resources. The logic is that each base station has a load-dependent maximum processing need, and it is unlikely to be needed simultaneously on many neighboring base stations. If they pool their resources, then one can save on resources and also enable more flexibility in the algorithms.
@@WirelessFuture Thanks Emil! That makes sense - so Ethernet packets are "aimed" at the user that it is intended for and will be "listening" with the same antenna formation for receive? I get that - it makes sense, as does pooling resources to reduce costs. I'm at the early stages of learning about 5G technology but I've been interested in radio comms since I was very young. Again, thank you for your time in replying to me.
These things are explained in a previous video: ruclips.net/video/m9wEAucKoWo/видео.html In short: 1. Massive MIMO enables you to serve multiple users at the same time and thereby increase the spectral efficiency. 2. It is very costly to learn the channel properly in FDD mode, since the resources spent on channel estimation must grow proportional to number of antennas.
Massive MIMO has two key properties. 1) Channel hardening makes all subcarriers equally good for a user, 2) Favorable propagation makes it possible to serve many users at the same time, and yet protect them from interference using adaptive beamforming. The combination of these properties enable us to serve all users at the same time on all subcarriers, using predictable data rates. There is no need for intricate scheduling algorithms or quick changes of the modulation and coding, two things that usually require substantial control signaling.
Thank you. Great explanation. If the number of antennas is increased, the mutual coupling between them is also increased. Does it affect when switching between many receivers?
I don’t think that mutual coupling will become worse in Massive MIMO. The arrays become larger but the inter-antenna distances remain the same. Mutual coupling is mainly a problem that affect an antenna and its neighboring antennas, so it should not be affected by increasing the array size.
@@WirelessFuture Thank you, dear professor. Therefore, its major effect changes the pattern shape or scattering parameters of the antennas, and it does not depend on the size of the array.
Hi Emil, one question. Throughout the video you kinda considered that the orthogonality principle holds (i.e., the beams are orthogonal). Is this true in practice? I'm working on adaptive beamforming for mmWaave Vehicular communications, and one assumption that makes the life better, yet it's not the case in the reality, is that orthogonality applies. In fact, we should expect interference between beams belonging to the same BS and interference between beams of different BSs. Thx
Hi Umberto, there is actually no assumption of orthogonality anywhere in the video. Interference between beams is an important thing that one must always consider, and it was included in all the simulation results. That is why we saw differences between different array geometries; some lead to more interference than others. My book Massive MIMO networks (massivemimobook.com) provides all the details.
@@WirelessFuture I mean, you don't explicitly assume orthogonality, but when you describe slide "Last Decade's Evolution of Adaptive Beamforming" and the subsequent one you talk about ORTHOGONAL 3D beams and orthogonal directions, just for this reason. Since I need to model a very complex and dynamic mmW vehicular scenario with multiple BSs, the choice of non-orthogonality is crucial. Anyhow, I downloaded the book, I shall go through it willingly!
umberto cappellazzo I probably misunderstood you. With M antennas, each beam is represented by an M-dimensional vector in an M-dimensional vector space. If you pick any set of basis vectors in that vector space, you get M orthogonal beams. However, it is highly unlikely that the users will be located so they only observe one of those beams. There will almost always be interference between beams, even if they are transmitted in orthogonal directions.
Thank you for the explanation. How do MIMO and Beamforming relate to each other? Is Beamforming a part of MIMO, or is MIMO used to detect the user whereas Beamforming is used to direct the signal to the user? Also, did you upload the slides somewhere?
There is a link to the slides in the description of the video. MIMO means that you have multiple antennas at both the transmitter and the receiver. One thing that these antennas can be used for is beamforming. I recommend my video series "Introduction to Multiple Antenna Communications": ruclips.net/p/PLTv48TzNRhaKz0C-dCAwimXSypV_5UTxg
Thank you, Sir, for the upload. As discussed that the spectral efficiency increases with the M/K ratio. What about the reliable detection at the receiver side in the case of Massive MIMO where the detection complexity linearly increases with the number of antennas. Is it not make the receiver in trouble to efficiently decode the transmitted information?
