Understanding the 5G NR PHY

Objective: This experiment has four goals. First, to gain an appreciation for the 5G NR
physical layer, i.e., the time-frequency resource grid in the OFDM access scheme. Second, to
understand how a packet is transmitted over this OFDM PHY in NetSim, and the assumptions
involved. Third, to analytically estimate (per 3GPP standards) the application throughput for a
simple use-case. And finally, to simulate and analyse throughput as different PHY parameters
are varied.

Introduction
OFDM: 5G uses Orthogonal Frequency Domain Multiplexing (OFDM) as the multiple access
scheme for both downlink and uplink transmissions with the flexibility of multiple subcarrier
spacing that supports diverse application scenarios. The smallest physical resource, known
as the resource element (RE), comprises one subcarrier and one OFDM symbol.
The time-domain transmission structure comprises of frames 10 ms (to support backwards
compatibility with LTE). Each frame is composed of 10 subframes of 1 ms each. The 1 ms
subframe is then divided into one or more slots in 5G, whereas LTE had exactly two slots in a
1
subframe. The slot size depends on the numerology, 𝜇, and is equal to ms. The number of
2𝜇

OFDM symbols per slot is 14 for a configuration using normal cyclic prefix. For extended cyclic
prefix, the number of OFDM symbols per slot is 12. Data is transmitted over these symbols.

Fig 1: NR Frame Structure when numerology 𝝁 is set to 3

In the time division duplex version, each frame is partitioned into downlink subframes and an
uplink subframes. The downlink part of each frame is used to send data from the gNB to the
UEs. The uplink part of the frame is used to send data from the UEs to the destinations, via
the gNB. The uplink-downlink ratio is a GUI parameter in NetSim. If Internet access is a major
application in a system, then the downlink part of the frame would be substantially larger than
the uplink part, due to the asymmetry of Internet access traffic.
In the frequency domain, the group of 12 consecutive sub-carriers forms a resource block
(RB). The sub carrier spacing (SCS) is also dependent on numerology, 𝜇 and is equal to
2𝜇 × 15 KHz. 5G supports total carrier bandwidth up to 400 MHz with a maximum of 275 RBs.

Fig 2: The OFDM frame structure. Slot times get shorter as the sub-carrier spacing gets larger.

Data Transmission in NetSim

• In TDD operation the UL and DL transmissions are separated in the time-domain over
different frames/subframes/slots/symbols and use the same carrier frequency. In FDD
operation UL and DL transmissions are separated in the frequency domain, with
different frequencies used for UL and for DL transmissions. NetSim does slot-based
scheduling. For example, if the DL:UL ratio is 4:1 then 4 slots are allotted to DL and 1
slot to UL.

• Higher layer packets arrive at the RLC buffer for each UE and each gNB.
▪ The MAC Scheduler determinates of the Transport block size (TBS) based on the
channel quality index (CQI). The CQI is determined by the Adaptive Modulation and
Coding (AMC) function based on the SNR.

▪ Now, the SNR is determined from a) large scale pathloss and shadowing calculated
per the 3GPP’s stochastic propagation models, and b) the small-scale fading which
leads to beamforming gains when using MIMO. These models provide signal
attenuation as an output. Several parameters are used in the model, including the
distance between the transmitter and the receiver. These computations are executed
each associated UE-gNB pair, in DL and UL, at the start of simulation and again at
every mobility event. In calculating SNR, the noise power is obtained from 𝑁 =
𝑘 × 𝑇 × 𝐵.

▪ Note that the SNR/CQI is not computed/fed-back using reference signals but is
computed on the data channel. Then it is assumed to be instantaneously known to the
transmitter and receiver. This assumption is known as perfect CSIT and CSIR. With
perfect CSIT the transmitter can adapt its transmission rate (MCS) relative to the
instantaneous channel state (SNR).

▪ Based on this SNR the AMC determines a wideband CQI which indicates the highest
rate Modulation and coding scheme (MCS), that it can reliably decode, if the entire
system bandwidth were allocated to that user. The rate adaptation is discrete (not
continuous), and the modulation and coding scheme (MCS) is selected from a
standards specified table. The modulation scheme defines the number of bits, that can
be carried by a single RE. Modulation schemes supported by 5G include QPSK (2
bits), 16 QAM (4 bits), 64 QAM (6 bits), and 256 QAM (8 bits). The code rate defines
the proportion of bits transmitted that are useful. It is computed as the ratio of useful
bits by total bits that are transmitted. The modulation order 𝑄𝑚 , which denotes the
number of bits per RE, and the code rate denoted by 𝑅 are jointly encoded as
modulation and coding scheme (MCS) index. These values of 𝑄𝑚 and 𝑅 are then
passed to the TBS determination function.

