5G beamforming has emerged as a scalable and economical choice among the MIMO techniques you can use for developing 5G systems. 5G beamforming separates the design between the RF and digital domains.
In this webinar, we will discuss the end-to-end 5G hybrid beamforming design workflow. Topics include:
Recorded: 15 Jan 2021
Hello, everyone. My name is Houman Zarrinkoub. I'm the product manager for the wireless products and MathWorks including 5G, LTE, wireless LAN, and communications toolboxes. And it is my pleasure to welcome you to this MathWorks webinar entitled 5G Beamforming Design.
Let's go over the agenda of this presentation. After some introductory remarks, I essentially going to present five different sections related to 5G beamforming. The first section is all about 5G waveform generation. This second topic is all about channel modeling and precoding. When you are working with beamforming scenarios, you have to be aware of the effect of channels on propagation environment, and you also have to design precoders that specially direct your transmissions.
Which brings us to the third topic, the antenna array design and the MATLAB tools we have to not only design antenna elements as well as antenna arrays that help with characterizing or beamforming performance. Another topic is hybrid beamforming. The complexity and the power consumption of a fully digital beamforming, can be mitigated by using hybrid beamforming approaches that divide the beamforming into two modes. In some direction is the analog, and in some direction is the digital approach.
Finally, you have to design the RF front-end in order to improve your performance, and for that, we have tools in MathWorks that help you characterize the power amplifiers and design DPDs and other elements that related to RF transmission chain. Finally, I'm going to provide some summary.
Let's go over to enabling technologies of 5G. As you know, 5G has multiple use cases, and one of the most familiar use cases is the 5G enhanced mobile broadband. And that is the challenge that the networks that are going towards 5G is dealing with right now, how do we substantially increase the throughput and data rates over our mobile communications [? definition. ?]
So I might think of the solutions that have been proposed is the following. You can increase the bandwidth by using your signals, a larger bandwidth. You can essentially increase your throughput. And you can also squeeze more from a given Hertz of bandwidth. That's called better spectral efficiency, how much more bits per second can you use per Hertz.
You can also use flexible interfaces. In this case, it means using multi-carrier communication systems that use OSM or OSDMA and multi-user MIMO. And also, you can use small cells such that you can essentially use MIMO techniques to achieve better received power and performance.
Another approach is also using a higher frequency band. And that's one of the main themes that 5G will exhibit. Instead of working at the typical below 6 gigahertz frequencies in the 4G and before similar communication systems, we are going beyond the [? hertz, ?] millimeter again. And with using that, you have large swaths of bands available for you, and you can essentially achieve the high throughputs you want.
Now, all these things are not without challenges. The minute you decide to go to millimeter wave and higher frequencies, the simple intonation becomes a real problem. Systems with very broad bandwidth, the frequency selective fading becomes a real problem. And you have to seriously look at multicarrier systems like OSTM, OSTMA, and channel estimation equalization in order to combat the effect of frequency selective fading. And also, you need scatter-rich propagation environments in order to make this viable or if you have systems possible. And then you have to constantly update your channel characteristics and your models.
How to put it all together. You're using high carrier frequencies, larger bandwidth and large antenna arrays. This is an approach that works with 5G, especially at millimeter wave. This is known as massive MIMO. And massive MIMO essentially means that you're using very large antenna rays in order of much about 64-- 64, 128, 256, and so on. And remember that if you are doing all digital beamforming are these large antenna arrays, behind every antenna that transmits this signal, there is an RF analog chain, wideband chain that requires a large amount of power and complexity.
So in order to mitigate the power and dissipation problems of these massive MIMO systems, hybrid structures have been proposed to achieve the best compromise. As you can see, in the vision, the massive MIMO technology of the future with a transmitter that has narrowed beams that targets each individual receiver, that's in multi-user MIMO scenarios. That requires large antennae arrays to achieve, essentially, narrow beams.
And we don't have those kind of a transmission in all directions isotropic anymore. We are doing focused transmission. And in order to achieve that, you need to essentially coordinate your precoding and beamforming together with RF chain such that you can custom design transmission to different users at different narrow beam angles. So what is the workflow for such hybrid beamforming 5G architectures?