A linear increase is not particularly worrisome. For many types of operations, that even mean that the operations can be parallelized on the chip so that run time won't be increased. If it would have been growing faster than linear it would be more of an issue.
While the goal of SU-MIMO is to increase the data rate of a single user, MU-MIMO is all about increasing the total data rate in the cell. Hence, MU-MIMO is designed to deal with congestion.
Dual polarized antennas are actually two different antennas that emit and receive signals with orthogonal polarization. So it is not directly related to whether or not one can transmit or receive at the same time. FDD can be used to transmit and receive at the same time, but at different frequencies.
I'm no expert, this question will prove that, Have you thought / Can you by, still using towers as the main transmitter and phones as the receiver also have phones as small transmitters thus boosting the towers overall as well as not needing as many towers? Lol see no expert.
Yes, this is possible. The phone will then act as a relay. There is academic research on this topic but I don't think we will see it in practice. Just imagine having the battery of your phone drained because another person is using it as a relay. Would you really like that?
I appreciate all your excellent courses, dear Professor.
Great Lecture Prof. Björnson
professor Emil Björnson very well explained. thanks!!!!!
Thank you Associate professor Emil Björnson for this good presentation about Massive MIMO. You may help us to more about Indoor 3D Massive MIMO deployment and channel modeling.
I don't think that anything particular needs to change when it comes to the signal processing. But yes, the channel modeling will be different and I'm not working on that topic. There are 3GPP models for many kinds of propagation scenarios. I recommend the Quadiga implementation: quadriga-channel-model.de
@@WirelessFuture Thanks for your response!!!
Awesome lecture sir.Huge respect from India
Thank you professor! really great course
Loved the presentation, Thank you!
Great video
Thanks for sharing
1- may I know @10:51 how did you calculate the 16 dbi if each element of the 8 is 7 dbi ?
2- what kind of beamforming is done here is it analog beamforming where we merely control the Phase ?
3-@11:40 why can't we beamform in the vertical domain if we have access to the radiating elements per column ?
4-why do we need a pilot sequence of length tau if we can send a single symbol and let the base station measure the channel ?
Thanks
1. 10log10(8) = 9 dB is the beamforming gain. Add that to 7 dBi to get 16 dBi.
2. It is a fixed beamforming where all antennas send the same signal. It is like analog beamforming with no phase shift.
3. All the elements in a column must send the same signal. Hence the vertical beam shape is fixed.
4. We can use tau=1 in the case that you describe.
@@WirelessFuture
Thanks but for
3-a why in this case horizontal beamforming was achieved
3-b can we apply different weights across columns to achieve vertical beamforming, said in another way what does it take to achieve vertical beamforming ?
4-what are the cases where we need tau >1
5-when we talk about up to eigh beams this means simultaneously or at a time we can have just one beam at a time ? Can I know how in both cases ? If your answer was all simultaneously or one at a time
6-In many papers I see they assume tau k but I don't get it any clue why is that ?
7-when we say we have dual polarized antenna do we have single beam per polarization ?
I appreciate you recommended books or reference papers used for teaching those stuff may be used at Linköping or KTH Univs
Thanks
3a: Each column is connected to a different RF chain, thus we can vary phase shifts in the horizontal direction and thereby perform horizontal beamforming.
3b: The individual elements in column must be connected to separate RF chains (or at least phase-shifters). This is what is done in Massive MIMO.
4: If we serve K users, then we want tau=K.
5: Simultaneously.
6: We want tau>=K to be able to separate the users during channel estimation. You can think of it like having a seminar where the audience wants to know what K persons have to say. The K persons take turns in talking and therefore you need to divide the lecture into tau=K parts. When one person talks, all the receivers (people in the audience) can listen simultaneously.
7: One can transmit one or multiple beams.
I recommend "Fundamentals of Massive MIMO" and "Massive MIMO networks" (massivemimobook.com)
@@WirelessFuture
For 3-a and 3-b to better imagine the answer may I know what is the math behind them , I mean what equations govern those answers you provided ? Or what title should l look for
8-do we send independent data stream per polarization ?