▪ At each gNB a frame of length 10ms is started. Each frame in turn starts 10 sub frames
each of length 1ms. Each sub frame then starts a certain number of slots based on
numerology.

• The PHY layer in NetSim then notifies the MAC about the slot start. The MAC sub layer
in turn seeks a buffer status report from the RLC layer and invokes the MAC scheduler.
It then notifies the RLC of the transmission. The RLC then transmits the transport block
to the PHY layer. The downlink and uplink data channels (PDSCH, PUSCH) receive
 this transport block as its service data unit (SDU), which is then processed and
transmitted over the radio interface.

Procedure:
1. Use the following download Link to download a compressed zip folder which contains
the workspace.
https://github.com/NetSim-
TETCOS/5G_Experiments_v13.0/archive/refs/heads/main.zip
2. Extract the zip folder.
3. The extracted project folder consists of a NetSim workspace file (*.netsim_wsp).
4. Go to NetSim Home window, go to Your Work and click on workspace options.

Fig 3: NetSim Home Window

5. Select more options.

Fig 4: NetSim Your Work window

6. Click on Import.

Fig 5: NetSim Your Work Window to import a workspace.

7. In the Import Workspace Window that appears browse and select the *.netsim_wsp
file from the extracted directory and for the Destination path browse to select a path in
your system where you want to set up the workspace folder.

Fig 6: NetSim Import Workspace Window

8. While importing the workspace, if a warning message indicating a software version

mismatch is displayed, it can be ignored by clicking on OK.
9. The Imported Project workspace will automatically be set as the current workspace.
 Fig 7: NetSim Your work window with list of workspaces

Network scenario:

NetSim UI would display the following network topology when you open the example
configuration file as shown below screenshot.

Fig 8: Network topology in this experiment

Settings:

The following settings were configured in the network setup.

1. The UE was placed 100m away from the gNB.

2. The following properties were set in Interface 5G RAN, Physical Layer of gNB.

gNB Interface 5G RAN

gNB Height (m) 10
Tx Power (dBm) 31, 34, 37, 40, 43 (Varied)
Tx Antenna Count 1* (varied from 1 to 8)
Rx Antenna Count 4
CA Type Single Band
CA Configuration n261
DL: UL Ratio 4:1
Numerology 3
Channel Bandwidth (MHz) 50, 100, 200, 400 (Varied)
MCS Table QAM256
 CQI Table TABLE2
Outdoor Scenario Rural Macro
Indoor Office Type Mixed Office
Pathloss Model 3GPPTR38.901-7.4.1
LOS Mode User Defined
LOS Probability 0
Shadow Fading Model None
Fading and Beamforming No Fading
O2I Building Penetration Model None

Table 1: gNB properties

3. Tx Antenna Count =4 and Rx Antenna Count is varied from 1 to 8 in UE > Interface 5G

RAN > Physical Layer
4. A downlink CBR application was configured from Wired Node to UE with Packet Size
1460B and IAT (varied for each layer count) and the start time was set to 1s.
5. Run simulation for 1.2s.
6. Run simulation for different MIMO Layers and different Tx Powers in gNB and note
down the throughput obtained.

Analytical Estimation of Data Throughput

We derive the throughput for the setting involving 2 × 2 MIMO (2-layers) with 100MHz
Bandwidth and 31 dBm Transmit power. The procedure for TBS determination given in the
steps below is per 3GPP TS 38.214 Section 5.1.3.2 (DL).

