How does 5G work? | A RARE look inside the 5G testbed facility @IITM | BP2B Labcast Ep 1

1. 0:00 – 0:23

   Intro

   1. SPSpeaker0:00

      a Jio telecom user and a Airtel telecom user. This happens with a periodicity of about 0.5 milliseconds. So this is our first, uh, fi- call on the indigenous 5G system from a commercial 5G phone. Here, this core network, we manage the traffic from all the deployments in the IIT Madras campus deploy.

   2. SPSpeaker0:17

      As far as I know, it took about four years-

   3. SPSpeaker0:19

      Yeah

   4. SPSpeaker0:19

      ... to finally finish up the test bed.

   5. SPSpeaker0:22

      Correct. [upbeat music]

2. 0:23 – 0:50

   Welcome to the Best Place to Build - Labcast

   1. SPSpeaker0:44

      Hi, and welcome to The Best Place To Build Labcast. Today, in the very first episode, we're gonna be visiting the 5G

3. 0:50 – 1:37

   Introduction to the 5G testbed at IIT Madras

   1. SPSpeaker0:50

      testbed facility of IIT Madras, the place where India's first-ever 5G phone call was made. Stay tuned, because we are gonna look at what they are building, how do they do it, and what are they building now that 5G is everywhere? Before we see what happens there, let's first understand how 5G transmission works. See, whenever you try to connect, your phone sends and receives radio signals to an antenna, the same things you see around you on towers. It processes these signals into digital signals and sends them to a baseband unit to handle the connection request. The BBU then forwards your traffic into the 5G core, which authenticates you, manages mobility, and routes your data to the Internet or another user. And now, let's have a look at it in more detail with Jeeva from the lab. So this radio unit is responsible for transmitting and receiving the radio signals. So this

4. 1:37 – 5:30

   Infrastructure involved in building the indigenous 5G

   1. SPSpeaker1:37

      is the component that is actually exposed, and when our mobile phone actually connects to something, it sends the radio waves. The number of radio elements have significantly increased. Typical 4G deployments will have one to four antenna elements, but in a 5G, if you see, there'll be 16, 32, and 64 antenna elements that are present over there.

   2. SPSpeaker1:56

      So why, why do you need such complexity? Why do you need to keep adding more antenna elements?

   3. SPSpeaker2:01

      So when I speak to you, I look at you and speak, but when a radio unit is talking to you, it doesn't know exactly where you are, or it cannot rotate itself. So what it does is, a typical radio unit, it'll broadcast the signal in all the directions. When I have multiple antennas, it is possible for me to direct my energy into the, uh, direction where a user is present, where it is required, so that I don't waste the energy which is going on other directions. Having said that, the catch is that every antenna element here has to be fine-tuned and calibrated, and this is not a simple calibration process. That, uh... This involves a lot of algorithms. And this radio unit, where does it sit? Somewhere on a tall tower, so it is not possible for anyone to actually go over there and calibrate it every time. So this has to happen-

   4. SPSpeaker2:49

      And how often do you need to calibrate it?

   5. SPSpeaker2:51

      So it depends on the algorithm, and very important part about calibration is it should not disturb an actual transmission and reception that is happening. We decide that as well based on the changing channel scenarios, where do you deploy it, how often the scenario is changing. It depends on the varying temperature as well, and aging as well... lot of parameters.

   6. SPSpeaker3:10

      If I am-- you know, I go to a Jio tower and sort of open it, how many antennas am I likely to see?

   7. SPSpeaker3:17

      So they also have multiple deployments, so typically you would be see somewhere about 32 antenna system. So this is 16 antenna system. The way that we built this, it's a modular structure. When it is a 16 antenna system, uh, inside the system, if you see, there would be two digital processing units, and if you see, um, 32 antenna system, you would have one inter-aggregator and two similar, uh, digital processing units.

   8. SPSpeaker3:43

      I do understand the function of the 16 antenna now. Um, what's this, and what's this, you know, black cap on the top?

   9. SPSpeaker3:51

      Okay, so you put this radio unit on the top, and it is emitting so much of power out of it. So isn't so much of heat also generated? If you put fans, then you're creating space for dust and water and other moisture and other ingredients to get in, but it is not possible over here. So that's why we have state-of-the-art, um, heat dissipation fins over here.

   10. SPSpeaker4:13

       Okay.

   11. SPSpeaker4:13

       These fins actually take heat pipes all the way from the junction point, where there's a huge temperature gradient, and we dissipate it through these fins. That is why-

   12. SPSpeaker4:20

       Okay, this is a thermal fluid cooling.

