Featured Examples

Derive from 3GPP standards the theoretical data rate and throughput for a 1gNB - 2UE scenario, and compare with simulation

Introduction

NetSim calculates the PHY rate per the 3GPP formula, which is explained in the infographic below.

Figure-1: 5G PHY Rate Formula in NetSim

A simple, approximate way to think of the above formula is

\[Data\ Rate = BW \times Q \times R \times N \times (1 - OH)\ \ \ \ \ \ \ \ \ \ \ \ \ldots(1)\]

where Data rate is the per carrier PHY rate, BW is the allocated bandwidth to the particular UE, Q is the modulation order, R is the code rate, N is the number of MIMO layers, and OH is the Overhead. In NetSim, OH is usually taken as 2/14 since we have 2 control symbols in a slot spanning 14 symbols.

How is the formula in (1) equivalent to the 3GPP formula? Let's examine the variables.

Q, R, N, and (1-OH) are common to both
The scaling factor in NetSim is assumed as 1
The data rate in (1) per carrier and needed to be summed up for multiple carriers

What remains to be shown is that BW is the same as N_prb * 12 / Ts. In this 12 is the number of subcarriers and Ts represents the symbol duration which varies with numerology. 10^-3 represents 1 ms, 14 is the number of OFDM symbols and 2^mu adjusts the slot duration based on the numerology. When we calculate N_prb*12/Ts, we are essentially calculating the total number of subcarriers allocated within the given symbol duration, which is nothing, but the bandwidth allocated to that device. The simplified formula we provided would yield a slightly higher estimate compared to the 3GPP formula because of the way N_prb is calculated in the standards.

When operating in TDD mode, the above computation would give the two-way (downlink + uplink) data rate. Therefore, the downlink data rate would be

\[DL - rate = Data\ Rate\ \times DL - Fraction\ \ \ \ \ \ \ \ \ \ \ \ \ldots(2)\]

While BW, OH, and N are based on user inputs in NetSim, Q and R are dependent on the modulation and coding scheme (MCS). The MCS i.e., Q and R, is chosen by looking up the the 3GPP spectral efficiency to MCS table assuming ideal Shannon rate whereby

\[Spectral - Efficiency = log2\ \left( 1 + SINR\lbrack linear\rbrack \right)\ \ \ \ldots(3)\]

The expression thus becomes

\[\frac{DL}{UL}Data\ Rate\ \lbrack Mbps\rbrack = \ BW\ \lbrack MHz\rbrack \times \ Q\ \left\lbrack \frac{bits}{symbol} \right\rbrack \times \ R\ \times \ N\ \ \times \ (1\ - \ OH) \times \left( \frac{DL}{UL}fraction \right)\ \ \ \ \ldots(4)\]

Bandwidth is the cycles per second which translates to the number of signal changes (or symbol transmissions) per second. Hence multiplying BW and Q gives Mbits/sec when BW is in MHz. The other terms are dimensionless.

Now, in 5G, the transmitter adapts its PHY layer MCS depending on the receiver's SINR. The SINR in turn depends on the received power, which is transmit-power less pathloss. In NetSim users can record the radio measurements to obtain the SINR and MCS (for each UE) over time if the channel is time varying.

Network Setup

Open NetSim, Select Examples ->5G NR ->5G data rate and throughput computation then click on the tile in the middle panel to load the example as shown in below screenshot.

Figure-2: List of 5G examples showing 5G data rate and throughput computation

Figure-3:Network setup for studying the 5G data rate and throughput computation.

Network Settings

Grid length is set to 6000m * 3000m by clicking on the grid panel on right.
Consider a single Macro cell gNB and two UEs. The distance from the gNB to the first UE should be 1900m, and the distance to the second UE should be 2100m.
Click on the gNB and expand property panel on right and set the properties in 5G RAN layer as mentioned in the below table.

Properties
Datalink Layer Properties
Scheduling type	Round Robin
Physical Layer Properties
CA type	Single band
CA Configuration	n78
CA1
Numerology	1
Channel Bandwidth	100 MHz
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Rural macro
LOS NLOS Selection	Used defined
LOS Probability	1
Shadow Fading Model	3GPP TR 38.901-7.4.1
Fast Fading Model	No Fading

Table-1: 5G RAN Datalink and Physical layer properties.

Set Transmitter and Receiver antenna count as shown in the below table.

Device	Tx Antenna Count	Rx Antenna Count
gNB	2	2
UE 1	1	1
UE 2	2	2

Table-2: Tx and Rx Antenna counts for gNB and UE

Create a CBR application for both UEs from the servers by clicking on the Set Traffic tab in the ribbon at the top. Keep the packet size at the default (1460 B) and change the inter-arrival time to 97.33 µs, thereby generating 120 Mbps of data for each UE.
Enable the LTE NR Radio Measurement log by clicking on Configure reports tab and plots.

Figure-4: Enabling the LTENR Radio Measurement log

Run the simulation for 10 seconds.

Results

Figure-5: Application metrics showing throughput results.

Open LTENR Radio measurement log from simulation results window and filter the channel to PDSCH, Layer ID to 1, To observe the MCS, CQI, Pathloss and SNR values for UE 1, filter the UE ID to UE 1

Figure-6: LTENR Radio Measurement log showing the MCS value for UE 1

Filter the UE ID to UE 2, to observe the MCS, CQI, Pathloss and SNR values for UE 2 alone.

Figure-7: LTENR Radio Measurement log showing the MCS value for UE 2

Note: The values obtained till 163 ms are during RRC association time, hence consider the values after 163 ms.

The PHY Data Rate Calculations for UE 1

\(PHY\ data\ rate(in\ Mbps) = 10^{- 6}\sum_{j = 1}^{J}{\left( v_{Layers}^{(j)} \right).Q_{m}^{(j)}.f^{(j)}.R\frac{N_{PRB}^{BW(j),\mu}.12}{T_{s}^{\mu}}}\left( 1 - {OH}^{(j)} \right)\),

Where \(T_{s}^{\mu} = \frac{10^{- 3}}{{14.2}^{\mu}}\)

For UE 1,

The number of layers \(v = 1,\ \)since the Tx and Rx antenna of UE count is 1*1, Obtained MCS is 13 which means the \(Q_{m}\) (Modulation order) is 6, \(f = 1,\) , \(N_{PRB} = 273,\ \) \(OH\) in downlink for FR1 is 0.14.

\[Data\ Rate\ \lbrack Mbps\rbrack = 10^{- 6}\left( 1\ \times 6 \times 1 \times \frac{567}{1024} \times \frac{273 \times 12}{\left( \frac{10^{- 3}}{{14.2}^{1}} \right)}\ \ \times (1 - 0.14)\ \right) = \ 262.08\ Mbps\]

This is total Data Rate which includes DL and UL. The DL data rate would be

\(DL\ Data\ rate = Data\ Rate\ (Mbps)\ \times \frac{DL}{DL + UL}\) =\(262.08\ \times \frac{4\ }{5} = 209.66\ Mbps\)

Since, we have 2 UE, with round robin resource allocation, alternate slots are allocated to each UE, therefore the PHY throughput for UE1 would be

\[PDSCH\ PHY\ Throughput\ \lbrack Mbps\rbrack = \ \frac{DL\ Data\ rate\ \lbrack Mbps\rbrack}{2} = \ \frac{209.66}{2} = 104.83\ Mbps\]

This is PHY layer throughput, and hence Application layer throughput would be

\[DL\ App\ throughput\ \lbrack Mbps\rbrack = PDSCH\ PHY\ Throughput\ \times \frac{App\ layer\ Pkt\ Size}{Phy\ layer\ Pkt\ Size}\]

The PHY Data Rate Calculations for UE 2

For UE 2,

The number of layers \(v = 2,\ \)since the Tx and Rx antenna of UE count is 2*2, Obtained MCS is 7 which means the \(Q_{m}\) (Modulation order) is 4, \(f = 1,\) , \(N_{PRB} = 273,\ \) \(OH\) in downlink for FR1 is 0.14.

\[Data\ Rate\ \lbrack Mbps\rbrack = 10^{- 6}\left( 2\ \times 4 \times 1 \times \frac{490}{1024} \times \frac{273 \times 12}{\left( \frac{10^{- 3}}{{14.2}^{1}} \right)}\ \ \times (1 - 0.14)\ \right) = \ 301.98\ Mbps\]

This is total Data Rate which includes DL and UL. The DL data rate would be

\(DL\ Data\ rate = Data\ Rate\ (Mbps)\ \times \frac{DL}{DL + UL}\) =\(301.98 \times \frac{4\ }{5} = 241.58\ Mbps\)

Since, we have 2 UE, with round robin resource allocation, alternate slots are allocated to each UE, therefore the PHY throughput for UE1 would be

\[PDSCH\ PHY\ Throughput\ \lbrack Mbps\rbrack = \ \frac{DL\ Data\ rate\ \lbrack Mbps\rbrack}{2} = \ \frac{241.58}{2} = 120.79\ Mbps\]

This is PHY layer throughput, and hence Application layer throughput would be

\[DL\ App\ throughput\ \lbrack Mbps\rbrack = PDSCH\ PHY\ Throughput\ \times \frac{App\ layer\ Pkt\ Size}{Phy\ layer\ Pkt\ Size}\]

Results and discussion

We run a simulation in NetSim per the above scenario and obtain the throughput values tabulated below.

	PHY Data Rate (Analytical throughput) in Mbps	Application throughput (Simulation) in Mbps
UE 1	101.69	95.36
UE 2	116.48	109.81

Table-3: Table showing Analytical and simulated throughput results.

UE 2 achieves higher throughput despite being further from the gNB due to 2x2 MIMO vs 1x1 MIMO UE 1.

The application layer throughput would be

\[DL\ App\ Throughput\ = \ DL\ DataRate\ \times \ (App\ Layer\ Packet\ Packet\ Size)\]

\(DL\ App\ throughput = DL\ Data\ Rate\ \times \ \frac{App\ layer\ packet\ size}{Phy\ layer\ packet\ size}\ \ \ \ \ \ldots(4)\)

The computation of the PHY layer packet size is complex. It involves various layers adding overheads: the Transport layer (UDP) contributes 8 B, and the Network Layer (IP) adds 20 B. The MAC layer introduces additional overhead, with the SDAP header contributing 1B and the PDCP header adding 16B. At this point, the packet size is the size of the application layer packet plus 45 B. The MAC layer in 5G further processes these packets, fitting them into transport blocks (TBs). These TBs are then divided into code blocks (CBs), which are grouped into code block groups (CBGs) for transmission over the air. The sizes of the TB and CB depend on various parameters, and additional overheads are incurred during this process. As a result, it's challenging to provide a simple analytical formula for PHY layer packet size. A reasonable estimate would be about 5 - 10% reduction between the PHY rate and the application throughput. This is what we observe when we compare the simulation results with the theoretical predictions in the above table.

The above discussion assumes a conservative MCS is selected, ensuring a Block Error Rate (BLER) of zero. However, if a more aggressive MCS is chosen, which typically has a higher throughput but also a higher t-BLER (e.g., 5% or 10%), the computation must account for this increased BLER.

Exercises

Explain how changing the DL:UL ratio would affect the results? Redo the theoretical calculations for DL:UL ratio of 1:1 and compare against simulation.
Change the gNB UE distances such that the MCSs seen by the UEs are different for e.g.,MCS 17 and MCS 23. Compute the theoretical throughput and compare against simulation.

Effect of distance on pathloss, SINR and MCS (7-Cell Hexagonal Layout, Urban Macro Propagation, NLOS)

The experiment aims to analyze how the performance of a UE changes as it moves away from its serving gNB. In real-world scenarios, users are mobile, and their distance from the base station affects signal strength. By simulating a linear movement of the UE away from the gNB, we can observe how parameters such as pathloss, SINR, and MCS change with distance. This helps determine the coverage limit of the gNB, identify when the signal is no longer sufficient for communication, and plan for handover decisions.

Network Layout for UE Coverage Analysis in Multi-Cell Scenario:

Figure-8: A logical diagram with then 7-cell hexagonal network topology; the UE is initially close to gNB1 and moves east towards gNB2.

The following network diagram illustrates what the NetSim UI displays when you open the example configuration file.

Figure-9:Equivalent scenario in NetSim. In this network setup to study how Pathloss, SINR, and MCS vary with distance

Settings done in example config file

Set distance between all gNB as 1500m
Set the gNB properties as follows. To configure it, click on gNB. On the right side, expand the property panel, go to property panel, go to the physical layer of the Interface (RAN) layer, and set the properties below

Properties
CA Configuration	n78
Antenna
TX Antenna Count	1
RX Antenna Count	1
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban macro
LOS NLOS Selection	User defined
LOS Probability	0
Shadow Fading Model	None
Interference Model
Downlink Interference Model	Exact geometric model

Table-4: gNB >Interface (5G RAN) >Physical layer properties

Set TX Antenna and RX Antenna as 1 in UE properties > Interface (5G RAN) > Physical Layer.
In the Device Position Properties of UE, set Mobility Model as File Based Mobility

The NetSim Mobility File (mobility.csv) is configured as follows:

#Time(s)	Device ID	X	Y
2	16	2600	1500
4	16	2700	1500
6	16	2800	1500
8	16	2900	1500
10	16	3000	1500
12	16	3100	1500
14	16	3200	1500
16	16	3300	1500
18	16	3400	1500
20	16	3500	1500

Table-5: The values set in the Mobility.csv file. The Y value remains constant while the X value is increased to configure movement towards the east.