Thanks
excellent video, thank you. subscribed =) hope to see your videos in future use-case.
Really nice lecture, enjoyed it a lot. Why can’t the industry take a different approach and devise what an ultimate use case for cellular communication might look like? And derive needed technologies and resources from that?
The ultimate use case would probably be so seldomly used that it isn’t economically feasible to deploy networks that support it. Most of the revenue that it is generated by wirelessly connected devices go to the device manufacturers and software/content providers (e.g., Google, Netflix), not to the telecom operators. That is why I think it is better to design future networks that can improve basic metric such as total capacity, uniformity of data rates, latency, reliability, and energy efficiency without making the infrastructure extremely expensive to deploy. Then the hardware/software companies will build products that make use of it as well as they can.
@@WirelessFuture my guess on what the ultimate case for the cellular communication might be is smthg that fully resembles and recreates in-person face-to-face interaction, capturing and re-creating it verbally and visually. IMHO, a face-to-face meeting is in ultimate act of communication the ICT whole industry is trying to re-create. No wonder why VR/glass free 3D content and tech are emerging. Yet, one can see how big the gap is between current capabilities and the final goal we strive to achieve.
Thanks Professor. Well done!
wow this information is great. thanks
Can't you use a "step up regulator" on the input to your antenna to increase range according to the size of the regulator? Not an electric voltage regulator. But a magnetic regulator. That's basically the same concept to increase the range?
Am I correct in assuming that beamforming is happening within ms to serve different users in different directions? Or does beamforming effectively act as serving static directions and moves over minutes? The amount of processing power to do active beamforming seems to be quite significant. Probably a stupid question but I'm trying to understand 5G physical layer / RU architecture and the benefits of functional splits primarily 7.1 and 7.2.
You are right that the beamforming happens at the ms level, at the frame level. The computational complexity might seem high, but it is actually just basic matrix/vector operations so it isn’t a big deal for a modern DSP or ASIC. The main benefit of functional splits and C-RAN is to gather many computational tasks at one point, and thereby save on processing resources. The logic is that each base station has a load-dependent maximum processing need, and it is unlikely to be needed simultaneously on many neighboring base stations. If they pool their resources, then one can save on resources and also enable more flexibility in the algorithms.
@@WirelessFuture Thanks Emil! That makes sense - so Ethernet packets are "aimed" at the user that it is intended for and will be "listening" with the same antenna formation for receive? I get that - it makes sense, as does pooling resources to reduce costs.
I'm at the early stages of learning about 5G technology but I've been interested in radio comms since I was very young.
Again, thank you for your time in replying to me.
Dear Pofessor, I have two questions.
1: How can we increase Spectral Efficiency in 5G Massive MIMO?
2: Why FDD is not suitable in Ma-MIMO?
These things are explained in a previous video: ruclips.net/video/m9wEAucKoWo/видео.html
In short:
1. Massive MIMO enables you to serve multiple users at the same time and thereby increase the spectral efficiency.
2. It is very costly to learn the channel properly in FDD mode, since the resources spent on channel estimation must grow proportional to number of antennas.
@@WirelessFuture Dear Professor, Thank you very much for your response.
you are amazing
thank u for these details please keep continuing, I'm a Ph.D. student do u have an online lectures i can attend
You can watch the videos from the course TSKS14: ruclips.net/p/PLTv48TzNRhaKb_D7SF3d1eNoqjrTJg34C
hello prof
how massive mimo makes most of the physical-layer control signaling redundant.?
what that mean?
THX
Massive MIMO has two key properties. 1) Channel hardening makes all subcarriers equally good for a user, 2) Favorable propagation makes it possible to serve many users at the same time, and yet protect them from interference using adaptive beamforming. The combination of these properties enable us to serve all users at the same time on all subcarriers, using predictable data rates. There is no need for intricate scheduling algorithms or quick changes of the modulation and coding, two things that usually require substantial control signaling.