   13. SPSpeaker4:23

       Exactly.

   14. SPSpeaker4:23

       Uh, and then over here, there's an interface so that the fins can dissipate heat out.

   15. SPSpeaker4:27

       Exactly. And this cap that you were talking about, this is an antenna that is used to synchronize our system with the, uh, GPS, uh, satellites, GNSS satellites. The other way is my base station and the other base stations which are nearby should also be in sync. So even though they are distributed across frequency, it is required that they are sync in time as well, so that they don't interfere with each other.

   16. SPSpeaker4:53

       Now, when communication happens between, say, you know, a, a Jio telecom user and a Airtel telecom user, this antenna would be used to sort of transmit the signal from, like, via the satellite to the other tower, or-

   17. SPSpeaker5:06

       Yes, there are-

   18. SPSpeaker5:07

       ... tower-

   19. SPSpeaker5:07

       There are satellites all over the globe.

   20. SPSpeaker5:09

       Mm.

   21. SPSpeaker5:09

       And these antennas actually receive the signals from the, uh, satellites. These are the common satellites, and from them, we receive the signal, and then we ensure that our timing, our transmission and reception timing, are synchronous to that satellite timing. That is what we take care of.

   22. SPSpeaker5:24

       Did you know that until 4G, we had an antenna handle only about 64 connections simultaneously

5. 5:30 – 8:25

   Why is 5G so much faster than 4G?

   1. SPSpeaker5:30

      at one single instance? That's an 8 by 8 multi-input, multi-output system. But this was the reason why it was so much faster than 3G. However, with 5G, they've been able to accomplish even greater speeds, thanks to 16 by 16, 32 by 32, or even 64 by 64 MIMO connectors.

   2. SPSpeaker5:48

      MIMO means that multi-input, multi-output, multiple inputs and multiple output. So we have separate chains which is processing all this data, and inside, in the next component that I'll be showing you-... all these, uh, signals received from multiple antenna is processed together, and then we make sense out of it.

   3. SPSpeaker6:05

      You are saying that each of these sixteen antennas are collecting input as well as output?

   4. SPSpeaker6:11

      Yes.

   5. SPSpeaker6:11

      Once an antenna converts a radio signal to a digital signal, the baseband unit handles the request, which means it assigns radio frequencies and time slots, controls the power and gain for signals, manages handovers between towers to ensure low latency connection. How does it do that?

   6. SPSpeaker6:27

      So a baseband unit consists of two parts. One is, uh, physical processing unit. One is the physical layer's significant part over here. So we reserved a lot of data from the radio unit. So how do we... where do we process all this data? So our part of processing happens in the FPGAs that are present in the radio unit itself, but the remaining processing all happens in this physical layer. So what we were receiving-

   7. SPSpeaker6:51

      How do you, how do you break down what processing happens there, what processing happens here?

   8. SPSpeaker6:55

      Yes. So this is, uh, defined by, um, O-RAN. O-RAN is a body that defines wh- where we can do these splits, so that baseband unit from one company and a radio unit from another vendor could talk to each other without any hassle. So we have implemented something called as O-RAN seven point two split. According to this split, a part of signal processing happens in the baseband unit in the ground, and the remaining signal processing happens on the radio itself. When there is a lot of noise in the environment, despite of all that, our algorithms are effective enough to remove all this noise and still make sense of what the user wanted to talk to you. So I understand that the hello that has come from you is the original signal, and all the other elements that I'm listening to is noise. So that is what I mean by channel estimation also. And this estimation is not something that happens once in a long while. It happens periodically, because-

   9. SPSpeaker7:50

      How periodically?

   10. SPSpeaker7:52

       This happens with a periodicity of about point five milliseconds. In some milliseconds, this channel keeps varying, and then we keep estimating it.

   11. SPSpeaker7:59

       So that ensures that even when I'm in a train or when I'm traveling in a rickshaw, like, the correct signal is traveling?