Create a CBR application between Wired Node 8 and UE 16 from the set traffic tab in the ribbon on top. Click on the created application, and in the right-side property panel, set the transport protocol to UDP, keeping the other application properties as default.
The LTENR Radio measurement log file must be enabled from the design window.
LTENR Radio measurement Log can be enabled by clicking on the icon in Configure Reports > Plots > Network Logs option as shown below

Figure-10: Enabling LTENR Radio Measurements Log.

Run simulation for 20s, after the simulation completes Go to results window click on logs options and open LTENR Radio Measurement Log.csv and note down the Pathloss, SINR and MCS.

Figure-11: Results window

Figure-12: LTENR Radio Measurement log.csv file

Filter channel to PDSCH, and vary the distance by filtering to 100, 200, 300, 400, 500, 600, 700, 800 and 900. Record the Pathloss, SINR, and MCS values from the log file.

Results

Distance(m)	Pathloss (dB)	SINR (dB)	MCS
100	102.7656	36.2866	27
200	114.4841	24.4183	27
300	121.3573	17.2854	21
400	126.2369	12.0169	15
500	130.0228	7.6774	11
600	133.1164	3.8080	5
700	135.7322	0.1114	3
800	137.9982	-3.6469	1
900	139.9971	-7.6709	0

Table-6: Results for Pathloss vs Distance, SINR vs Distance, MCS vs Distance

As the UE moves away from gNB 10 along the defined mobility path, the signal quality gradually decreases due to increasing pathloss. Initially, at close proximity (100–200 meters), the SINR remains high, supporting a high Modulation and Coding Scheme (MCS) index of 27, which enables maximum throughput. However, as the distance increases, the signal experiences significant attenuation. By the time the UE reaches 500 meters, the SINR drops to 7.68 dB and the MCS reduces to 11, indicating reduced spectral efficiency. Beyond this point, the SINR continues to decline rapidly, becoming negative past 700 meters. At 900 meters, the SINR falls to -7.67 dB and the MCS drops to 0, meaning the UE can no longer sustain a viable communication link with gNB 10. This point effectively marks the edge of the gNB’s coverage area, beyond which the UE must perform a handover to a neighboring cell or face radio link failure.

Figure-13: Pathloss vs Distance – As the UE moves farther from the gNB, signal attenuation increases, leading to higher pathloss values.

Figure-14: SINR vs Distance – The SINR is highest when the UE is near the gNB and progressively decreases with increasing distance, reflecting the impact of signal degradation and rising interference from neighboring cell.

Figure-15: MCS vs Distance – As the UE moves farther from the gNB and signal quality declines, the MCS value decreases. This reflects the network’s adaptive modulation strategy to maintain reliable communication at lower data rates.

Figure-16: MCS vs SINR – The graph illustrates the relationship between SINR and MCS. At higher SINR the link adapts to a higher a MCS and at a lower SINR the link adapts to a lower MCS. The highest MCS per 3GPP standards in MCS 28.

CQI Interpretation and MCS Selection Using 3GPP 38.214

The 3GPP standards Spectral Efficiency vs. MCS table is used to select the appropriate MCS (Modulation and Coding Scheme). This selection can be based on the 64QAM, 256QAM, or 64QAMLOWSE table, depending on the configuration chosen by the user. In this example, we have used the 256QAM table.

The CQI (Channel Quality Indicator) indices and their corresponding interpretations are taken from 3GPP 38.214 Table 5.2.2.1-3, which defines CQI reporting for QPSK, 16QAM, 64QAM, and 256QAM.

It is recommended that users configure the same MCS table for both PDSCH (Physical Downlink Shared Channel) and PUSCH (Physical Uplink Shared Channel)

MCS	Modulation	SINR (dB)	Spectral Efficiency	CQI
27	256QAM	36.28	7.4063	15
25	256QAM	21.98	6.9141	14
23	256QAM	20.52	6.2266	13
21	256QAM	17.28	5.5547	12
19	64QAM	15.56	5.1152	11
17	64QAM	14.49	4.5234	10
15	64QAM	12.01	3.9023	9
13	64QAM	10.64	3.3223	8
11	64QAM	7.67	2.7305	7
9	16QAM	7.09	2.4063	6
7	16QAM	5.88	1.9141	5
5	16QAM	3.80	1.4766	4
3	QPSK	0.11	0.8770	3
1	QPSK	-3.64	0.3770	2
0	QPSK	-7.67	0.1523	1

Table-7: Spectral efficiency to MCS table defined in 3GPP Standards for 256 QAM.

In 5G networks, Modulation and Coding Scheme (MCS) plays a critical role in determining the spectral efficiency, which is a measure of how efficiently the available bandwidth is used to transmit data. The 3GPP specifications, particularly 38.214, define a mapping between MCS indices and their corresponding modulation orders and coding rates. Each MCS index corresponds to a specific spectral efficiency value, calculated based on the number of bits per symbol and the applied coding rate. For example, a low MCS index such as 0 uses QPSK with a low coding rate, resulting in low spectral efficiency but high robustness, while a high MCS index like 27 or 28 uses 256-QAM with a high coding rate, achieving high spectral efficiency suitable for strong channel conditions. This standardized mapping ensures that 5G systems can dynamically adapt transmission parameters to channel quality, maximizing throughput while maintaining reliability.

Effect of UE distance on throughput in FR1 and FR2

In this example we understand how the downlink UDP throughput of a UE varies as its distance from a gNB is increased. Open NetSim, Select Examples ->5G NR ->Distance vs Throughput then click on the tile in the middle panel to load the example as shown in below screenshot.

Figure-17: List of scenarios for the example of Distance vs Throughput

The following network diagram illustrates what the NetSim UI displays when you open the example configuration file.

Figure-18: Network setup for studying the Distance vs Throughput

Frequency Range - FR1

Settings done in example config file.

Set grid length as 2000m x 1000m from grid setting property panel on the right. This needs to be done before any device is placed on the grid.
Set distance between gNB 9 and UE 10 as 100m.
Click on gNB and expand the property panel on right side go to Interface (5G RAN) -> PHYSICAL LAYER, set the following properties as shown below.

Properties
CA Type	Inter band CA
CA Configuration	CA_2DL_1UL_n39_n41
CA1
Numerology	2
Channel Bandwidth	40 MHz
CA2
Numerology	2
Channel Bandwidth	100 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM64LOWSE
CSI Report Configuration
CQI Table	TABLE3
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	User Defined
LOS Probability	0
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-8: gNB >Interface (5G RAN) >Physical layer properties

Set Tx Antenna Count and Rx Antenna Count in gNB as 2 and 2.
Set Tx Antenna Count and Rx Antenna Count in UE as 2 and 2.
Go to Application properties and set the following properties as shown below.

Application Properties
Source Id	8
Destination Id	10
QoS	UGS
Transport Protocol	UDP
Packet Size	1460 Bytes
Inter Arrival time	23 μs
Start Time	1 s

Table-9: Application properties

The LTENR Radio measurement log file must be enabled from the design window.

LTENR Radio measurement Log can be enabled by clicking on the icon in Configure Reports > Plots > Network Logs option as shown below.

Figure-19: Enabling log files in NetSim GUI.

Run Simulation for 2s, after simulation completes go to metrics window and note down throughput value from application metrics.

Go back to the scenario and change the distance between gNB and UE as 200, 300, 400, 500, 600, 700, 800, 900, and 1000m and note down throughput from the results window. The other parameters in table shown below can be noted down from the LTE NR Radio measurement logs.

Frequency Range - FR2

Settings done in example config file

Set grid length as 1000m x 500m from grid setting property panel.
Set distance between gNB 9 and UE 10 as 50m.
Click on gNB and expand the property panel on right side go to Interface (5G RAN)  PHYSICAL LAYER, set the following properties as shown below.

Properties
Physical Layer Properties
CA Type		Intra Band Contiguous CA
CA Configuration		CA_n258G
	Numerology	Channel Bandwidth (MHz) per carrier	Frequency Range
CA1,CA2	3	400	FR2
Channel Model
Pathloss Model		3GPP TR 38.901-7.4.1
Shadow Fading Model		None
Fast Fading Model		No Fading
Outdoor Scenario		Urban macro
LOS NLOS Selection		User defined
LOS Probability		0
MCS Table		QAM256
CQI Table		TABLE2

Table-10: gNB >Interface (5G RAN) >Physical layer properties

Set Tx Antenna Count and Rx Antenna Count in gNB as 2 and 2.
Set Tx Antenna Count and Rx Antenna Count in UE as 2 and 2.
Go to Application properties and set the following properties as shown below.

Application Properties
Source Id	8
Destination Id	10
QoS	UGS
Transport Protocol	UDP
Packet Size	1460 Bytes
Inter Arrival time	2μs
Start Time	1s

Table-11: Application properties

The LTENR Radio measurement log file can be enabled as per the information provided above in Step 7.
Run Simulation for 1.05s, after simulation completes go to results window and note down throughput value from application metrics.

Go back to the scenario and change the distance between gNB and UE as 50, 100, 150, and 200 and note down throughput from the results window. The other parameters in the table shown below can be noted down from the LTENR Radio Measurement log.csv.

Results

Note: Filter the CC ID to 1 in the LTENR Radio measurement log file and same values have been considered in the tables given below. (SNR and CQI are shown for downlink Layer1).

Distance(m)	Pathloss (dB)	SNR (dB)	CQI Index	Modulation	Code Rate R*[1024] (MCS)	Throughput (Mbps)
100	97.33	37.46	15	64QAM	772	507.57
200	109.05	25.74	15	64QAM	772	507.57
300	115.92	18.86	15	64QAM	772	507.57
400	120.80	13.98	15	64QAM	772	453.63
500	124.59	10.20	13	64QAM	567	292.91
600	127.68	7.11	11	16QAM	616	185.82
700	130.30	4.49	10	16QAM	490	131.36
800	132.56	2.22	8	QPSK	602	80.65
900	134.56	0.22	7	QPSK	449	52.05
1000	136.35	-1.55	6	QPSK	308	36.67

Table-12: FR1 - Variation of pathloss, SNR, CQI, Modulation, code rates and throughput as the distance of the UE from the gNB is increased.

Distance(m)	Pathloss (dB)	SNR (dB)	MCS Index	Modulation	Code Rate R*[1024] (MCS)	Throughput (Mbps)
50	108.43	16.36	19	64QAM	873	2726.11
100	120.01	4.77	7	16QAM	490	1025.50
150	126.86	-2.06	1	QPSK	193	188.74
200	131.73	-6.94	0	QPSK	120	107.93

Table-13: FR 2 - Variation of pathloss, SNR, MCS, Modulation, code rates and throughput as the distance of the UE from the gNB is increased.

Increase in distance leads to an increase in pathloss, which in turn hence leads to lower received power (and lower SNR). The lower SNR leads to a lower MCS, in turn a lower CQI and thereby results in lower throughputs. The drop for FR2 happens at a much faster rate in comparison to FR1. Note that the number of information bits is obtained from the Transport Block Size Determination calculations given in 3.9.12. The throughput would depend on the TBS.

Impact of MAC Scheduling algorithms on throughput, in a multi-UE scenario

In this example we understand how the scheduling algorithm affects the UDP download throughput of a multi-user (UE) system where the UEs are at different distances from the gNB. Open NetSim, Select Examples ->5G NR ->Scheduling then click on the tile in the middle panel to load the example as shown in below screenshot

Figure-20: List of scenarios for the example of Scheduling

Multi UE throughput with UEs at different distances and channel is not time varying.

The following network diagram illustrates what the NetSim UI displays when you open this example configuration file.

Figure-21: Network set up for studying the Scheduling example.

Configuring the scheduling algorithm, and parameter settings in example config files

Set grid length as 12000m x 6000m from grid property panel on the right.
Set distance as follows.

gNB 9 to UE 10 = 1500m

gNB 9 to UE 11 = 2000m, and

gNB 9 to UE 12 = 2500m

Go to gNB properties -> Interface (5G RAN), set the following properties as shown below Table-14. In the first sample the scheduling type is set to Round Robin, in the second to Proportional fair, and in the third to Max throughput.

Properties
Datalink Layer Properties
Scheduling Type	Varies: Proportional Fair, Max throughput, Round Robin
Physical Layer Properties
CA type	Single Band
CA Configuration	n78
CA1
Numerology	1
Channel Bandwidth	100 MHz
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban macro
LOS NLOS Selection	User defined
LOS Probability	1
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-14: gNB >Interface (5G RAN) > Datalink and Physical layer properties

Set Tx Antenna Count as 1 and Rx Antenna Count as 1 in gNB properties.
Set Tx Antenna Count as 1 and Rx Antenna Count as 1 in all the UEs.
Go to the Set Traffic tab in the top ribbon and create a CBR application as shown in the table below. To change the transport protocol, QoS, and IAT, click on the application and change the properties in the right-side property panel.