You need to generate a waveform. These waveforms essentially determine how much bandwidth you're using, and there's lot of flexibility there in 5G, and I talked to you about that. You have to design antennas and antenna arrays enable you to achieve those narrow beams that participate in the process of multi-user MIMO massive, MIMO structures. You have to have a representative channel model with lots of scattering that included in order to make the essential estimation of the angle of arrival, angle of departure, the locality of the transmitter receiver.
All of that possible. And based on that, if you want to use special multiplexing and send multiple streams on different antenna ports, you have to compute the precoding matrices, which are the digital beamforming requirement. The digital precoding matrices that multiply [? weights ?] to the digital beamforming architecture in order to steer your signal digitally to the direction that you want. And if you are doing hybrid beamforming, you also have to consider the analog beamforming parts, those elements that use phase shifters in order to essentially steer in the analog context without using all the RF chain to steer your beam in another direction.
And after all of that is done, you have to consider the effect of the RF front end, the amplifiers, the filters, and so on, and make sure that within your band interest, the amplification is linear, otherwise, you suffer a lot. So all of these considerations, which are call system-level design considerations come into play to produce a successful 5G hybrid beamforming architecture.
Let's talk about the first element, the 5G waveform generation. 5G standardized in 2018. And in MathWorks, we have introduced our 5G toolbox, our MathWorks product that allows you to model, simulate, and test 5G systems in the same year, 2018. And one of the characterizing features of 5G technology is that it is based on OFDM, however, unlike 4G when a subcarrier spacing or the resolution in frequency were constant, in 5G, subcarrier spacing of the resolution frequency is a variable.
So your difference in frequency between two subcarriers can be 15 kilohertz or multiples of 15 kilohertz or a power of 2, 30, 60, 120, and so on. Now, we have a duality between time and frequency, which means that if your subcarrier spacing or something is short, the duration in time is long. So for 15 kilohertz, 1 millisecond slot in time, and as you increase subcarrier spacing, the slot in time reduces in size.
But you know that the minute that we use multiple of subcarrier spacing, we achieve, immediately, a higher bandwidth. So if you send the same information over a 15-kilohertz subcarrier, you are at the 50 megahertz, if you like, bandwidth. But by doubling the subcarrier spacing, you are occupying, essentially, 100 megahertz. These are the considerations that show the flexibility of 5G in terms of time frequency resolution. And these are the type of waveforms you have to generate in order to test and validate your 5G and 5G beamforming designs.
As I mentioned, 5G toolbox supports both uplink and downlink 5G wavefrom generation. And as you can see, we have the waveform and 5G tool box, essentially, allow you to have mixed numerologies, signals, and waveforms of various subcarrier facing, multiple bandwidth parts, each bandwidth part associated with different user or EU, multiple physical downlink, or physical uplink shared channels. They're all fully parameterizable and allow for the synchronization signals and bursts, SS bursts, that are needed to synchronize everything in 5G.
I'm going to show you with a very easy-to-use app called Wireless Waveform Generator app, which has been introduced in our tools ever since 2018 B release, how easy it is for you to generate 5G waveforms. You're going to test your wireless designs interactively essentially, using an app by clicking the right parameters. Even if you're not a proficient MATLAB programmer, you can interactively parameterize and generate 5G waveforms.
And the Waveform IQ samples that are at the base band, you can actually see the MATLAB code that generates them. And you can transmit them over the air right from MATLAB environment if you're using an RF instrument that connects to MATLAB. Let's go to MATLAB. Take a look at this app known as Wireless Waveform Generator App that makes the process of 5G waveform generation very easy.
In MATLAB, the version released 2020A, we can go to the Apps tab and look for the Wireless Waveform Generator app. Now, this app enables you to generate a variety of waveforms without having to write a single line of MATLAB code. As you can see, multiple standard-based waveforms are available here, the 5G and R, DLT or 4G, the wireless LAN or Wi-Fi, different flavors, and the Bluetooth Low Energy.
Now, we're focusing on the 5G waveform generation. Notice that you can generate, uplink, or downlink fixed reference channels, or we can use the new radio test models in RTM. In this case, we're going to choose an RTM. We have multiple options here, the frequency range FR1, which is below 6 gigahertz frequencies or FR2D, the [INAUDIBLE], and choose the FR1, and the multiple standard-based test models available here. Each of them represent different modulation schemes and use of the bandwidth.
We're going to choose the 64 QAM full band that looks by TM3.1. In terms of channel bandwidth, as you can imagine, we have multiple bandwidth possibilities here. We can choose a 10-megahertz bandwidth. And the subcarrier spacing can be values 15, 30, and 60. We're going to choose a 30 kilohertz and choose our duplex mode, FTP. Of course, we can do also filtering, if needed.