Thank you. Great explanation. If the number of antennas is increased, the mutual coupling between them is also increased. Does it affect when switching between many receivers?
I don’t think that mutual coupling will become worse in Massive MIMO. The arrays become larger but the inter-antenna distances remain the same. Mutual coupling is mainly a problem that affect an antenna and its neighboring antennas, so it should not be affected by increasing the array size.
@@WirelessFuture Thank you, dear professor. Therefore, its major effect changes the pattern shape or scattering parameters of the antennas, and it does not depend on the size of the array.
Hi Emil, one question. Throughout the video you kinda considered that the orthogonality principle holds (i.e., the beams are orthogonal). Is this true in practice? I'm working on adaptive beamforming for mmWaave Vehicular communications, and one assumption that makes the life better, yet it's not the case in the reality, is that orthogonality applies. In fact, we should expect interference between beams belonging to the same BS and interference between beams of different BSs. Thx
Hi Umberto, there is actually no assumption of orthogonality anywhere in the video. Interference between beams is an important thing that one must always consider, and it was included in all the simulation results. That is why we saw differences between different array geometries; some lead to more interference than others. My book Massive MIMO networks (massivemimobook.com) provides all the details.
@@WirelessFuture I mean, you don't explicitly assume orthogonality, but when you describe slide "Last Decade's Evolution of Adaptive Beamforming" and the subsequent one you talk about ORTHOGONAL 3D beams and orthogonal directions, just for this reason. Since I need to model a very complex and dynamic mmW vehicular scenario with multiple BSs, the choice of non-orthogonality is crucial.
Anyhow, I downloaded the book, I shall go through it willingly!
umberto cappellazzo I probably misunderstood you. With M antennas, each beam is represented by an M-dimensional vector in an M-dimensional vector space. If you pick any set of basis vectors in that vector space, you get M orthogonal beams. However, it is highly unlikely that the users will be located so they only observe one of those beams. There will almost always be interference between beams, even if they are transmitted in orthogonal directions.
Thank you for the explanation. How do MIMO and Beamforming relate to each other? Is Beamforming a part of MIMO, or is MIMO used to detect the user whereas Beamforming is used to direct the signal to the user? Also, did you upload the slides somewhere?
There is a link to the slides in the description of the video. MIMO means that you have multiple antennas at both the transmitter and the receiver. One thing that these antennas can be used for is beamforming. I recommend my video series "Introduction to Multiple Antenna Communications": ruclips.net/p/PLTv48TzNRhaKz0C-dCAwimXSypV_5UTxg
Thank you, Sir, for the upload. As discussed that the spectral efficiency increases with the M/K ratio. What about the reliable detection at the receiver side in the case of Massive MIMO where the detection complexity linearly increases with the number of antennas. Is it not make the receiver in trouble to efficiently decode the transmitted information?
A linear increase is not particularly worrisome. For many types of operations, that even mean that the operations can be parallelized on the chip so that run time won't be increased. If it would have been growing faster than linear it would be more of an issue.
SU-MIMO may not perform well in a congested environment. How does MU-MIMO's perform in a congested environment?
While the goal of SU-MIMO is to increase the data rate of a single user, MU-MIMO is all about increasing the total data rate in the cell. Hence, MU-MIMO is designed to deal with congestion.
Can these dual polarised antennas send & receive at the same time?
Dual polarized antennas are actually two different antennas that emit and receive signals with orthogonal polarization. So it is not directly related to whether or not one can transmit or receive at the same time. FDD can be used to transmit and receive at the same time, but at different frequencies.
@@WirelessFuture thanq very much
link of the paper in video: www.sciencedirect.com/science/article/pii/S1051200419300776
I'm no expert, this question will prove that, Have you thought / Can you by, still using towers as the main transmitter and phones as the receiver also have phones as small transmitters thus boosting the towers overall as well as not needing as many towers? Lol see no expert.
Yes, this is possible. The phone will then act as a relay. There is academic research on this topic but I don't think we will see it in practice. Just imagine having the battery of your phone drained because another person is using it as a relay. Would you really like that?