   12. SPSpeaker8:06

       Exactly. So when you are moving, your location continuously keeps changing, so the channel also has to be estimated so periodically so that it can get it to you. We are looking at the core network. Core network is like the brain and backbone of the 5G infrastructure. This is an important system which manages the traffic, user information, the internet,

6. 8:25 – 9:40

   Glimpse at the 5G core network system

   1. SPSpeaker8:25

      uh, request, and the policy-related information, if you have a recharge pack, if you don't have a recharge pack, and if you also can connect to this particular base station at this point of time, all these are managed by this core network. Just to give a scale, it's a similar core network that is deployed one... at one place for the whole of Chennai by an operator, like Jio or Airtel. So they- it manages the traffic from all the nodes. Similarly, here, this core network, we manage the traffic from all the deployments in the IIT Madras campus deployed. Till 4G, this core network was typically, uh, deployed in a custom hardware because general purpose hardware was slowly developing over days. But for 5G, it is completely a software-based deployment. So what it means is the whole algorithm, all the features, all the network functions of the core network, I can typically put it on a laptop that you use it on every day, but the scale of that would be very limited. I could probably be connecting only, like, ten or fifteen users, and with the scale of this, we could hun- we could connect, like, thousands of users, and with a scale of, like, one in Chennai, we will be connecting lacks of users, millions of users.

   2. SPSpeaker9:34

      After we got a run-through of the lab, I also got to hear the story of first 5G phone call in India from

7. 9:40 – 12:34

   Story of India’s 1st official 5G phone call

   1. SPSpeaker9:40

      researcher's perspective, followed by: What are they doing now? Is there a plan for 6G? Where are we? First question from me is, uh, you know, first of all, we st- decided at some point that, you know, we have to start working on 5G. I think that was around 2018 when we realized that, oh, my God, we've already been delayed in implementing 4G in India. Don't want to do that to 5G. What were the first steps, and, like, how did the timeline look like?

   2. SPSpeaker10:06

      So it is not that we started only in 2018. A lot of, uh, work in the academia has been happening since a very long time. For 5G, what are the new features that has to be incorporated, or what are the new algorithms that has to be put in place to meet the required data rates? All these things we have been discussing, and we have been implement... putting it on papers. There has been a lot of discussions, but in 2018 is when the government convened a meeting across all the IITs, and different Indian Institute of Technology have expertise in different concepts. Why not all of them put together, build something that is state of the art, that can be con- as a big contribution to the industry asset?

   3. SPSpeaker10:48

      That's how a consortium was formed, and-

   4. SPSpeaker10:49

      Correct

   5. SPSpeaker10:50

      ... we started working, and as far as I know, it took about four years-

   6. SPSpeaker10:54

      Yeah

   7. SPSpeaker10:54

      -to finally finish up the testbed.

   8. SPSpeaker10:56

      Correct. It was only in April 2022 that we read the... we made the first official 5G call. So we were all working in the lab, like, every day, and there are, like, multiple steps that we have to cross to ensure that the phone latches to the base station. The previous day, it was a Friday, we, uh, tested, we took a lot of logs, and then we understood that these are some power parameters that has to be changed, and we were fixing them, and it was late, so we went. And in the early morning, we came, and then we put all those parameters, and we checked, and for the first time, we saw a phone got connected to our 5G base station. So that was, like, a eureka moment, and, uh, we were, like, overjoyed, and our professor was already on the way to our lab. So he came over there, and then we explained to him that we were able to achieve this. Immediately, he was overjoyed, and he called Professor Bhaskar Ramamurti. He rushed over to the lab to the- to witness that, uh, first 5G call through our own base station. Then he called Professor Kamakoti, then it went on media, it went on news. Okay, so this is our first, uh, fi- call on the indigenous 5G system from a commercial 5G phone. Completely indigenous stack. Congratulations to the entire team. So that particular day, if you see, it was like, um-... uh, reward to all the hard work that a lot of people have been, uh, contributing for those many days. So it took us after that, about a year to, uh, say that, uh, concretely, when we put a system outside in the field, that works without any problem, and the phone latches, and we can do a long field test.

   9. SPSpeaker12:25

      You mentioned that you design it end-to-end, so-

   10. SPSpeaker12:27

       Correct

   11. SPSpeaker12:27

       ... that means there would be mechanical engineers, uh-

   12. SPSpeaker12:30

       Yes

   13. SPSpeaker12:30

       ... designers.

   14. SPSpeaker12:31

       So there's a thermal team and a mechanical team. They design

8. 12:34 – 14:30

   The teams behind the 5G innovation

   1. SPSpeaker12:34

      all this chassis that you see over here are custom designed, and there is a RF team which designs the antennas, which, uh, also takes care of the amplifiers that are present, uh, in the radio unit. And there is a hardware team who design the hardware, the PCB boards, and along with that, there's a very big team that builds the code base, which has to sit on this particular FPGA. Along with this, we have another simulation team. So that team mainly reads the 3GPP spec, and they also build algorithms, and they do end-to-end simulations on tools like MATLAB, C, Python, to understand this algorithm, does it actually theoretically meet the spec requirements?