Just by setting the parameters and clicking on the Generate button, you can see that we can generate the corresponding 5G and R waveform. As you can see the bandwidth is 10 megahertz as required, and you can see that we have a channel view with the channel edges and the guard band and the transmission bandwidth. And we have a representation of various elements in the resource grid.
Notice that you can go to Visualize and add other elements to your signal, including time scopes to look at the IQ samples. You can alternatively go to the impairment section and add various impairments to your signal. This is when you are actually modeling the transmission or reception. Doing the simulation in software, you can add various impairments as you can see here.
Finally, you can export the signal to workspace or to a file to be used for further processing. All right, let's take a look at the second topic, the channel modeling and precoding. Now, 5G standard, one of the first things that was published with 5G standard was the 5G channel models. These channel models are introduced in a technical report, 38.901, and they allow you to have a very rigorous spatial channel modeling characteristics for your MIMO channel models.
Essentially, not only we have the traditional channel modeling, you have profiles like delay profiles, the TDL and CDL profiles and delay spreads at the, essentially, impulse response and Doppler shifts. We also have antenna array geometry specifications that enable you to figure out when a beam form signal is transmitted over the air, what is the effect of scattering and all other channel modeling based on the clusters that we see at the scattering locations. So 5G allows you to specify a very good channel modeling specification for MIMO and beamform-based transmissions.
The other tools in MathWorks, we have Communication Toolbox, our based product, as well as Phased Array System Toolbox that provide much more support in terms of propagation channels. We have free space path loss, which can be used for large distances or attenuation due to atmospheric gases and fog and clouds. MIMO multipath fading channels. We have Winner II fading channel models and scattering MIMO channels, which allow you to specify exactly if you know the location of transmitter, location of receiver and the location of scatterers, you can do it all in MATLAB and find out what is the actual received power signal arrived at the receiver as a result of propagation channel model.
Let's look at the MIMO OFDM data flow to the foundation of LTE and 5G and other similar things. So when you want to do beamforming or special multiplexing, especially MIMO signal processing or OFDM signal, the process is done in two phases, essentially. You have to do what's known as channel sounding to understand the impulse response or the frequency response of the channel.
You send a preamble to transfer that signal. It goes through the MIMO or your propagation environment. It's received at the receiver of the OFDM. Signal processing is done OFDM demodulation, and you do a MIMO channel estimation. At each subcarrier, you perform, essentially, you compute a channel matrix. Provides a matrix of gains at each subcarrier relating D transmit ports to the receive ports.
For every matrix using SVD, it can be diagonalized, and you can find the precoding matrices that can diagonalize your system, which means that if you use those precoding matrices, V, in this case, then you can steer your transmission digitally in a direction that enables maximum retention by your receiver. So that's what the channel sounding is.
In the next step, you essentially go through waveform generation, and then from the sounding environment, you used the precoding matrix computed, and you compose your OFTM transmitting single with precoding. And then the MIMO channel happens. They receive the beam steering and the OFDM happens. And again, through channel estimation and equalization, you actually re-compute the precoding matrix that either transmits, or use that to recover database. This is the process of MIMO OFDM data flow used in 5G systems.
So precoding plays a big role, and therefore, channel solving plays a big role in successful 5G precoding. The practical approach for 5G has been the use of CSI and reference signals, and they use the codebooks to find the best precoding vector match at the UE and transmit that in the uplink to base stations so that that precoding method can be used in subsequent transmissions. But this architecture informs how for digital precoding, the channel funding and the transmission go hand in hand.
So back in MATLAB environment, I want to start with typing the word doc, documentation. And in documentation, look for massive MIMO. Now, that gets you to the example that I want to show you today. It's called massive MIMO hybrid beamforming. It's a very good example because it provides the MATLAB program that exactly talks about channel sounding and use of massive MIMO. As you can see here in this example, we are looking at a hybrid beamforming scenario for multi-user and single-user systems, which is exactly what 5G requires.
The partitioning of precoding and the digital and RF component is done. At the end, some EVM and BER measurements are completed to characterize your system. You see that we're using a scattering-based spatial channel modeling, as I told you before. And essentially, we are looking at a multi-user MIMO FDM system with the beamforming and application of large antenna arrays.