Application Properties
	Application 1	Application 2	Application 3
Application Type	CBR	CBR	CBR
Source ID	8	8	8
Destination ID	10	11	12
QoS	UGS	UGS	UGS
Transport Protocol	UDP	UDP	UDP
Packet Size	1460 Bytes	1460 Bytes	1460 Bytes
Inter-arrival time	58.4 μs	58.4 μs	58.4 μs
Start Time	1s	1s	1s

Table-15: Application properties

Run Simulation for 10 s and note down throughput value in the results window in each sample. Recall that each sample has a different scheduling algorithm configured.

Results and discussions

The results with all the three UEs simultaneously downloading data is as given below.

Throughput (Mbps)
Scheduling	Application 1	Application 2	Application 3	Aggregate
Round Robin	64.65	37.21	16.65	118.51
Proportional Fair	64.65	37.22	16.64	118.51
Max Throughput	193.94	0	0	193.94

Table-16: UDP download throughputs for different scheduling algorithms when all three 3 UEs simultaneously downloading data

Figure-22: Aggregate throughput for different scheduling algorithms

Next, consider a scenario with only one of the UEs seeing DL traffic (we don’t provide inbuilt configuration file for this, and since it is a simple exercise for a user) First, run for the UE at 1500m, then for UE at 2000m and finally for UE at 2500m. This gives the maximum achievable throughput per node since the gNB resources (bandwidth) is not shared between 3 UEs and is fully dedicated to just one UE. The results are below.

Distance from gNB (m)	Application ID	Throughput (Mbps)	Remarks
1500	1	193.94	UE 1 alone has full buffer DL traffic
2000	2	111.66	UE 2 alone has full buffer DL traffic
2500	3	49.95	UE 3 alone has full buffer DL traffic

Table-17: UE throughputs if they were run standalone (without the other UEs downloading data)

The PHY rate is decided per the received SNR. Therefore, a UE closer to the gNB will get a higher data rate than a UE further away. In this example the distances from the gNB are such that UE12 Distance > UE11 Distance > UE10 Distance.

In Round Robin PRBs are allocated equally among all three nodes. However, throughputs are in the order UE10 Distance > UE11 Distance > UE12 Distance because of their distances from the gNB. The individual throughputs seen by each of the UEs is exactly 1/3 of the throughput as shown in Table-17. The PF scheduler results will match that of the RR scheduler since the channel is not time varying. In Max throughput scheduling the PRBs are allocated such that the system gets the maximum download throughput. The nearest UE will get all the resources and its throughput will be 3 times the throughput of the UE which got the max throughout in RR.

Multi UEs at different distances with a time varying channel

Configuring the scheduling algorithm, and parameter settings will remain the same for the case below.

Changes in the gNB properties are as follows.

Click on gNB and go to Interface (5G RAN), set the following properties as shown below. In the first sample the scheduling type is set to Round Robin, in the second to Proportional fair, and in the third to Max throughput.

Properties
Datalink Layer Properties
Scheduling Type	Varies: Proportional Fair, Max throughput, Round Robin
Physical Layer Properties
CA Type	Single Band
CA Configuration	n78
CA1
Numerology	1
Channel Bandwidth	100 MHz
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	User Defined
LOS Probability	1
Fast Fading Model	Rayleigh
MIMO Beamforming Model	Eigen

Table-18: gNB >Interface (5G RAN) >Physical and Datalink layer properties.

Run Simulation for 10s and note down throughput value in the results window in each sample.
Enable EWMA MAC throughput plot

Figure-23: Enabling the EWMA MAC Throughput log

Results and discussions

The results with all the three UEs simultaneously downloading data are as given below.

Throughput (Mbps)
Scheduling	Application 1	Application 2	Application 3	Aggregate
Round Robin	50.21	29.68	17.36	97.27
Proportional Fair	64.28	38.60	22.30	125.18
Max Throughput	129.87	28.77	4.68	163.32

Table-19: UDP download throughputs for different scheduling algorithms when all three 3 UEs simultaneously download data with time varying channel.

While running the Proportional fair sample enable Application throughput vs time and EWMA MAC throughput plot to observe the throughput differences.

Figure-24: Aggregate throughput for different scheduling algorithms

A difference in the performance of the RR and PF schedulers can be seen when the channel is time varying (of the order of the coherence time which is 10ms). To induce time varying randomness in the channel we enable fading and beamforming. Thus, after every 10ms, NetSim draws an i.e. fading random variable, as the additional loss. Under these conditions, the RR scheduler would allot resources to the UEs in a round robin fashion, whereas the PF scheduler would give preference to the UE which sees the best channel (highest SINR). The reason why the RR scheduler yields lower throughputs than the PF scheduler is that the RR scheduler is not “opportunistic,” i.e., it does not take advantage of the knowledge that a UE has a good channel in the next slot and continues to serve the UEs cyclically. The results are shown in Table-19; observe how this is different from Table-17 where the channel is not time varying.

Figure-25: EWMA MAC throughput stacked for UE 10, UE 11, and UE 12

In the earlier results we observed the average (overtime) throughput while in Figure-25 we observed the MAC throughput vs Time for all 3 UEs. Key points are:

Channel Coherence Time: The wireless channel fading gain changes at this time scale, causing fluctuations in the signal quality and, consequently, the achievable throughput.
Proportional Fair Scheduler Behavior: The Proportional Fair (PF) scheduling algorithm aims to balance fairness and throughput by allocating resources to users based on their current channel quality relative to their average throughput

This dynamic allocation process leads to throughput variations as the scheduler continuously adjusts resource assignments to maintain fairness while exploiting favorable channel conditions. Thus, when a user experiences good channel conditions relative to their average, they receive more resources, leading to increased throughput. Conversely, when channel conditions degrade or other users are prioritized, a user's throughput will decrease. The MAC throughput would be higher than the application throughput because of the overheads of the various layers.

Max Throughput for different MCS and CQI

Open NetSim, Select Examples ->5G NR ->Max Throughput vs MCS and CQI then click on the tile in the middle panel to load the example as shown in below screenshot.

Figure-26: List of scenarios for the example of Max Throughput vs MCS and CQI

The following network diagram illustrates what the NetSim UI displays when you open the example configuration file.

Figure-27: List of scenarios for the example of Max Throughput vs MCS and CQI

Settings done in example config file:

Set grid length as 500m x 250m from grid property panel.
Go to gNB properties  Interface (5G RAN), set the following properties as shown below Table-20.

Properties
Physical Layer Properties
CA TYPE		Intra Band Contiguous CA
CA Configuration		CA n258G
	Numerology	Channel Bandwidth (MHz)	Frequency Range
CA1	3	400		FR2
CA2	3	400		FR2
Pathloss Model		None

Table-20: gNB >Interface (5G RAN) >Physical layer properties

Go to Application properties and set the following properties as shown below Table-21.

Application Properties
Source Id	8
Destination Id	10
Transport Protocol	UDP
Start Time	1 s
Packet Size	1460 Bytes
Inter Arrival time	1 μs
Generation Rate	11680 Mbps

Table-21: Application properties

Set Tx Antenna Count as 2 and Rx Antenna Count as 1 in gNB properties.
Set Tx Antenna Count as 1 and Rx Antenna Count as 2 in UE properties.
Run Simulation for 1.002s, after simulation completes go to results window and note down throughput and delay value from application metrics.

For this Scenario set MCS Table as QAM64LOWSE and CQI Table as TABLE3 and note down throughput.

Go Back to the Scenario and set MCS Table as QAM64 and CQI Table as TABLE1 and note down throughput.

Go Back to the Scenario and set MCS Table as QAM256 and CQI Table as TABLE2 and note down throughput.

Result:

MCS Table	CQI Table	Throughput (Mbps)
QAM64LOWSE	TABLE3	2084.88
QAM64	TABLE1	2633.84
QAM256	TABLE2	3439.76

Table-22: Results Comparison.

Figure-28: Plot for Max throughput obtained for different MCS/ CQI tables

Load balancing in 5G using Cell Individual Offset (CIO)

Introduction

Overview

Mobility load balancing is a 3GPP Release 17 AI/ML for NG RAN Use Case. It involves transferring load from overloaded cells to under-loaded neighboring cells, for optimizing network performance and user experience. This study describes the network setup, presents the simulation results before and after applying CIO-based load balancing, and discusses the observed outcomes.

Concept

Default Association: The default user equipment (UE) association with a base station is based on Maximum signal strength.
Load Balancing Goal: Modify the association/handover criteria to distribute network load efficiently across available cells.

Role of Cell Individual Offset (CIO)

Cell Individual Offset (CIO) is a configurable parameter used to artificially modify the signal quality measurement of a target cell during the handover evaluation process. In NetSim, the CIO value is added to the measured SINR of the candidate cell.

\[SINR\_ eff\ = \ SINR\_ actual\ + \ CIO\]

A positive CIO increases the effective signal value.
A negative CIO decreases the effective signal value.

CIO is typically applied to control handovers and implement load balancing across cells.

Network Setup

Figure-29: Network setup for load balancing. There is a total of 500 UEs, 75 near cell, 125 mid cell and 300 cell edge UEs. Without load balancing most of the UEs would associate with the low band (1.5 GHz) gNB given the lower path loss. We use CIO for load balancing.

Network Settings

Environment size is set to 600m x 500m
Consider three Macro cell gNB sector antennas and 500 UEs spread across the network grid such that

75 UEs are located near the gNBs (represented in green)

125 UEs are located near at the cell center (represented in yellow)

300 UEs are located at the cell edge. (represented in red)

The 3 gNBs (with sector antennas) are co-located at the top left with 120°.
1. gNB 1 operates in the n50 (1.5 GHz band) b. gNB 2 operates in the n38 (2.6 GHz band) c. gNB 3 operates in the n78 (3.5 GHz band)
The gNB properties and the Traffic model properties are set as follows:

System Model and Parameters
No of gNBs	3 (3 sector carriers)
No of UEs	500
Band	n78, n50, n38
gNB Properties
Numerology	1
Channel Bandwidth (MHz)	40
Antenna	4T4R
Pathloss model	Log Distance
Pathloss Exponent (\(\eta)\)	3.8
Shadowing Model	Log Normal
Standard Deviation (dB)	5
Simulation Time (s)	10
Traffic Model
Traffic Type	Custom
Traffic generation rate	467.2 Kbps
Packet size	1460B
Inter packet arrival time	25000\(\ \mu s\) (Exponential)

Table-23: Device and Application properties

Set the Cell Individual Offset present under Interface (RAN) > Datalink Layer > Handover > Cell Individual Offset as follows:

gNB1 1475 MHz: -4.77 dB

gNB2 2595 MHz: 0 dB

gNB3 3550 MHz: 4.77 dB.

These negative CIO value shifts the load away from gNB1; the positive CIO value shifts the load towards gNB3.

Enable PRB Utilization vs time Plots and Radio resource allocation log
Run the simulation for 10 seconds.

Results

After simulation, plot the PRB Utilization for the three gNBs i.e., gNB1 1475MHz, gNB2 2595MHz, gNB3 3550MHz with CIO and without CIO.

PRB Utilization vs time plot without CIO (default configuration)

Figure-30: (Exponentially weighted moving) Average PRB Utilization of the 3 gNBs without load balancing

PRB Utilization vs time plot with CIO enabled

Figure-31: (Exponentially weighted moving) Average PRB Utilization of the 3 gNBs with load balancing

Discussion

These graphs show resource (PRB) usage of different 5G base stations over time, comparing scenarios with and without load balancing.
Without load balancing (Figure-30): Base station 1 (gNB1_1475MHz) is overloaded at ~100% while gNB2 (2575 MHz) and gNB3 (3550 MHz) are underused and operating around 30% and 25% PRB utilization respectively.
With load balancing(Figure-31): Resource usage is more evenly distributed across all base stations, with gNB1 around 60-70% and other gNBs operating at 30-40%.

We can observe the association of UEs with gNBs before and after load balancing.

To analyze this, we use the LTENRRadioMeasurementsLog.csv and DeviceList.xlsx files.

A Python script reads the UE association entries at two time points—initial and post-load balancing—and generates corresponding plots.