   2. SPSpeaker13:14

      What have you been, say, working on the last six months, two months?

   3. SPSpeaker13:18

      In the recent days, what we have been, uh, testing is many features like mobility. Mobility is a feature when a phone travels very fast from, uh, one base station to the other base station, how can it connect with minimal, uh, latency? That is one feature. And how to better optimize all these receiver and transmit algorithms. These are all some things that we regularly work on.

   4. SPSpeaker13:41

      Oh, that's awesome. It's sort of grown in pute- uh, in scope beyond 5G just being a testbed to a lab. Uh, do we also work on 6G here, and, like, how is it going?

   5. SPSpeaker13:52

      Correct. Very rightly said. For 5G, we built this testbed, we built this whole lab, and we have, like, the entire infrastructure architecture. Taking learnings from all this, we are building next-generation radio units as well, and next-generation baseband units, so that they could potentially be a solution for coming... upcoming technologies like 5G Advanced and 6G. We were showing about a sixteen-channel radios unit, thirty-two channel, and sixty-four channel. Probably for sixty-four, uh, for the 6G, the number of radio elements, MIMO elements, could tremendously increase.

   6. SPSpeaker14:25

      You also mentioned that, like, I've also sort of come to see on professor's research page also,

9. 14:30 – 17:00

   How is AI/ML used in the 5G testbed facility?

   1. SPSpeaker14:30

      that there's a lot of AI ML applications in 6G that we're working on. Uh, first of all, why is AI ML needed? I mean, on, on what particular things is AI ML needed? And secondly, what, what are we researching?

   2. SPSpeaker14:43

      So there are conventional algorithms for many solutions that we have put forward. When you are able to frame a problem very properly, some of them are easily solvable. But for a very dynamic, changing scenario and new use cases come up, which- for which AI ML has proven to, uh, give or present a better performance.

   3. SPSpeaker15:06

      Okay, can you give an example of one such scenario, like-

   4. SPSpeaker15:09

      Correct

   5. SPSpeaker15:09

      ... uh, dynamic scenario?

   6. SPSpeaker15:10

      So one pos- one scenario is, uh, user positioning. So how can we do a centimeter-level positioning with the base station and when a user is present? For multiple applications. In an indoor factory scenario, you want to tag where a particular product is on the conveyor belt, so you need centimeter-level accuracy. Can we reduce the amount of time that we take for doing all these measurements and leverage AI, artificial intelligence-based algorithms, and try to solve this issue, is something that we look at.

   7. SPSpeaker15:39

      Quite a bit of, uh, work there, right? In terms of hardware, in terms of infrastructure, is there any work happening? Do you think that we'll be able to use the same 5G testbed for, uh, 6G, or, like, how would things change?

   8. SPSpeaker15:51

      So every day, changes are happening, uh, for 6G. 6G, people are still anticipating what could be potential use cases of 6G and what are the algorithms that could potentially be a part of it. So we are also taking active part in that and identifying use cases and identifying good number of problems that could be solved, uh, in 6G, which are not yet, uh, solved in 5G kind of infrastructure.

   9. SPSpeaker16:16

      Awesome. So in terms of what we have come to understand at this lab, what are some, you know, current issues with 5G that you would sort of love to solve in 6G?

   10. SPSpeaker16:26

       So one issue is there is a lot of overhead on the reference signals. So for channel estimation and other purposes, we are transmitting a lot of known signals, and that also occupies a significant part of bandwidth. So how can we reduce that? And, uh, in current 5G, the main use case that has actually kicked off is the end user, EMBB, we call it, so end user, uh, application. So users are the main connecting nodes, but in 6G, we are, uh, foreseeing a lot of IoT-based nodes which could

10. 17:00 – 17:34

    Closing thoughts

    1. SPSpeaker17:00

       come- which will do a very small processing, and they can still connect to the existing base station itself. That is something that we are foreseeing.

    2. SPSpeaker17:07

       So it will be like each IoT device would become a receiver and a transceiver at the same time?

    3. SPSpeaker17:12

       Yes.

    4. SPSpeaker17:13

       I'm really excited to see what you guys come up with.

    5. SPSpeaker17:16

       Definitely.

    6. SPSpeaker17:17

       Um, I think that's all, Jeeva. Um, thank you so much for your time.

    7. SPSpeaker17:21

       No problem.

    8. SPSpeaker17:21

       While the researchers continue their work, we'll continue ours and take you around to other labs of IIT Madras, where cutting-edge research keeps happening. Subscribe to Best Place To Build for more.

Episode duration: 17:39