Now, if you open the script, you'll notice that we are looking at MATLAB. Code is all in MATLAB, so it's easy to understand, so we're using some parameters to go through the steps. We're using four users, and each user is allocated different number of streams, so you're using special multiplexing context. And the number of base station transmission antennas are essentially the power of 2 multiple of those streams, so in this case, 24 and so on.
We can use multiple modulation schemes, and notice that we are simulating a millimeter wave situation, 28 gigahertz, and a large number of arrays, in this case, 500. It's a MIMO OFDM system, so we have to specify OFDM parameters. In this case, we're using the 256 FFD, OFDM, and so on. So we use the phase array system toolbox functionality like a partition array or uniform linear array to compose the antenna arrays. And you notice here how we use the channel state information.
For each user, we are computing, essentially, the preamble signal, and we are transmitting it, scaling it. You have OFDM demodulation tool, perform channel estimate, and through SVD, we compute the channel matrix estimate. And here is where we compute the hybrid weights on the transmitter sites. For each user, we divide in the multi-user sense with the base band and RF, which are beamforming, and compute the matrix, V. And we go through the convolutional encoding modulation and all that look like performance metrics such as bit error rate and EVM.
As you can see, for every user, we compute the RMS EVM values, the bit error rates, a number of errors as well as we look at the constellation after equalization for each of those streams for user number 1, which are three streams, user number 2, number 3, number 4. And look how, for each of the four users, we have computed the beamforming going at different angles, and we visualize that using the phase array functionality. So it is easy to understand in MATLAB how theoretical aspects of massive MIMO high beamforming are represented as practical procedures at [INAUDIBLE].
What else do we need for a successful 5G beamforming design? We need antenna arrays. You can use Antenna Toolbox, one of the most popular toolboxes you have in MathWorks, but enabling it not only to design antenna elements, but also antenna arrays enables you to visualize and analyze antenna data in detail, the beam with the coupling and all the other stuff that are needed to characterize your antenna transmissions.
And not only we have done it at the point level, we have also introduced this RF propagation visualization analysis over maps. So we can in MATLAB, open the maps of different geographic areas, insert in a specific location transmitter and receivers, and observe the received power as INR and all these other characteristics using Antenna Toolbox.
I'm going to take you now, again, to the MATLAB App tab and show you two apps, the Antenna Designer as well as the Antenna Array Designer, which enables you to interactively design antenna elements and antenna arrays that can be used in beamforming patterns. Like in MATLAB, going to the App tab, Enforcing and Fast Communications app are located beside the 5G generator that we saw before. I want you to take a look at Antenna Designer and Antenna Array Designer.
First, look at Antenna Designer app. We notice that we can start designing an antenna from scratch by clicking a Plus button, and then you can choose from a variety of antenna design galleries, in this case using a dipole, but we can use all kinds of existing antennas, structures. And then you can specify whether radiation of each antenna is directional or omnidirectional in all directions.
We can specify the polarization and the bandwidth as well as specify the operating frequency. Let's just look at a 28-gigahertz frequency millimeter wave, and you say Accept. When you do that, each antenna is characterized with all these geometry and load properties. Can apply that. And look at different analysis capabilities of the Antenna Toolbox.
For this particular antenna element, I like the 3D pattern, which shows, essentially, a symmetric directional approach, but you can choose any antenna you want. So you can save this and export it as a script, so you can put it in your simulations. There's another app, Antenna Array Designer. We are not looking at the design of each antenna, and then you're looking at the antenna array.
So the first parameter is the array size. Let's just go toward a more elaborate array like 8 by 8, and this was 64 antenna element. As you do that, notice that the app immediately organizes the antenna array for you based on geometric array characteristics you have here. Can be rectangular array, circular array, conformal array, and so on. And then you specify your desired frequency. In this case, I put 28, and except it.
You will see that the geometry and all the characteristics here are specified for you, can modify all these parameters, apply, and look at things like antenna propagation array patterns. Here, look at the 3D [INAUDIBLE] pattern of this particular antenna array, at the desired frequency of 28 gigahertz, and you'll see we can compute other patterns such as remote and elevation. And finally, we can export the results to our MATLAB environment.