Figure-32: Initial UE association (left) and association after load balancing (right). The left clearly shows a higher concentration of UEs (green dots) associated with gNB1_1475MHz, which operates on the lower frequency band The right panel shows a more uniform distribution of UEs across the three gNBs due to the CIO application

Initial Association of UEs
gNB	Count of Associated UEs
gNB1 1475 MHz	321
gNB2 2595 MHz	122
gNB3 3550 MHz	57
Total	500

Table-24: Initial association count of UEs

Association of UEs after Load Balancing
gNB	Count of Associated UEs
gNB1 1475 MHz	192
gNB2 2595 MHz	105
gNB3 3550 MHz	203
Total	500

Table-25: Association count of UEs after load balancing

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import sys
import os

def get_color_map():
    return {
        'GNB1_1475MHZ': '#228B22',  # Green
        'GNB2_2595MHZ': '#FFD700',  # Yellow
        'GNB3_3550MHZ': '#d62728',  # Red
    }

def plot_combined_association(log_df, device_list_df, output_dir):
    snapshots = {
        161.5: 'UE-gNB Initial Association',
        220: 'UE-gNB Association After Load Balancing'
    }
    fig, axs = plt.subplots(1, 2, figsize=(14, 6), sharex=True, sharey=True)
    color_map = get_color_map()
    gnb_names = list(color_map.keys())

    all_x = device_list_df['X Pos/LON']
    all_y = device_list_df['Y Pos/LAT']
    x_min, x_max = all_x.min() - 20, all_x.max() + 20
    y_min, y_max = all_y.min() - 20, all_y.max() + 20

    for ax, (time_snapshot, title) in zip(axs, snapshots.items()):
        filtered_df = log_df[
            (log_df['Time(ms)'] == time_snapshot) &
            (log_df['Channel'] == 'PDSCH') &
            (log_df['isAssociated'] == True)
        ]
        merged_df = pd.merge(filtered_df, device_list_df, how='left',
                             left_on='UE Name', right_on='Device Name')

        for gnb in gnb_names:
            ue_group = merged_df[merged_df['gNB or eNB Name'] == gnb]
            ax.scatter(ue_group['X Pos/LON'], ue_group['Y Pos/LAT'],
                       label=f'{gnb} UEs', s=30, marker='o', color=color_map[gnb])

        gnb_data = device_list_df[device_list_df['Device Type'].str.contains('GNB')]
        for _, gnb in gnb_data.iterrows():
            ax.scatter(gnb['X Pos/LON'], gnb['Y Pos/LAT'],
                       label=gnb['Device Name'], s=70, marker='^', color='black')

        ax.set_title(title, fontweight='bold', fontsize=12, pad=10)
        ax.set_xlabel('X Coordinate', fontsize=10)
        ax.set_xlim(x_min, x_max)
        ax.set_ylim(y_max, y_min)
        ax.grid(True, linestyle='--', linewidth=0.5)

    axs[0].set_ylabel('Y Coordinate', fontsize=10)
    handles, labels = axs[1].get_legend_handles_labels()
    fig.legend(handles, labels, loc='lower center', fontsize=14, ncol=3, markerscale=2)

    plt.tight_layout(rect=[0, 0.08, 1, 1])
    output_path = os.path.join(output_dir, 'combined_association_plot.png')
    plt.savefig(output_path, dpi=300)
    plt.close()

    print(f"Combined plot saved: {output_path}")


if __name__ == '__main__':
    if len(sys.argv) != 4:
        print("Usage:   python   combined_association.py   <log_csv_path>   <device_list_excel_path>   <output_directory>")
        sys.exit(1)

    log_path = sys.argv[1]
    device_path = sys.argv[2]
    output_dir = sys.argv[3]

    log_df = pd.read_csv(log_path)
    device_df = pd.read_excel(device_path)

    plot_combined_association(log_df, device_df, output_dir)

4G vs. 5G: Capacity analysis for video downloads

Open NetSim, Select Examples ->5G NR -> 4G vs 5G then click on the tile in the middle panel to load the example as shown in below screenshot.

Figure-33: List of scenarios for the example of 4G vs 5G

4G

Under 4G click on 40 Nodes Sample, the following network diagram illustrates what the NetSim UI displays when you open the example configuration file.

Figure-34: Network setup for studying the 4G

Settings done in example config file:

Set grid length as 3300*5200m from grid property panel on the right.
Set the following property as shown in below given Table-26.

eNB Properties -> Interface (LTE)
CA Type	Intra Band Non-Contiguous CA
CA Configuration	CA_4DL_42C_42C_2UL_42C_BCS1
	Frequency Range		DL UL Ratio		Numerology		Channel Bandwidth
CA1	FR1	1:1		0		20 MHz
CA2	FR1	1:1		0		20 MHz
CA3	FR1	1:0		0		20 MHz
CA3	FR1	1:0		0		20 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM64
CSI Report Configuration
CQI Table	TABLE1
Channel Model
Pathloss Model	None

Table-26: eNB >Interface (LTE) >Physical layer properties

Frequency range FR1, Numerology = 0, Bandwidth = 20 MHz with QAM 64 MCS table represents a 4G configuration.
Set Uplink speed and Downlink speed as 10000 Mbps and BER as 0 in all wired links.
Set Tx Antenna Count as 2 and Rx Antenna Count as 1 in eNB > Interface LTE > Physical Layer.
Set Tx Antenna Count as 1 and Rx Antenna Count as 2 in UE > Interface LTE > Physical Layer.
Configure 40 applications with Source id as 3 and Destination id as 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43 and 44 and set the properties as shown below. This would generate 2.5 Mbps of traffic per user. Transport Protocol is set to UDP in all the applications.

Application Properties
Frame Per Sec	50
Pixel Per Frame	50000
Mu	1
Start Time	1s

Table-27: Application properties

Run simulation for 2 sec. After simulation completes go to results window and note down throughput and delay value from application metrics.

Increase the number of UE’s and number of applications as 40, 80,120,160, 200, 240 and 280 and note down throughput and delay value from application metrics.

5G

Under 5G click on 40 Nodes Sample, the following network diagram illustrates what the NetSim UI displays when you open the example configuration file.

Figure-35: Network setup for studying the 5G.

Settings done in example config file:

For the above 5G scenario set the following given properties Table-28.

gNB Properties -> Interface (5G RAN)
CA Type	Intra Band Contiguous CA
CA Configuration	CA n258G
	DL UL Ratio	Frequency Range	Numerology	Channel Bandwidth (MHz)
CA1, CA2	1:1	FR2	3	400
PDSCH and PUSCH Configuration
MCS Table	QAM256
CSI Report Configuration
CQI Table	TABLE2
Channel Model
Pathloss Model	None

Table-28: gNB >Interface (5G RAN) >Physical layer properties

The Tx Antenna Count was set to 2 and Rx Antenna Count was set to 1 in gNB > Interface 5G RAN > Physical Layer.
The Tx Antenna Count was set to 1 and Rx Antenna Count was set to 2 in UE > Interface 5G RAN > Physical Layer.
Frequency range FR2, Numerology = 3, Bandwidth = 400 MHz with QAM 256 MCS table represent a 5G configuration
The Uplink and Downlink speed was set to 10000 Mbps and BER as 0 in wired links.
Run simulation for 2 sec. After simulation completes go to results window and note down throughput and delay value from application metrics.
Increase number of UE’s and number of applications as 40, 80,120,160, 200, 240 and 280 and note down throughput and delay value from application metrics.

\[Throughput\ Per\ User\ (Mbps) = \frac{Sum\ of\ throughputs\ (Mbps)}{Number\ of\ User}\]

\[Delay\ Per\ User\ (\mu s) = \frac{Sum\ of\ Delays\ (\mu s)}{Number\ of\ User}\]

Theoretical PHY Rate Calculation

The 4G/5G PHY data rate is given by the expression

\(PHY\ data\ rate(in\ Mbps) = 10^{- 6}\sum_{j = 1}^{J}{\left( v_{Layers}^{(j)} \right).Q_{m}^{(j)}.f^{(j)}.R\frac{N_{PRB}^{BW(j),\mu}.12}{T_{s}^{\mu}}}\left( 1 - {OH}^{(j)} \right)\),

Where \(T_{s}^{\mu} = \frac{10^{- 3}}{{14.2}^{\mu}}\)

This expression gives the PHY rate; the application throughput would be lower than the PHY rate given the overheads in the various layers.

4G:

Number of carriers: 4, Number of layers = \(Min\left( N_{t}\ (gNB),\ N_{r}(UE) \right) = Min\ (2,\ 2) = 2\), Numerology: 0. The BW per carrier is 20 MHz, with each carrier having 100 PRBs. In this experiment settings the DL:UL Ratio is 1:1 for 2 carriers and 1:0 for 2 carriers. Avg DL: UL ratio is 3:1 and hence the DL fraction = \(\frac{3}{3 + 1} = \frac{3}{4}.\ \)

Applying the 4G PHY data rate formula

\[PHY\ Rate\ = 10^{- 6}\left( 2\ \times 6 \times 1 \times \frac{948}{1024} \times \frac{100 \times 12}{\left( \frac{10^{- 3}}{{14.2}^{0}} \right)}\ \ \times (1 - 0.25)\ \right) \times 4 = 559.40\ Mbps\]

Where \(4\) is the number of carriers

Multiplying by the DL fraction we obtain the downlink PHY rate as\(\ 559.40\ \times \ \frac{3}{4}\) = \(419.93\ Mbps\)

5G:

Number of carriers: 4, Number of layers = \(Min\left( N_{t}\ (gNB),\ N_{r}(UE) \right) = Min\ (2,\ 2) = 2\), Numerology: 3. The BW per carrier is 400 MHz with each carrier having 264 PRBs. In this experiment settings, the DL:UL Ratio: 1:1 for both carriers

Applying the 5G PHY data rate formula, we get

\[PHY\ Rate = 10^{- 6}\left( 2\ \times 8 \times 1 \times \frac{948}{1024} \times \frac{264 \times 12}{\left( \frac{10^{- 3}}{{14.2}^{3}} \right)}\ \ \times (1 - 0.18)\ \right) \times 2 = \ 7568.22\ Mbps\]

Where \(2\) is the number of carriers

Multiplying by the DL fraction we obtain the downlink PHY rate as\(\ 7568.22\ \times \ \frac{1}{2} = 3784.11\ Mbps\).

We vary the UE count from 40 to 280 in steps of 40. Each UE is downloading video at a rate of 2.5 Mbps. Post simulation, we plot the throughput per UE for 4G and 5G as the UE count is increased from 40 to 280.

Results:

Number of Users	4G (Devices downloading video)			5G (Devices downloading video)
	Throughput (Mbps)		Delay (μs)	Throughput (Mbps)			Delay (μs)
	Per user	Aggregate	Average delay	Per user		Aggregate		Average delay
40	2.43	97.23	3113.92	2.45	98.08		392.12
80	2.44	195.41	5539.42	2.44	195.69		649.40
120	2.44	293.13	7965.39	2.44	293.29		908.43
160	2.41	386.24	10175.24	2.45	392.18		1171.72
200	2.06	412.60	82771.02	2.44	489.94		1430.69
240	1.71	412.70	152826.58	2.44	587.86		1689.47
280	1.47	412.36	202854.93	2.44	685.70		1946.88

Table-29: Aggregated and Average throughput and delay per user with different number of users for LTE 4G and 5G NR

In the earlier section, we had predicted a PHY rate of 419 Mbps. We observe that the aggregate application throughput of 4G saturates at 407 Mbps. The \(\approx 10\%\ \)difference is due to the overheads in the various layers. The required rate for each video application is \(\approx 2.5\ \)Mbps and we see that 4G is able to support full rate for upto 160 UEs. At 200 UEs the capacity required is \(\approx 200 \times 2.5 = 500\ Mbps\ \)which is more than what is available. Therefore, the rate per user starts decreasing for UE counts of 200, 240 and 280. In line with this, we see that the average delay increases exponentially from 200 UEs onwards.

On the other hand, 5G can handle a PHY rate of \(3784\) Mbps or \(\approx 3400\) Mbps of application throughput. Thus, we see that 5G is easily able to provide full rate of \(\approx 2.5\) Mbps to each UE even when the total UE count is 280.

Throughput per user vs. Number of users for 4G and for 5G

Figure-36: Throughput per user vs Number of Devices for 4G and 5G. The 4G per user throughput starts falling after 160 devices.

Average delay vs. Number of users for 4G and for 5G

Figure-37: Delay vs Number of Devices. The 5G Network average delay is insignificant i.e., many orders of magnitude lower, and hence not visible in the plot.

5G-Peak-Throughput

Open NetSim, Select Examples ->5G NR -> 5G Peak Throughput then click on the tile in the middle panel to load the example as shown in below screenshot

Figure-38: List of scenarios for the example of 5G Peak Throughput

3.5 GHz n78 band

The following network diagram illustrates what the NetSim UI displays on clicking.

Figure-39: Network setup for studying the 5G Peak Throughput

Settings done in example config file:

Set the following property as shown in below given Table-30.

gNB Properties -> Interface (5G RAN)
CA Type	Single Band
CA Configuration	n78
CA1
Frequency Range	FR1
DL/UL Ratio	4:1
Numerology	2
Channel Bandwidth	50 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM256
CSI Report Configuration
CQI Table	TABLE2
Channel Model
Pathloss Model	None

Table-30: gNB >Interface (5G RAN) >Physical layer properties

The Tx Antenna Count was set to 8 and Rx Antenna Count was set to 4 in gNB > Interface 5G RAN > Physical Layer.
The Tx Antenna Count was set to 4 and Rx Antenna Count was set to 8 in UE > Interface 5G RAN > Physical Layer.
Set 2 applications Downlink source node as 8, and destination node as 10, Uplink source node as 10, and destination node as 8. Transport Protocol is set to UDP in all the applications.

Application Properties
App CBR UDP DL
Start Time (s)	1
Packet Size (Byte)	1460
Inter Arrival Time (µs)	2.92
App CBR UDP UL
Start Time (s)	1
Packet Size (Byte)	1460
Inter Arrival Time (µs)	5.84

Table-31: Application properties

Enable the Throughput vs time plot under Application and link and run simulation for 1.1 sec. After simulation completes go to results window and note down throughput value from application metrics.

Go back to the Scenario and change channel bandwidth to 100 MHz, run simulation for 1.1 sec and note down throughput value from application metrics.

