Is there a future for 2D Radio Design in 5G

[henry3499]

Introduction to 5G Radio Design
The 5th generation of mobile telecommunications standards called 5G NR (New Radio) become a reality and its deployment started in many countries around the world.
Any new mobile technology can't be deployed without going through an important phase called Radio Network Design (RND). The main output of this phase are given in below points:

• The geographical location of 5G NR base station which are co-located with 4G
• The new geographical location of 5G NR base station to fill coverage and capacity gap
• Optimized Radio parameters configuration (Antenna direction, Antenna height, Tilts and radiation power).
• Number of required site and carriers to address the coverage and capacity requirements

The radio network design engineers use planning software to design and simulate the mobile network in terms of coverage, capacity and quality.
In this article, we will discus the challenges of traditional 2D radio network design approach in 5G NR network.
Design inputs
Most of current 5G NR deployment is based on existing LTE network. Below diagram shown high level design process for 5G.[/COLOR]

Other than physical data and traffic information, one of the main input for planning tools is the geographical information of the scope area, called Clutter and elevation maps. The clutter map represents the different class of the area at each pixel (location). Example of clutter class are: Dense Urban, Urban, Sub Urban and Rural but it does not contain the information of the building heights Above Ground Level (AGL) and buildings depth.
Clutter maps are used to compute the indoor prediction and generate traffic map. The user can specify penetration loss per clutter type and planning tool will apply it while calculating the received signal level. Clutter map can be used to generate traffic or user map using the clutter weight defined by the planning engineer. other information such slow and fast fading margin can be set for each clutter class.
Below Figure showing sample clutter file:

The elevation map contains the information of above mean sea level (AMSL) of each pixel for the concerned area. It represents the terrain profile. The elevation map is used by the propagation model to calculate the received signal level at each location of the area by calculating the path loss from transmitter to receiver.
Below figure is showing sample elevation file

The geographical data inputs can be provided in 2D, 2.5D or 3D format.The two-dimension (2D) format means only the clutter class and elevation map files are provided (as explained in above paragraph) and in 2.5D additional file called Clutter Height is provided (it represent the height of clutter type in reference to ground level). Finally, 3D format will include additional building information (height and the width of the buildings) and optionally the vegetation vectors is given which is very important to use while calculating the receive signal level at higher frequencies (example frequency > 28 GHz).
Below Figure is showing sample Clutter height file:

The next figure is showing sample buildings vectors with 3D view

Challenges of 2D Radio Design in 5G
Previous mobile Technologies like GSM, UMTS and LTE used two-dimension (2D) radio network design approach and the prediction used the clutter class and elevation only. The design results was satisfactory and accurate.
The 5G technology is poised to bring massive improvements compare to previous technologies in terms of capacity, low latency and reliability. The new cellular technology come-up with new architecture and support for new frequencies in mid-band (Sub-6 GHz) and high band (>28 GHz) which are not supported by LTE (4G). As well 5G NR is capable to use hundreds of antenna arrays called M-MIMO (Massive Multiple Input Multiple Output antennas) that can serve multiple users at the same time.
There are many differences between 5G and previous technologies, but in next we will focus only on differences that are making the 2D radio network design approach based is not much reliable in 5G NR radio network design.

• New spectrum: The deployment of 5G network can be built on flexible spectrum in mid bands (3.5GHz – 6GHz) and High bands (24GHz – 40GHz) spectrum. At high frequency (millimeter wave) the signal is more sensitive to obstacles and attenuation. Rain absorption loss and other loss due to diffraction, diffusion and reflection on buildings can’t be negligible at high frequencies.

This will require the use of 3D propagation model that can consider the detail propagation calculation on buildings and vegetation. The model must consider the depth and the height of each buildings. Similar case is applicable for the vegetation.

• Advance Adaptive System Antenna (AAS): It consist of many sub-element antennas integrated in one single physical antenna. Each antenna sub-elements generate narrow beams for one or multiple users multiplexed in special domain. The antenna beam is generated in the direction of the user. The system is called Massive Multiple Input Multiple Output (M-MIMO). Below figure showing sample M-MIMO antenna pattern.

In Dense Urban environment, most of the users are located inside the buildings at different floors. The AAS system generate hundreds of beams (beamforming) directed to the user locations. The interference calculation in such scenario is calculated at each bin (location) by considering 3D interference pattern calculation and the total interference pattern is then the sum of all these narrow beams’ patterns.
In M-MIMO system and by using only the ground prediction, it will be not possible to generate the coverage prediction at different floors of the building and therefore the vertical gain (Z- axis) is not captured and the interference calculation generated by the system is not accurate. The throughput (connection speed) calculation is based on total bandwidth and the amount of interference (Signal to Noise Ratio). In such conditions the throughput simulation plots are not reliable.

• Traffic Map: or user map are generated by assigning for each location of the area the number of user or the amount of traffic payload generated. The Traffic map are used by the tool Automatic Cell Planning (ACP) to optimize the network radio parameters such as antenna direction, tilts and power at multi-storey levels (buildings floor).