Let's go to the next topic, hybrid beamforming. So what is hybrid beamforming? When you decide that the full, complete digital beamforming is expensive in terms of power usage and dissipations, you may consider using hybrid beamforming, where the RF beamforming done in RF front ends without the use of full RF chain. That's analog beamforming. And digital beamforming is done in another direction or another dimension using the precoding method I told you about.
So that division of task into analog or digital is not a hybrid beamforming, and it's a very pragmatic and practical approach to massive MIMO beamforming. So I'm going to go over an example and show you how this decomposition is done. So we are looking at a system where we have estimated the angle of arrival and departure, PX and RX, and we are using a hybrid structure where we are combining digital transmit array with analog phase shifters to achieve a hybrid beamforming solution.
For example, we got to use digital beamforming in Azimuth direction, so the precoding matrix is applied there. And we're going to do RF beamforming in the elevation. And you can see we can reach multiple UEs in a multi-user MIMO format using these architectures. If you look closely using channel sounding, we compute the V matrix and multiply different stream by those digital rates.
But if you look at the analog for RF beamforming structure, you see that we are using phase shifters. The angle is derived from the weight stat we computed in the different direction. This architecture enables you to add the characteristics-- the thermal noise, phase noise, image rejection, and so on. And it creates a very realistic approach to hybrid beamforming for your design.
Finally, we're going to use the 5G in our transmitter. Those waveforms are generated as you saw before. You're asked to use as the transmitter signal, and we're going to use the 5G in our receiver in order to make measurements and estimations.
Let's look at the next at a final stage, design of RF front ends. So the elements that undergo digital beamforming, they require you to have an RF chain associated with each of those antenna points. How do you design the RF chains that take the baseband signal and bring it to RF, and then they are subject to beamforming before they're transmitted over the air.
We can use another app in our MATLAB environment known as RF Budget Analyzer app. This app is part of the RF toolbox. Allows to implement power, noise, IP3, and all other RF analysis and computations. It helps you come up with the RF processing chain composed of filters, low-noise amplifiers, the mixers, and power amplifiers that needed to essentially take the signal to the RF and transmit over the antenna.
Now, when you open the RF Budget Analyzer App, you see how easily you can add various RF chain components and cascade them together. Look at the overall gain, [INAUDIBLE], all the characteristic needed to come up with budget analysis and to compute the received power.
One interesting and important part of designing any RF component is power amplifier or PA characterisation, as well as legalization with digital pre-distortion. Over the large input ranges in DBM of signals, especially as the bandwidth increases substantially, if your power amplifier fails to be linear, the performance is severely impeded. So what you need to do is come up with closed-loop behavioral simulation and closed-loop designs that linearize the power amplifier across the operating range and frequencies in order to make your transmission successful.
Now, you can use our RF toolbox and RF [INAUDIBLE] capabilities for PA modeling workflows with DPD design. And I will refer you to a complete presentation on that by my friends shortly at the end of summary. But essentially, the gist is that you can get the IQ time domain wideband measurements data from your PA. And using our MATLAB polynomial fitting functionality, you can actually fit the data with the polynomial as you can see on the bottom. And using that, you can pre-distort with the inverse of those polynomials and essentially achieve linearization.
To summarize, we can learn more about hybrid beamforming and its applicability to 5G designs by exploring our examples that we have in mathworks.com. Just got to mathworks.com, and look for four more detailed presentations and examples of reference designs that we have provided for you, introduction to hybrid beamforming, hybrid MIMO beamforming with QSHB and HBPS algorithms, massive MIMO hybrid beamforming, and modeling of RF millimeter wave transmitter with hybrid beamforming.
They all can be found on our wireless communication solution page at mathworks.com/so lution/wirelesscommunication. If you want to learn more about 5G, you can look at the my product page, mathworks.com/products/5G, and watch my 5G toolbox video as well as learn all about 5G technology and how it's different from LT and all the components of a 5G [INAUDIBLE] system by learning through watching a series of 5G explained videos, each of them about 10 minutes, and there are 11 of them that are produced by my friend, Mark Bers. And you can learn about different characteristics of 5G technology in those video series.
To summarize, using MATLAB and Simulink, you can design antenna, RF, and signal processing systems in a single environment; we can generate 5G NR waveforms and perform receiver operations; the design MIMO phased arrays including complex operating structures; partition your beamforming design into hybrid structures intelligently; and use RF and baseband domains, and model MIMO channels and wireless communication systems; and explore architectural choices and trade offs. Thank you again.
You can also select a web site from the following list:
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.