Result:

Bandwidth (MHz)

Throughput (Mbps)

CBR UDP UL

Throughput (Mbps)

CBR UDP DL

DL/UL Ratio of 4:1, with 8 DL MIMO and 4 UL MIMO layers

50

128.71

1597.47

100

270.62

3366.87

Table-32: Results Comparison

Figure-40: Plot for DL and UL throughput for 50 and 100 MHz bandwidth of 3.5 GHz and n78 band

26 GHz n258 band

The following network diagram illustrates what the NetSim UI displays on clicking.

Figure-41: Network setup for studying the 5G Peak Throughput

Settings done in example config file:

Set the following property as shown in below Table-33.

gNB Properties -> Interface (5G RAN)
CA Type	Single Band
CA Configuration	n258
Component Carrier 1
DL/UL Ratio	4:1
Frequency Range	FR2
Numerology	3
Channel Bandwidth	200 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM256
CSI Report Configuration
CQI Table	TABLE2
Channel Model
Pathloss Model	None

Table-33: gNB >Interface (5G RAN) >Physical layer properties

The Tx Antenna Count was set to 8 and Rx Antenna Count was set to 4 in gNB > Interface 5G RAN > Physical Layer.
The Tx Antenna Count was set to 4 and Rx Antenna Count was set to 8 in UE > Interface 5G RAN > Physical Layer.
Set 2 applications Downlink source node as 8 destination node as 10, Uplink source node as 10 destination node as 8. Transport Protocol is set to UDP in all the applications.

Application Properties
App CBR UDP DL
Start Time (s)	1
Packet Size (Byte)	1460
Inter Arrival Time (µs)	1
App CBR UDP UL
Start Time (s)	1
Packet Size (Byte)	1460
Inter Arrival Time (µs)	4

Table-34: Application properties

After simulation completes go to results window and note down throughput value from application metrics.

Go back to the Scenario and change channel bandwidth to 400 MHz, run simulation for 1.1 sec and note down throughput value from application metrics.

Result:

Bandwidth (MHz)

Throughput (Mbps)

CBR UDP UL

Throughput (Mbps)

CBR UDP DL

DL/UL Ratio of 4:1, with 8 DL MIMO and 4 UL MIMO layers

200

518.35

6283.72

400

1041.15

11648.46

Table-35: Results Comparison

Figure-42: Plot for DL and UL throughput for 200 and 400 MHz bandwidth of 26 GHz and n258 band

Impact of distance on throughput for n261 band in LOS and NLOS states

Objective: We observe throughput of a UE (operating in the n261 band with a channel bandwidth of 100 MHz), moving away from the gNB from 1m to 3.5 Km. The variation of throughput is plotted in both LOS and NLOS states. Since 5G simulations take a long time to complete, and given our goal of studying throughput vs. distance, we have set an unrealistic speed of 20m every 10ms to complete the UE movement in a short time duration.

Open NetSim, Select Examples ->5G NR -> Distance vs Throughput n261 band then click on the tile in the middle panel to load the example as shown in below Figure-43.

Figure-43: List of scenarios for the example of Distance vs Throughput n261 band

Figure-44: Network setup for studying the Distance vs Throughput n261 band

DL: UL Ratio 4:1

LOS and NLOS

The following settings were done to generate this sample:

Step 1: A network scenario is designed in NetSim GUI consisting of 1 gNB, 5G-Core, and 1 UE and 1 Router and 1 Wired Node in the “5G NR” Network Library.

Step 2: Grid length was set to 8000 m x 4000 m.

Step 3: The device positions are set as per the table given below.

Device	UE_10	gNB_9
x- axis	500	500
y- axis	1	0

Table-36: Device position properties

Step 4: The following properties were set in Interface (5G RAN) of gNB

Parameter	Value
Tx Power	40
gNB Height	10m
CA Type	Single Band
CA Configuration	n261
Component Carrier 1
DL-UL Ratio	4:1
Numerology	3
Channel Bandwidth	100 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM64LOWSE
CSI Report Configuration
CQI Table	TABLE3
Channel Model
Pathloss Model	3GPPTR38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	User Defined
LOS Probability	1
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-37: gNB >Interface (5G RAN) >Physical layer properties

Step 5: Set Tx Antenna Count and Rx Antenna Count as 2 and 2 in gNB properties > Interface(5G RAN) > Physical Layer.

Step 6: Set Tx Antenna Count and Rx Antenna Count as 2 and 2 in UE properties > Interface(5G RAN) > Physical Layer.

Step 7: Two CBR Applications were generated from between the Server 8 and UE 10 with the following values.

Parameter	Value
APP1 CBR DL
Source	Server 8
Destination	UE 10
Start Time (s)	1
Packet Size (Bytes)	1460
IAT (µs)	11.68
Generation Rate (Mbps)	1000
Transport Protocol	UDP
APP2 CBR UL
Source	UE 10
Destination	Server 8
Start Time (s)	1
Packet Size (Bytes)	1460
IAT (µs)	97.33
Generation Rate (Mbps)	120
Transport Protocol	UDP

Table-38: Application Properties

Step 8: In the Device Position Properties of UE 10, set Mobility Model as File Based Mobility

File Based Mobility: In File Based Mobility, users can write their own custom mobility models and define the movement of mobile users. Create a mobility.csv file for UE’s involved in mobility with each step equal to 4 sec with distance 100 m. The NetSim Mobility File (mobility.csv) format is as follows:

#Time(s)	Device ID	X	Y	Z
1	10	500	50	0
1.01	10	500	70	0
1.02	10	500	90	0
1.03	10	500	110	0
.	.	.	.	.
2.65	10	500	3350	0
2.66	10	500	3370	0
2.67	10	500	3390	0
2.68	10	500	3410	0
2.69	10	500	3430	0
2.7	10	500	3450	0
2.71	10	500	3470	0
2.72	10	500	3490	0
2.73	10	500	3510	0

Table-39: Mobility.csv file

Step 9: Enable application throughput vs time plot under Plots tab in the NetSim GUI.

Step 10: Run simulation for 2.75s.

Step 11: Similarly, in LOS, set the LOS Probability to 0 in gNB properties and simulate the scenario for 2.75s.

Results:

Downlink Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) Plots

Figure-45: Downlink Application Throughput Plot in LOS mode.

Figure-46: Downlink Application Throughput Plot in NLOS mode.

Uplink Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) Plots

Figure-47: Uplink Application Throughput Plot in LOS mode.

Figure-48: Uplink Application Throughput Plot in NLOS mode.

Discussion: The downlink throughput of 478.1 Mbps is maintained till ~450m in LOS whereas, it is maintained till 150m in NLOS. Similarly, the uplink throughput of 114 Mbps is maintained till 150m in LOS whereas, it is maintained till 130m in NLOS. The Uplink throughput falls to the lowest level at ~750m in LOS and at ~150m in NLOS.

DL: UL Ratio 3:2

LOS and NLOS

Step 1: All the properties were set as in DL: UL-Ratio 4:1.

Step 2: In the gNB properties-> Interface 5G RAN, the DL:UL ratio was set to 3:2.

Step 3: The following settings were done in application properties:

Parameter	Value
APP1 CBR DL
Source	Server 8
Destination	UE 10
Start Time (s)	1
Packet Size (Bytes)	1460
IAT (µs)	11.68
Generation Rate (Mbps)	1000
Transport Protocol	UDP
APP2 CBR UL
Source	UE 10
Destination	Server 8
Start Time (s)	1
Packet Size (Bytes)	1460
IAT (µs)	38.93
Generation Rate (Mbps)	300
Transport Protocol	UDP

Table-40: Application Properties

Step 3: Run simulation for 2.75s.

Step 4: Similarly, in LOS, set the LOS Probability to 0 in gNB properties and run simulation for 2.75s.

Results:

Downlink Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) Plots

Figur-49: Downlink Application Throughput Plot in LOS mode.

Figure-50: Downlink Application Throughput Plot in NLOS mode.

Uplink Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) Plots

Figure-51: Uplink Application Throughput Plot in LOS mode

Figure-52: Uplink Application Throughput Plot in NLOS mode

Inference: The downlink throughput of 359.74 Mbps is maintained till ~550m in LOS whereas, it is maintained till 150m in NLOS. Similarly, the uplink throughput of 120 Mbps is maintained till 170m in LOS whereas, it is 35.97 Mbps maintained till 130m in NLOS. The Uplink throughput falls to the lowest level at ~750m in LOS and at ~150m in NLOS.

gNB cell radius for different data rates

Open NetSim, Select Examples->5G NR ->gNB cell radius for different data rates then click on the tile in the middle panel to load the example as shown in below screenshot

Figure-53: List of scenarios for the example of gNB cell radius for different data rates

3.5 GHz n78 urban gNB cell radius for different data rates

The following network diagram illustrates what the NetSim UI displays on clicking.

Figure-54: Network setup for studying the gNB cell radius for different data rates.

Setting done in example config file:

Set the following property as shown in below Table-41.

gNB Properties -> Interface (5G RAN)
gNB Height	10m
Tx Power	40
CA Type	Single Band
CA Configuration	n78
Component Carrier 1
DL: UL	4:1
Numerology	2
Channel Bandwidth	50 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM256
CSI Report Configuration
CQI Table	TABLE2
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	3GPP TR 38.901-Table7.4.2-1
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-41: gNB >Interface (5G RAN) >Physical layer properties

Set the Tx Antenna Count as 8 and Rx Antenna Count as 1 in gNB> Interface 5G RAN > Physical Layer.
Set the Tx Antenna Count as 1 and Rx Antenna Count as 8 in UE> Interface 5G RAN > Physical Layer.
Set the following application properties:

App_1_CBR
Source Id	8
Destination Id	10
Packet Size	1460
IAT	1.94 µs
Start time	1s
Transport Protocol	UDP
Generation Rate	6 Gbps

Table-42: Application properties

Run simulation for 1.1 sec. After simulation completes go to results window and note down throughput value from application metrics.

Go back to the Scenario and change distance between gNB and UE to 100m, 130m, 150m, 170m, 190m, 200m, 300m, 330m, and 350m and run simulation for 1.1 sec.

Result:

Cell Radius (m)	Data Rate (Mbps). Downlink
\(\mathbf{\approx}\)1500 Mbps Downlink
100	1597.47
130	1355.11
150	1222.89
\(\mathbf{\approx}\)1000 Mbps Downlink
170	1112.75
190	969.44
200	837.22
\(\mathbf{\approx}\)500 Mbps Downlink
300	506.79
330	418.61
350	308.46

Table-43: Results Comparison

26 GHz n258 urban gNB cell radius for different data rates

Setting done in example config file:

Set the following property as shown in below given table:

gNB Properties -> Interface (5G RAN)
gNB Height	10m
Tx Power	40
CA Type	Single Band
CA Configuration	n258
Component Carrier 1
DL: UL	4:1
Numerology	2
Channel Bandwidth	200 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM256
CSI Report Configuration
CQI Table	TABLE2
Channel Model
Pathloss Model	3GPPTR38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS_NLOS Selection	3GPPTR38.901-Table7.4.2-1
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-44: gNB >Interface (5G RAN) >Physical layer properties

Set the Tx Antenna Count as 8 and Rx Antenna Count as 1 in gNB> Interface 5G RAN > Physical Layer.
Set the Tx Antenna Count as 1 and Rx Antenna Count as 8 in UE> Interface 5G RAN > Physical Layer.
Set the following application properties:

App 1 CBR
Source Id	8
Destination Id	10
Packet Size	1460
IAT	1.94 µs
Start time	1s
Transport Protocol	UDP
Generation Rate	6 Gbps

Table-45: Application properties

Run simulation for 1.1 sec. After simulation completes go to results window and note down throughput value from application metrics.

Go back to the Scenario and change distance between gNB and UE to 20m, 110m, and 150m and run simulation for 1.1 sec.

Result:

Cell Radius (m)	Data Rate (Mbps). Downlink
\(\mathbf{\approx}\)6000 Mbps Downlink
20	5989.03
\(\mathbf{\approx}\)1000 Mbps Downlink
110	735.48
\(\mathbf{\approx 5}\)00 Mbps Downlink
150	302.86

Table-46: Results Comparison

Impact of numerology on a RAN with phones, sensors, and cameras

Open NetSim, Select Examples ->5G NR -> Impact of numerology on a RAN with phones sensors and cameras then click on the tile in the middle panel to load the example as shown in below Figure-55.

Figure-55: List of scenarios for the example of Impact of numerology on a RAN with phones sensors and cameras

Network Scenario: To model a real-world scenario, we base our simulation on the setup shown in Figure-56. The link between the gNB and the L2_Switches that represents the Core Network (CN) is made with a point-to-point 10 Gb/s link, without propagation delay. The Radio Area Network (RAN) is served by 1 gNB, in which different UEs share the connectivity. We have 25 smartphones, 6 sensors, 3 IP cameras. The bandwidth is 100MHz and Round Robin MAC Scheduler. The position of the devices in the reference scenario depicted in Figure-56 is quasi-random.

Figure-56: Network setup for studying with 25 smartphones, 6 sensors and 3 cameras communicating with respective cloud servers.