In mobile network, 80% of the users are indoor and for accurate planning it is recommended to create different traffic map for each floor of the building. If only 2D map is available, then only the traffic map at ground can be created. The ground level traffic map is the traffic aggregation of all users inside the buildings and the users located on the roads.
Experience shown that 30 to 40% of traffic is captured when planning using 3D traffic map.
It is very important to generate Traffic map which is covering the complete storey of the buildings. Call traces or buildings information (AGL) can be used to distribute the user among the building floor.
Below figures are showing the distribution of traffic in 2D and 3D mode:


By having different traffic distributed at each floor of the buildings, will enable accurate calculation of amount of interference and therefore optimum configuration changes provided by Automatic Cell Planning algorithm (ACP).
The ACP tilts changes will not only address the ground user but also the users located on higher floors of the buildings. It is expected that 2D optimization will focus on ground level coverage and capacity only. The coverage and throughput will be poor at high-rise buildings.
2D Vs 3D Simulation and Comparison
Multiple prediction simulation are run with 2D propagation model. 2D General Model (2D GM) is used with pre-calibrated slope and intercept per clutter type.
As buildings information are used in 3D prediction, the output coverage is an indoor prediction and in order to compare with 2D predictions, additional indoor penetration loss added for each clutter type as shown in below figure.

The Coverage, capacity and quality comparisons are performed between the 2D and 3D predictions.

• Down-link Coverage comparison

Below figure are showing the prediction comparison between both 2D and 3D prediction at ground level.
The 3D coverage is showing better signal strength on streets and open area compare to 2D predictions. Reason is that in 2D models, propagation is defined by slope and intercept (K1 & K2 factors). The slope define the attenuation of signal per decade and it is directly related to distance between the transmitter and receiver. Usually in 2D prediction we see signal attenuation as far as we go far from the site.
The 3D models attenuation is defined by a fix wall loss (A one-time penetration loss is applied at the main building wall) and pass through loss (based on the distance traveled by the wave inside the building).The 3D model concept are different from 2D propagation formula and some of 3D models use ray tracing algorithm and different diffraction method on buildings. The 2D models are mainly empirical models were signal attenuation is proportional to distance between the transmitter and receiver. The indoor prediction in 2D model is generated by applying a fix attenuation loss per clutter type but in 3D model it is calculated based on buildings wall thickness and buildings width (linear slope ex. 0.6 dB/m).
Even by tuning both the models , it is normal to have a delta coverage deviation between 2D and 3D predictions. The difference can be brought to smaller delta by increasing the buildings wall loss and the pass through loss. It is always better to use drive test as reference for comparison.

• Down-link Quality comparison

The quality or the signal to noise ratio (SNR) is the ratio between the signal of serving cells and the level of all interfere cells. The following plots are showing the 2D and 3D interference plots at 70% load of the network.

• Overall, there is 5-10% deviation between 2D and 3D quality statistic. 3D quality plot is showing more optimistic SINR compare to 2D due to better signal strength.
• Down-link Capacity comparison

As the throughput plots is directly linked to quality plots (SNR), we notice 5-10% deviation in throughput statistic between 2D and 3D models.

• Automatic Cell Planning (ACP) Comparison
• 2D Vs 3D Site Selection

The aim of site selection process is to provide the list of sites to be activated for 5G technology among the co-located sites to 4G. The first step is to duplicate all the 4G site to 5G sites and then change the 5G site configuration inline with 5G technology (spectrum, propagation model, antenna, power, frame settings,..).
[COLOR=rgba(0, 0, 0, 0.7)]The second step is to define a coverage threshold to be used for site selection. In our case -118 dBm used as SS-RSRP coverage threshold.
The site selection is run with 2D and 3D propagation models and the following figures are showing the proposed number of site for each case:


To achieve 85% of the coverage area , it require the deployment of 57 sites based on 3D SS and 62 sites based on 2D SS. The 3D SS provide ~ 8% saving in terms of site count for similar coverage target. The coverage gain remain constant post addition of 66 sites in 3D site selection and gain keep increasing by adding more site in 2D SS.

• 2D Vs 3D ACP Optimization

In ACP optimization, antenna direction , tilts are optimized for each sectors. The optimization goal is to improve the coverage SS-RSRP at threshold of -105dBm.
Below chart are showing the SS-RSRP improvement for both mode 2D and 3D


[COLOR=rgba(0, 0, 0, 0.7)]The number of change per configuration are given in below table:

We conclude following

• The number of 3D ACP changes are less compare to 2D ACP changes. The 2D ACP provide much more higher coverage gain compare to 3D ACP as the prediction didn't account the details loss on buildings.
• The 3D ACP time consumption takes more than 2 hours whereas the 2D ACP take only 10 minutes.
• The 3D ACP provide better coverage improvement on high rise buildings by directing the antenna direction and tilt into hot spot area.

Conclusion

The 2D based radio network design approach is a cost efficient solution for operators as it not required to buy buildings and vegetation vectors which are very expensive but it has limitation in high-rise buildings scenario.
The drawback of 2D based prediction are mainly the inability to optimize the network at different heights rather than the ground level. In High rise buildings scenario, it is not accurate to optimize the network using only ground level predictions as 40% of traffic is located at higher floors.
At higher frequencies (> 28 GHz), it is not accurate to use 2D based prediction due to sensitivity of signal to foliage and rain which require more sophisticated 3D models.
It is expected that 3D based optimization, will enable ACP to provide optimum coverage and throughput at high-rise buildings. The ACP will direct the antenna directions and tilts towards areas with high user density.
The accurate 3D prediction will result in optimum 5G site count proposal.The 3D site selection results in 8% in terms of site count savings compare to 2D site selection. The 3D maps cost could be compensated with the savings from site count and ACP changes.
Despite to benefits of 3D based design, the approach is time taking in terms of ACP optimization and 2D design is showing better scalability. The 2D design approach can still work well in cities with average buildings height of ~ 20m (~ 4 floor buildings) were the ground prediction and prediction at average antenna height is more or less similar.
THANKS FOR READING