In terms of application data traffic, the camera (video) and sensor nodes have one UDP flow each, that goes in the UL towards a remote node on the Internet. These flows are fixed-rate flows: we have a continuous transmission of 5 Mb/s for the video nodes, to simulate a 720p24 HD video, and the sensors transmit a payload of 500 bytes each 2.5 ms, that gives a rate of 1.6 Mb/s. For smartphones, we use TCP as the transmission protocol. These connect to database servers. Each phone has to download a 25 MB file and to upload one file of 1.5 MB. These flows start at different times: the upload starts at a random time between the 25^th and the 75^th simulation seconds, while each download starts at a random time between the 1.5^th and the 95^th simulation seconds.

	Flows (No of devices)	Traffic Rate (Mbps)	Segment / File Size (B)	RAN Dir.	TCP ACK Dir.
Camera (UDP)	3	5	500	UL
Sensor (UDP)	6	1.6	500	UL
Smartphone Upload (TCP)	25		1,500,000	UL	DL
Smartphone Download (TCP)	25		25,000,000	DL	UL

Table-47: Various parameters of the Traffic flow models for all the devices

The numerology \(\ \mu\) can take values from 0 to 3 and specifies an SCS of \(15 \times 2^{\mu}\) kHz and a slot length of \(\frac{1}{2^{\mu}}\) ms. FR1 support \(\mu = 0,\ 1\) and \(2\), while FR2 supports \(\mu = 2,\ 3.\ \)We study the impact of different numerologies, and how they affect the end-to-end performance. The metrics measured and analyzed are a) Throughput of TCP uploads & downloads, and b) Latency of the UDP uploads.

Settings done in example config file:

For the above scenario set the following given properties:

gNB Properties -> Interface (5G RAN)
Pathloss Model	None
CA Type	Inter Band CA
CA Configuration	CA_2DL_2UL_n40_n41
CA1
DL UL Ratio	1:4
Frequency Range	FR1
Numerology	0, 1, and 2
Channel Bandwidth	50 MHz
CA2
DL UL Ratio	1:4
Frequency Range	FR1
Numerology	0, 1, and 2
Channel Bandwidth	50 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM64
CSI Report Configuration
CQI Table	TABLE1
Channel Model
Pathloss Model	None

Table-48: gNB >Interface (5G RAN) >Physical layer properties

The following Application properties set to the above scenario:

Sensor UL UDP
Generation Rate (Mbps)	1.6
Transport Protocol	UDP
Application Type	Custom
Packet Size (Bytes)	500
Inter Arrival Time (μs)	2500

Table-49: Sensor Application Properties for UL UDP

Camera UL UDP
Generation Rate (Mbps)	5
Transport Protocol	UDP
Application Type	Custom
Packet Size (Bytes)	500
Inter Arrival Time (μs)	800

Table-50: Camera Application Properties for UL UDP.

Phone DL TCP
Transport Protocol	TCP
Start Time (s)	\(1.5\ + 4(i),\ \)Where, \(i = 0,\ 1,\ 2,\ldots\ldots,\ 48\)
Stop Time (s)	95
File Size (Bytes)	25,000,000
Inter Arrival Time (s)	200 (Simulation ends at 110s and hence only one file is sent)
Application Type	FTP

Table-51: Phone Application Properties for DL TCP

Phone UL TCP
Application Type	FTP
Transport Protocol	TCP
Start Time (s)	\[4.5 + 4(i - 1)\\] Where, \[i = 1,\ 2,\ldots\ldots,\ 25\]
Stop Time (s)	100
File Size (Bytes)	1,500,000
Inter Arrival Time (s)	200 (Simulation ends at 110s and hence only one file is sent)

Table-52: Phone Application Properties for UL TCP

The Tx Antenna Count was set to 2 and Rx Antenna Count was set to 4 in gNB > Interface 5G RAN >Physical Layer.
The Tx Antenna Count was set to 4 and Rx Antenna Count was set to 2 in UE > Interface 5G RAN >Physical Layer.
Run simulation for 110 sec. After simulation completes go to results window and note down throughput and delay value from application metrics.

Result and Analysis:

Numerology(μ) = 0
Camera Uplink		Sensor Uplink		Smartphone
Camera Uplink		Sensor Uplink		Downlink	Uplink
Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Throughput (Mbps)
5.00	1801.94	1.60	2252.64	101.68	86.32
5.00	1803.13	1.60	2254.25	101.68	86.32
5.00	1804.67	1.60	2256.92	101.68	86.32
		1.60	2256.38	101.68	86.32
		1.60	2257.37	101.68	86.32
		1.60	2254.70	101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32
				101.68	86.32

Table-53: Throughput and delay for Camera, Sensors and Smartphones, when μ=0

Numerology(μ) = 1
Camera Uplink		Sensor Uplink		Smartphone
Camera Uplink		Sensor Uplink		Downlink	Uplink
Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Throughput (Mbps)
5.00	901.86	1.60	1502.24	156.43	172.61
5.00	902.57	1.60	1503.20	156.43	172.61
5.00	903.47	1.60	1505.04	156.43	172.61
		1.60	1504.59	156.43	172.61
		1.60	1505.48	156.43	172.61
		1.60	1503.65	156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61
				156.43	172.61

Table-54: Throughput and delay for Camera, Sensors and Smartphones, when μ=1

Numerology(μ) = 2
Camera Uplink		Sensor Uplink		Smartphone
Camera Uplink		Sensor Uplink		Downlink	Uplink
Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Delay (μs)	Throughput (Mbps)	Throughput (Mbps)
5.00	451.78	1.60	751.98	151.75	345.10
5.00	453.33	1.60	752.68	151.75	345.10
5.00	456.98	1.60	754.26	151.75	345.10
		1.60	753.81	151.75	345.10
		1.60	754.71	151.75	345.10
		1.60	753.12	151.75	345.10
				151.75	345.10
				151.75	345.10
				151.75	345.1
				151.75	345.1
				151.75	345.1
				151.75	345.1
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.11
				151.75	345.10
				151.75	345.10
				151.75	345.10
				151.75	345.10

Table-55: Throughput and delay for Camera, Sensors and Smartphones, when μ=2

Figure-57: The average uplink throughput for camera and sensors remains the same as numerology is increased. This is because the flow is UDP.

Figure-58: Smartphone Uplink, and Smartphone Downlink average throughput vs. Numerology (µ)

Figure-59: Camera Uplink, and Sensor Uplink Latency vs. Numerology. The latency drops as the numerology increases

For UDP applications the μ does not impact the throughput. This is because throughput of UDP over 5G only depends on the "capacity" of the OFDM time-frequency grid. Changing the numerology does not change the OFDM capacity, given the inverse relationship between subcarrier spacing and numerology. However, higher μ leads to an obviously lower delay. The variation of delay vs. μ is as follows:

	Avg Delay (Camera)	Avg Delay (Sensor)
\[\mathbf{\mu = 0}\]	1.838 ms	2.26 ms
\[\mathbf{\mu = 1}\]	0.930 ms	1.51 ms
\[\mathbf{\mu = 2}\]	0.476 ms	0.75 ms

Table-56: Variation of delay vs. numerology for Camera and Sensors

The TCP throughput is inversely proportional to round trip time. Therefore, for applications running over TCP the throughput increases with higher numerology. This is because higher Numerology leads to reduced round-trip (end-to-end) times.

Impact of UE movement on Throughput

Open NetSim, Select Examples ->5G NR -> UE Movement vs Throughput then click on the tile in the middle panel to load the example as shown in below Figure-60.

Figure-60: List of scenarios for the example of UE Movement vs Throughput

NetSim UI displays the configuration file corresponding to this experiment as shown below in Figure-61

Figure-61: Network setup for studying Throughput vs. UE Movement

The following set of procedures were done to generate this sample:

Step 1: A network scenario is designed in NetSim GUI consisting of 1 gNB, 5G-Core, and 1 UE and 1 Wired Node in the “5G NR” Network Library.

Step 2: Grid Length was set to 7000 m x 3500 m.

Step 3: The device positions are set as per the table given below Table-57.

Device	UE_10	gNB_9
x- axis	500	500
y- axis	600	0

Table-57: Device general properties

Step 4: The following properties were set in Interface (5G RAN) of gNB

Parameter	Value
Tx Power	40
gNB Height	10m
CA Type	Single Band
CA Configuration	n78
Component Carrier 1
DL-UL Ratio	4:1
Numerology	0
Channel Bandwidth	10 MHz
PDSCH and PUSCH Configuration
MCS Table	QAM64LOWSE
CSI Report Configuration
CQI Table	TABLE3
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Propagation Model	Urban Macro
LOS NLOS Selection	User Defined
LOS Probability	0
Shadow Fading Model	None
Fast Fading Model	No Fading

Table-58: gNB >Interface (5G RAN) >Physical layer properties

Step 5: Set Tx Antenna Count and Rx Antenna Count as 2 and 1 in gNB properties > Interface(5G RAN) > Physical Layer.

Step 6: Set Tx Antenna Count and Rx Antenna Count as 1 and 2 in UE properties > Interface(5G RAN) > Physical Layer.

Step 7: In the Position Properties of UE 8, set Mobility Model as File Based Mobility

File Based Mobility: In File Based Mobility, users can write their own custom mobility models and define the movement of the mobile users. Create a mobility.csv file for UE’s involved in mobility with each step equal to 4 sec with distance 100 m.

The NetSim Mobility File (mobility.csv) format is as follows:

Figure-62: mobility.csv file.

Step 8: A CBR Application was generated from set traffic tab in top ribbon between Wired node and UE 10 (Source as Server and destination as UE) with Packet Size of 1460 Bytes and Inter Arrival Time of 1168 µs.

Step 10: The Transport Protocol is set to UDP. Additionally, the “Start Time(s)” parameter is set to 1s. To configure it, click on created application, change the properties accordingly in the right-side property panel.

Step 11: Application throughput vs time plot under Application and Link performance is enabled from the configure reports tab in plots tab in the NetSim GUI. Additionally, LTE Radio measurements log is enabled for detailed analysis.

Step 12: Run simulation for 105s.

Results:

Figure-63: Plot of Throughput (Mbps) vs Time (sec).

Discussion

As the UE moves away from the gNB, the Application throughput starts reducing. The maximum throughput of 10 Mbps is obtained until 11.9 sec. At 16s the UE is 1000m away from the gNB, then the throughput drops to 6.30 Mbps and at time 36.6 sec (when UE is 1800m away from gNB), the throughput drops to 1.86 Mbps and subsequently keeps dropping as till the end of the simulation as the UE continues to move further away from the gNB.

Simulate and study the 5G Handover procedure.

Introduction

The handover logic of NetSim 5G library is based on the Strongest Adjacent Cell Handover Algorithm (Ref: Handover within 3GPP LTE: Design Principles and Performance. Konstantinos Dimou. Ericsson Research). The algorithm enables each UE to connect to that gNB which provides the highest Reference Signal Received Power (RSRP). Therefore, a handover occurs the moment a better gNB (adjacent cell has offset stronger RSRP, measured as SNR in NetSim) is detected.

This algorithm is similar to 38.331, 5.5.4.4 Event A3 wherein Neighbor cell’s RSRP becomes Offset better than serving cell’s RSRP. Note that in NetSim report-type is periodical and not event Triggered since NetSim is a discrete event simulator and not a continuous time simulator.

This algorithm is susceptible to ping-pong handovers; continuous handovers between the serving and adjacent cells on account of changes in RSRP due mobility and shadow-fading. At one instant the adjacent cell's RSRP could be higher and the very next it could be the original serving cell's RSRP, and so on.

To solve this problem the algorithm uses:

Hysteresis (Hand-over-margin, HOM) which adds a RSRP threshold (Adjacent cell RSRP – Serving cell RSRP > Hand-over-margin or hysteresis), and

Time-to-trigger (TTT) which adds a time threshold.

This HOM is part of NetSim implementation while TTT can be implemented as a custom project in NetSim.

Network Setup

Open NetSim and click on Examples> 5G NR> Handover in 5GNR> Handover Algorithm then click on the tile in the middle panel to load the example as shown in below Figure-64.

Figure-64: List of scenarios for the example of Handover in 5GNR

Handover Algorithm

NetSim UI displays the configuration file corresponding to this experiment as shown below Figure-65.

Figure-65: Network setup for studying the 5G handover

Procedure for 5G Handover

The following set of procedures were done to generate this sample:

Step 1: A network scenario is designed in NetSim GUI consisting of 5G-Core devices, 2 gNBs, and 1 UE in the “5G NR” Network Library.

Step 2: The device positions are set as per the table given below Table-59.

	gNB 7	gNB 8	UE 9
X Coordinate	500	4500	500
Y Coordinate	1500	1500	3000

Table-59: Device positions

Step 3: In the Position properties of UE 9, set Mobility Model as File Based Mobility.

File Based Mobility:

In File Based Mobility, users can write their own custom mobility models and define the movement of the mobile users. Create a mobility.csv file for UE’s involved in mobility with each step equal to 0.5 sec with distance 50 m.

The NetSim Mobility File (mobility.csv) format is as follows:

#Time(s)	Device ID	X	Y	Z
0	9	550	2500	0
0.5	9	1000	2500	0
1	9	1050	2500	0
1.5	9	1100	2500	0
2	9	1150	2500	0
2.5	9	1200	2500	0
3	9	1250	2500	0
3.5	9	1300	2500	0
.	.	.	.	.
.	.	.	.	.
.	.	.	.	.
38	9	3900	2500	0
39	9	3950	2500	0
40	9	4000	2500	0

Table-60: mobility.csv file

Step 4: Click on the gNB 7 and expand the right-side property panel and set as following Table-61.

Interface 4(5G RAN) Properties
CA Type	Single Band
CA Configuration	n78
CA Count	1
Component Carrier 1
DL UL Ratio	4:1
Numerology	0
Channel Bandwidth (MHz)	10
PRB Count	52
PDSCH Configuration
MCS Table	QAM64LOWSE
X Overhead	XOH0
PUSCH Configuration
MCS Table	QAM64LOWSE
CSI Report Configuration
CQI Table	Table 3
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	User Defined
LOS Probability	1
Shadow Fading Model	None
Fast Fading Model	No Fading
O2I Building Penetration Model	None
Additional Loss Model	None

Table-61:gNB 7 > 5G RAN Interface Properties Window

Similarly, it is set for gNB 8.

Step 5: The Tx Antenna Count was set to 2 and Rx Antenna Count was set to 1 in gNB > Interface (5G RAN) > Physical Layer.

Step 6: The Tx Antenna Count was set to 1 and Rx Antenna Count was set to 2 in UE > Interface (5G RAN) > Physical Layer.

Step 7: Configure CBR application from Server 12 to UE 9 by clicking on the set traffic tab in ribbon on the top. Then, click on the created application and expand the application property on the right and set the start time to 40 seconds, and QOS to UGS.

Step 8: Packet Trace is enabled by clicking on Configure reports tab. At the end of the simulation, a very large .csv file contains all the packet information and is available for the users to perform packet level analysis.

Step 9: LTENR Radio measurement, Handover log and SNR vs Time plot under LTENR Radio Measurements plots are enabled by clicking on plots/logs from right panel for detailed analysis.

Step 10: Run the Simulation for 50 Seconds.

Results and Discussion

Handover Signaling

Figure-66: Control packet flow in the 5G handover process

The packet flow depicted above can be observed from the packet trace.

UE will send the UE MEASUREMENT REPORT every 5 ms to the connected gNB
The initial UE-gNB connection and UE association with the core takes place by transferring the RRC and Registration, session request response packets.
As Per the configured file-based mobility, UE 9 moves towards gNB 8.
After 18.5 s gNB 7sends the HANDOVER REQUEST to gNB 8.
gNB 8 sends back HANDOVER REQUEST ACK to gNB 7.
After receiving HANDOVER REQUEST ACK from gNB 8, gNB 7 sends the HANDOVER COMMAND to UE 9
After the HANDOVER COMMAND packet is transferred to the UE, the target gNB will send the PATH SWITCH packet to the AMF via Switch 4.
When the AMF receives the PATH SWITCH packet, it sends MODIFY BEARER REQUEST to the SMF
The SMF on receiving the MODIFY BEARER REQUEST provides an acknowledgement to the AMF.
On receiving the MODIFY BEARER RESPONSE from the SMF, AMF acknowledges the Path switch request sent by the target gNB by sending the PATH SWITCH ACK packet back to the target gNB via Switch 4.
The target gNB sends CONTEXT RELEASE to source gNB, and the source gNB sends back CONTEXT RELEASE ACK to target gNB. The context release request and ack packets are sent between the source and target gNB via Switch 6.
RRC Reconfiguration will take place between target gNB and UE 9.
The UE 9 will start sending the UE (SS/PBCH) MEASUREMENT REPORT to gNB 8.

Figure-67: Screenshot of NetSim packet trace file showing the control packets involved in handover. Some columns have been hidden before the last column.

Plot of SNR vs. Time

Figure-68: Plot of the DL SNR over time seen by the UE from the serving cell (gNB 7) and the target cell (gNB 8). The handover process does not commence with Adj. cell SNR is greater than Serving cell SNR but only commences with Adj. cell SNR is greater than Serving cell SNR by the Handover margin (3 dB in this case).

This chart can be obtained in NetSim by enabling the option to plot SNR vs. time prior to the simulation. First, plot the SNR curve for gNB7 and UE9 keeping the channel as SSB. Then select "Add as new series" and select the gNB/eNB as gNB8 and UE name as UE9. Click on plot, and you would then obtain the above "stacked" plot

At 15.6 seconds, the signal-to-noise ratio (SNR) from both gNB7 and gNB8 is 16.84 dB. This is the point where the SNR curves for both gNBs intersect.
At 18.6 seconds, the SNR from gNB7 is 15.21 dB and the SNR from gNB8 is 18.54 dB. This is the point where Adj cell RSRP from gNB8 exceeds the serving cell RSRP by the handover margin (HOM) of 3 dB.

Throughput and delay variation during handover

NetSim UI displays the configuration file corresponding to this experiment as shown below Figure-69.

Figure-69: Network set up for studying the Throughput and delay variation during handover.

Procedure for Effect of Handover on Delay and Throughput

The following set of procedures were done to generate this sample:

Step 1: A network scenario is designed in NetSim GUI consisting of 2 gNBs, 5G Core, 1 Router, 1 Wired Node and 1 UE in the “5G NR” Network Library.

Step 2: The device positions are set as per the table given below Table-62.

	gNB 7	gNB 8	UE 9
X Coordinate	500	4500	500
Y Coordinate	500	500	1000

Table-62: Device positions

Step 3: Click on the gNB 7 and expand the right-hand side properties, and set as following

Interface(5G RAN) Properties
CA Type	Single Band
CA Configuration	n78
CA Count	1
Component Carrier 1
DL UL Ratio	4:1
Numerology	0
Channel Bandwidth (MHz)	10
PRB Count	52
PDSCH Configuration
MCS Table	QAM64
X Overhead	XOH0
PUSCH Configuration
MCS Table	QAM64
CSI Report Configuration
CQI Table	Table 1
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS_NLOS Selection	User Defined
LOS Probability	1
Shadow Fading Model	None
Fast Fading Model	No Fading
Additional Loss Model	None

Table-63: gNB 7> Interface(5G RAN) Properties Setting

Similarly, it is set for gNB 8.

Step 4: The Tx Antenna Count was set to 2 and Rx Antenna Count was set to 1 in gNB > Interface (5G RAN) > Physical Layer.

Step 5: The Tx Antenna Count was set to 1 and Rx Antenna Count was set to 2 in UE > Interface (5G RAN) > Physical Layer.

Step 6: In the position properties of UE 9, set Mobility Model as File Based Mobility.

Step 7: The BER and propagation delay was set to zero in all the wired links.

Step 8: Configure application between server and UE by selecting an application from Set Traffic Tab. Click on the application flow App1 CBR, expand the application property panel on the right and set the start time to 1s, QOS to UGS and Inter arrival time to 233.6µs by keeping the packet size as default.

Additionally, the “Start Time(s)” parameter is set to 1, while configuring the application.

File Based Mobility:

In File Based Mobility, users can write their own custom mobility models and define the movement of the mobile users. Create a mobility.csv file for UE’s involved in mobility with each step equal to 0.5 sec with distance 50 m.

The NetSim Mobility File (mobility.csv) format is as follows:

#Time(s)	Device ID		Y
0	9	500	1000
0.5	9	750	1250
1	9	1000	1500
1.5	9	1250	1750
2	9	1500	2000
2.5	9	1750	2250
3	9	2000	2500
3.5	9	2250	2750
4	9	2500	3000
4.5	9	2750	2750
5	9	3250	2250
5.5	9	3500	2000
6	9	3750	1750
6.5	9	4000	1500
7	9	4250	1250
7.5	9	4500	500

Table-64: mobility.csv file

Step 9: The LTENR Radio measurement log file can be enabled per the information provided in Section LTENR Results, Packet Trace and Plots of 5G technology library document and enable the Latency vs. Time and Throughput vs. Time under application performance plots

Step 10: Run the Simulation for 20 Seconds.

Results and Discussion

UDP Throughput Plot

Figure-70: We see how throughput varies with time, and the reasons for this variation, as the UE moves from the source gNB to the target gNB.

The application starts at 1s. The generation rate is 50 Mbps and we see the network is able to handle this load, and the throughput is equal to the generation rate. We then observe that the throughput starts dropping from 2.5s onwards because the UE is moving away from the gNB. As it moves as the SNR falls, and therefore a lower MCS is chosen leading to reduced throughput. At 3s there is a further drop in throughput and then a final dip at 3.9s. The time the handover occurs is 5.04 sec. At this point we see the throughput starts increasing once UE attaches to gNB8. The throughput for a short period of time is greater than 50 Mbps because of the transmission of queued packets in the s-gNB buffer which get transferred to the t-gNB buffer over the Xn interface.

Figure-71: Plot of Delay vs. Time

Since the application starts at 1s, the UDP plot begins at 1000 ms. The initial UDP delay is ≈ 1 ms, and hence the curve is seen as close to 0 on the Y axis. We then see that the packet delay starts increasing as the UE moves away from the gNB. This is because the link capacity drops as the CQI falls. The peak delay experienced shoots up to ≈ 1.1s at ≈ 5.5s when the handover occurs. Once the handover is complete the delay starts reducing and returns to ≈1 ms. The reason is that as the UE moves closer to the gNB its CQI increases and hence the 5G link can transmit at a higher rate (see Figure-70).

Impact of Handover margin and Time-To-Trigger on the performance of a 5G heterogeneous network

In a 5G heterogeneous network we analyze how the handover margin and time-to-trigger parameters influence two performance metrics: the number of handovers and the sum throughput (aggregate throughput of all UEs).

Open NetSim, Select Examples ->5G NR -> Impact of Handover margin and Time-To-Trigger on the performance of a 5G heterogeneous network, then click on the tile in the middle panel to load the example as shown in below screenshot.

Figure-72: List of scenarios for the example of a 5G heterogeneous network

The following network diagram illustrates what the NetSim UI displays while opening the example configuration file.

Figure-73: Network setup for studying 5G heterogeneous network.

System model

The study is based on a 3-tier 5G HetNet simulation. The network comprises gNB tiers at 1.5 GHz, 2.1 GHz, and 3.5 GHz.

gNB tiers	Frequency (GHz)	Pathloss Exponent (\(\mathbf{\eta)}\)	Transmit Power(dBm)	Antenna Type	Antenna Height (m)	Channel Bandwidth (MHz)
Tier 1 gNBs	1.5	2.9	37	Sector (3 Nos)	30	10
Tier 2 gNBs	2.1	2.9	37	Sector (3 Nos)	30	15
Tier 3 gNBs	3.5	3.9	30	Omni-Directional	20	50

Table-65: System parameters for the gNBs in the three different tiers

Each tier has a specific pathloss exponent influencing signal attenuation. The transmit power, antenna types (sector and omni-directional), and antenna heights vary across tiers. The simulation area is 10 km², with 60 User equipments (UEs) distributed randomly and 18 tier-I gNBs, 18 tier-II gNBs, and 12 tier-III gNBs distributed randomly. The gNBs across tiers will not interference since they operate at different frequencies.

Simulation parameters include gNB and UE antenna configurations, pathloss models, interference models, and mobility settings. Shadowing effects are modeled using a lognormal distribution with a standard deviation of 5 dB.

Simulation Area	10 km \(\mathbf{\times \ }\)10 km
Number of UEs	60 (distributed randomly)
Number of Tier 1 gNBs	18 (distributed randomly)
Number of Tier 2 gNBs	18 (distributed randomly)
Number of Tier 3 gNBs	12 (distributed randomly)
gNB Tx*Rx Antenna Count	1*1
UE Tx*Rx Antenna Count	1*1
gNB Pathloss Model	Log Distance
Downlink Interference Model	Exact Geometric Model
Mobility Model	Random Walk
Velocity	33 m/s
Calculation (update) interval for mobility	0.12 s (120 ms)
Measurement Interval (ms)	160
Shadowing	Lognormal. Std. dev. = 5 dB
Time to trigger (ms)	Varies; 128, 256, 512, 1024
Handover Margin (dB)	Varies; 0, 1, 2, 3, 4, 5, 6
Handover model	A3 event based
Simulation Time	30s
Traffic model	Saturated (full buffer) DL

Table-66: System parameters for the scenario being simulated,

The Time to Trigger (TTT) and Handover Margin (HO Margin) are variables in the study. An A3 event-based handover model is used. An Event A3-based HO is triggered when,

C1. The SINR of a user from target gNB becomes higher than the SINR of the user from the serving gNB by an offset. This offset is termed as handover margin.

C2. And this condition (C1) is maintained for a duration known as the time to trigger.

The model focuses on the interaction of these parameters and their effect on network performance, measured in terms of handover count and sum throughput.

Procedure to simulate the scenario using Multiparameter Sweeper

Click on the first experiment tile to open the scenario in NetSim. Save this scenario and open the experiment in the file explorer and open Configuration.netsim in Visual Studios.

Figure-74: Opening saved file location from ‘Your Work’ window.

Within the Datalink Layer of all gNB, in HANDOVER tag replace the HANDOVER_MARGIN="{1}" and TIME_TO_TRIGGER="{0}" representing an input variable for the multi-parameter sweeper.

Figure-75: input variables for Multi-Parameter Sweeper

Save the configuration file and rename it as input.xml.
Download the multi-parameter sweeper from the link https://github.com/NetSim-TETCOS/5g-Heterogeneous-Networkv14.4/archive/refs/heads/main.zip
Paste input.xml and Config support folder into the 5g-Heterogeneous-Networkv14.3 folder.
Open the multi-parameter-sweeper.py file in text editor, change the NETSIM_PATH suitably (line #14).

Run via CLI from 5g-Heterogeneous-Networkv14.3 folder as shown below.

The multi-parameter sweeper runs a total of 28 simulations, varying handover margin from 0 to 6 dB and time to trigger from 128, 256, 512, 1024 for all gNB’s. It generates an output file named "result.csv" which stores sum throughputs of all applications and the handover count. (It took us approximately 4 hours to complete all 28 simulations; we used a machine with a i5 processor and with 8 GB RAM).

Results and discussion

We tabulate below the handover count and sum throughput for various values of time to trigger (ms) and Handover margin (dB) which is obtained in results.csv file in multiparameter sweeper folder.

Time-to-trigger (ms)	Handover margin (dB)	Sum throughput (Mbps)	Handover count
128	0	207.01	470
128	1	206.55	418
128	2	207.68	292
128	3	198.46	330
128	4	210.85	282
128	5	206.95	275
128	6	201.77	172
256	0	225.77	200
256	1	216.60	180
256	2	218.39	155
256	3	213.70	131
256	4	212.72	109
256	5	209.87	98
256	6	219.78	82
512	0	206.32	84
512	1	200.89	52
512	2	222.01	46
512	3	214.25	41
512	4	208.27	31
512	5	213.92	25
512	6	203.96	25
1024	0	218.04	26
1024	1	226.85	24
1024	2	213.71	17
1024	3	211.79	14
1024	4	214.24	14
1024	5	218.94	10
1024	6	217.08	8

Table-67: Table demonstrates Sum Throughput and Handover Count variation with changes in Handover Margin and Time to Trigger (TTT).

Figure-76: Plot representing Handover Count vs Handover Margin for different TTTs.

It is evident from the plot that the handover count decreases as the handover margin increases. This trend is consistent across different TTT values, suggesting that a higher handover margin generally results in fewer handovers. The rationale behind this trend is that increased handover margin leads to more stringent conditions for handover and thereby reduces the frequency of handover occurrences.

We also observe that the handover count decreases as TTT increases. Shorter TTT values lead to quicker responses to signal changes, resulting in more frequent handovers, while longer TTT values delay the handover process, thereby reducing the handover count. The plot highlights the effects of both the handover margin and the TTT on handover count.

Figure-77: Plot representing Sum Throughput vs Handover Margin for Different TTTs

In the second chart we see that sum throughput generally rises and then falls with the increasing handover margin. For each handover margin we see the throughput again roughly increases and then drops as TTT increases. Initially, with a higher handover margin and/or higher TTT unnecessary and frequent handovers between cells are avoided. This leads to better throughput, but only to a certain extent. Beyond a point, a high handover margin and/or high TTT causes delayed handovers. Users stay connected to a weaker cell longer, despite being closer to a stronger cell, leading to poorer signal quality and thus lowering throughput.

QoS in 5G using GBR

Introduction

This experiment explores a new approach to providing Quality of Service (QoS) guarantees in 5G networks by modifying the traditional Proportional Fair Scheduling (PFS) algorithm. The study focuses on implementing Guaranteed Bit Rate (GBR) requirements using index bias (Lagrange multiplier) in the scheduler, and understanding how this modification impacts network performance under various scenarios.

The experiment investigates three cases: First, a baseline scenario using standard PFS where all User Equipment (UEs) are static but at different distances from the gNB (base station). Second, the same setup but with the modified PFS algorithm incorporating GBR guarantees for one UE. Finally, the study examines how the GBR mechanism performs when one UE is mobile. In all cases, we simulate Rayleigh fading channels between the gNB and UEs, creating dynamic channel conditions that reflect real-world wireless propagation.

Through these scenarios, the experiment demonstrates how the scheduler dynamically adjusts resource allocation to maintain throughput guarantees for specific UEs, at the expense of reducing resources to other users. The study is particularly useful for understanding how modern 5G networks can provide differentiated services and maintain quality guarantees in real-world conditions, where users may be at varying distances from the base station and potentially mobile.

Methodology

Open NetSim and click on Examples> 5G NR> QoS in 5G using GBR then click on the tile in the middle panel to load the example as shown below

Figure-78: List scenarios for the example of a QoS in 5G using GBR network

Case 1: Proportional Fair Scheduling (PFS). All UEs are static

NetSim UI displays the configuration file corresponding to this experiment as shown below

Figure-79: Network scenario

Set grid length as 6000m and width as 12000m from grid property panel on the right.
Set distance as follows:

gNB 9 to UE 10 = 1500m

gNB 9 to UE 11 = 2000m, and

gNB 9 to UE 12 = 2500m

Go to gNB properties → Interface (5G RAN), set the following properties as shown below Table. In the first case the scheduling type is set to PFS.

Properties
Datalink Layer Properties
Scheduling Type	PFS, PFS with RG
Physical Layer Properties
CA Type	Single band
CA Configuration	n78
CA1
Numerology	1
Channel Bandwidth	100 MHz
Channel Model
Pathloss Model	3GPP TR 38.901-7.4.1
Outdoor Scenario	Urban Macro
LOS NLOS Selection	User defined
LOS Probability	1
Shadow Fading Model	None
Fast Fading Model	No fading

Table-68: gNB >Interface (5G RAN) > Datalink and Physical layer properties

Set Tx Antenna Count as 1 and Rx Antenna Count as 1 in gNB properties.
Set Tx Antenna Count as 1 and Rx Antenna Count as 1 in all the UEs.
Go to the Set Traffic tab in the top ribbon and create a CBR application as shown in the table below. To change the transport protocol, QoS, and IAT click on the application and change the properties in the right-side property panel.

Application Properties
	Application 1	Application 2	Application 3
Application Type	CBR	CBR	CBR
Source ID	8	8	8
Destination ID	10	11	12
QoS	UGS	UGS	UGS
Transport Protocol	UDP	UDP	UDP
Packet Size	1460 Bytes	1460 Bytes	1460 Bytes
Inter-arrival time	58.4 μs	58.4 μs	58.4 μs
Start Time	1s	1s	1s

Table-69: Application properties

Make sure you enable these two plots under plots section as shown in figure

Figure-80: Enabling the plots

Run Simulation for 100 s and note down throughput value in the results window for each UE.
Here, we can see the resulting plots of the case:

Figure-81: EWMA MAC Throughput vs Time

Case 2: PFS with RG using Guaranteed Bit Rate (GBR). All UEs are static.

Now, for the same scenario above we just need to disable the GBR Configuration.
For doing that we need to change the Scheduling algorithm from Proportional Fair to PFS with RG as shown below.

Figure-82: Changing the Scheduling Algorithm

Then, Configure GBR UEs via GUI as shown below. Change the Downlink of UE12 to 27.57Mbps and click on update.

Figure-83: Adding GBR DL to UE

Now, run the scenario for 100 s.
Now plot the values using plots section in Results dashboard.

Figure-84: EWMA MAC Throughput vs Time

Figure-85: Index Bias vs Time

Case 3: PFS with RG using GBR. One of the UE’s is mobile.

In this case we are working with GBR again.
Now we need to add Mobility for UE-12, click on UE-12 then select Position ->Mobility Model ->File Based Mobility -> via File.

Figure-86: Adding file-based mobility to UE

Now add Mobility as shown below:

Figure-87: Mobility.csv file

In this way, you can add mobility by increasing 10 m of distance in the interval of 1 s and keep adding up to 100 s then save the file.

Now, run the simulation for 100 s.
The plots are obtained from results dashboard:
So, we can see a lot of spike variations in plots due to mobility in UE 3.

Figure-88: EWMA MAC Throughput vs Time

Figure-89: Index Bias vs Time

Results and Discussion

Let’s analyze the Application Throughputs and MAC Throughputs obtained by 3 UEs in results dashboards in all the three different cases discussed above:

Case #

Description

UE1

(Mbps)

UE2

(Mbps)

UE3

(Mbps)

1

Proportional Fair Scheduling (PFS). All UEs static.

64.65

37.22

16.65

2

PFS with RG using Guaranteed Bit Rate (GBR). All UEs static.

54.97

31.65

21.63

3

PFS with RG using GBR. UE 3 is mobile.

28.15

16.20

20.44

Table-70: Application Throughputs

Case #

Description

UE1

(Mbps)

UE2

(Mbps)

UE3

(Mbps)

1

Proportional Fair Scheduling (PFS). All UEs static.

82.34

47.40

21.20

2

PFS with RG using Guaranteed Bit Rate (GBR). All UEs static.

70.02

40.31

27.55

3

PFS with RG using GBR. UE 3 is mobile.

35.86

20.64

26.04

Table-71: MAC Throughputs

Case #

Description

UE1

(%)

UE2

(%)

UE3

(%)

1

Proportional Fair Scheduling (PFS). All UEs static.

33.33

2

PFS with RG using Guaranteed Bit Rate (GBR). All UEs static.

28.35

28.34

43.31

3

PFS with RG using GBR. UE 3 is mobile.

14.51

70.97

Table-72: Resource allocation

Case-1: Proportional Fair (PF) Algorithm

Under standard PFS, the algorithm distributes Physical Resource Blocks (PRBs) uniformly among the three UEs, with each receiving approximately 33.33% of the resources. This results in decreasing throughputs as distance from the gNB increases - UE1 achieves 64.65 Mbps, UE2 gets 37.22 Mbps, and UE3 receives 16.65 Mbps at the application layer.

Case-2: GBR with Static UEs

When GBR is enabled for UE3 with a target rate of 27.57 Mbps, we observe the index bias mechanism actively working to guarantee this rate. The scheduler increases the bias factor for UE3, resulting in its PRB allocation increasing from 33.33% to 43.31%.
This resource reallocation successfully raises UE3's throughput from 16.65 Mbps to 21.63 Mbps at the application layer, achieving the target MAC layer throughput of 27.55 Mbps.
However, this comes at the cost of reduced resources for other UEs - UE1's throughput drops from 64.65 Mbps to 54.97 Mbps, and UE2's decreases from 37.22 Mbps to 31.65 Mbps.

Case-3: GBR with Mobile UE

As UE3 begins moving away from the gNB, we observe a complex interplay between distance, channel conditions, and the index bias mechanism.
Initially, as UE3's throughput starts dropping due to increased distance, the index bias increases to compensate, pulling more resources from UE1 and UE2. Their throughputs drop significantly - UE1 falls to 28.15 Mbps and UE2 to 16.20 Mbps.
As UE3 moves even further away, the scheduler dramatically increases the index bias in an attempt to maintain the GBR. This results in UE3 being allocated nearly 71% of all PRBs, leaving only about 14.5% each for UE1 and UE2.
Despite this extreme resource reallocation, UE3's throughput still falls to 20.44 Mbps, unable to meet the GBR target due to poor channel conditions at the increased distance. Meanwhile, the throughputs of UE1 and UE2 are severely impacted due to their minimal resource allocation.

This behavior demonstrates both the power and limitations of the index bias mechanism in GBR implementation - while it can effectively guarantee bit rates under most conditions by redistributing resources, there are physical limitations that cannot be overcome simply by increasing resource allocation when channel conditions become too poor.

Featured Examples

Derive from 3GPP standards the theoretical data rate and throughput for a 1gNB - 2UE scenario, and compare with simulation

Introduction

Network Setup

Network Settings

Results

Results and discussion

Exercises

Effect of distance on pathloss, SINR and MCS (7-Cell Hexagonal Layout, Urban Macro Propagation, NLOS)

Effect of UE distance on throughput in FR1 and FR2

Frequency Range - FR1

Frequency Range - FR2

Impact of MAC Scheduling algorithms on throughput, in a multi-UE scenario

Multi UE throughput with UEs at different distances and channel is not time varying.

Multi UEs at different distances with a time varying channel

Max Throughput for different MCS and CQI

Load balancing in 5G using Cell Individual Offset (CIO)

Introduction

Network Setup

Network Settings

Results

Discussion

4G vs. 5G: Capacity analysis for video downloads

4G

5G

5G-Peak-Throughput

3.5 GHz n78 band

26 GHz n258 band

Impact of distance on throughput for n261 band in LOS and NLOS states

DL: UL Ratio 4:1

DL: UL Ratio 3:2

gNB cell radius for different data rates

3.5 GHz n78 urban gNB cell radius for different data rates

26 GHz n258 urban gNB cell radius for different data rates

Impact of numerology on a RAN with phones, sensors, and cameras

Impact of UE movement on Throughput

Simulate and study the 5G Handover procedure.

Introduction

Network Setup

Handover Algorithm

Throughput and delay variation during handover

Impact of Handover margin and Time-To-Trigger on the performance of a 5G heterogeneous network

QoS in 5G using GBR

Introduction

Methodology

Case 1: Proportional Fair Scheduling (PFS). All UEs are static

Case 2: PFS with RG using Guaranteed Bit Rate (GBR). All UEs are static.

Case 3: PFS with RG using GBR. One of the UE’s is mobile.

Obtaining the EWMA MAC Throughput and Resource share

Results and Discussion