1. Initially, the Pathloss (dB) is calculated based on the Pathloss models specified in the
3GPP standards1. Pathloss in this example turns out to be 122.26 𝑑𝐵
2. The Total Loss is then calculated using the following equation:
𝑇𝑜𝑡𝑎𝑙𝐿𝑜𝑠𝑠 (𝑑𝐵) = 𝑃𝑎𝑡ℎ𝑙𝑜𝑠𝑠 (𝑑𝐵) + 𝑆ℎ𝑎𝑑𝑜𝑤𝐹𝑎𝑑𝑖𝑛𝑔 𝐿𝑜𝑠𝑠 + 𝑂2𝐼 𝐿𝑜𝑠𝑠 + 𝐴𝑑𝑑𝑖𝑡𝑖𝑜𝑛𝑎𝑙 𝐿𝑜𝑠𝑠

In this example, 𝑆ℎ𝑎𝑑𝑜𝑤 𝐹𝑎𝑑𝑖𝑛𝑔 𝐿𝑜𝑠𝑠 = 0, 𝑂2𝐼 𝐿𝑜𝑠𝑠 = 0, 𝐴𝑑𝑑𝑖𝑡𝑖𝑜𝑛𝑎𝑙 𝐿𝑜𝑠𝑠 = 0

𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑠𝑠 = 122.26 + 0 + 0 + 0 = 122.26 𝑑𝐵

3. The Received power (Per layer) is calculated, using the Tx Power (per layer), the
Total Loss and the Beamforming Gain (per layer). Since fading is turned off the
Beamforming (BF) gain per layer is 0 dB.
𝑅𝑥 𝑃𝑜𝑤𝑒𝑟𝐿𝑎𝑦𝑒𝑟 (𝑑𝐵𝑚) = 𝑇𝑥 𝑃𝑜𝑤𝑒𝑟𝐿𝑎𝑦𝑒𝑟 (𝑑𝐵𝑚) − 𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑠𝑠 (𝑑𝐵) + 𝐵𝐹𝐺𝑎𝑖𝑛𝐿𝑎𝑦𝑒𝑟

1
We do not get into the details of the pathloss computations here. The specifics are explained in Experiment 4:
Understand 3GPP 5GNR pathloss models.
 𝑅𝑥 𝑃𝑜𝑤𝑒𝑟𝐿𝑎𝑦𝑒𝑟 1 = 27.98 − 122.26 + 0 = −94.28 𝑑𝐵𝑚
𝑅𝑥 𝑃𝑜𝑤𝑒𝑟𝐿𝑎𝑦𝑒𝑟 2 = 27.98 − 122.26 + 0 = −94.28 𝑑𝐵𝑚
4. Thermal Noise computation
𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑖𝑠𝑒 = 𝑘 × 𝑇 × 𝐵
𝑘 (𝐵𝑜𝑙𝑡𝑧𝑚𝑎𝑛𝑛′ 𝑠 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡) = 1.38 ∗ 10−23 , 𝑇 (𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒) = 300 𝐾, 𝐵 = 100 𝑀𝐻𝑧
𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑖𝑠𝑒 = = 4.14 × 10−13 𝑊
= −93.829 𝑑𝐵𝑚
5. From the Rx Power and Thermal Noise, SNR is calculated
a) 𝑅𝑥 𝑃𝑜𝑤𝑒𝑟 𝑖𝑛 𝑑𝐵𝑚 𝑖𝑠 𝑐𝑜𝑛𝑣𝑒𝑟𝑡𝑒𝑑 𝑖𝑛𝑡𝑜 𝑚𝑊
𝑅𝑥 𝑃𝑜𝑤𝑒𝑟, 𝑃 = −94.28 𝑑𝐵𝑚 = 3.73 × 10−10 𝑚𝑊
b) 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑖𝑠𝑒 𝑖𝑛 𝑑𝐵𝑚 𝑖𝑠 𝑐𝑜𝑛𝑣𝑒𝑟𝑡𝑒𝑑 𝑖𝑛𝑡𝑜 𝑚𝑊
𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑖𝑠𝑒, 𝑁 = −93.82 𝑑𝐵𝑚 = 4.14 ∗ 10−10 𝑚𝑊
𝐸 𝑅𝑥 𝑃𝑜𝑤𝑒𝑟 𝑃 3.73×10−10
c) 𝑆𝑁𝑅(𝐿𝑖𝑛𝑒𝑎𝑟) = 𝑁𝑏 = 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑖𝑠𝑒 = 𝑁 = 4.14×10−10 = 0.902
0

6. From SNR, the Spectral Efficiency is calculated as follows:

𝐸𝑏
𝑆𝑝𝑒𝑐𝑡𝑟𝑎𝑙 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝐿𝑎𝑦𝑒𝑟 = log 2 (1 + ( ))
𝑁0

= log 2 (1 + (0.902)) = 0.927

7. The CQI Index is then looked up from the respective CQI Table using the spectral
efficiency obtained. The table is given below:
{0, Modulation_Zero, 0, 0}, //out of range
{1, Modulation_QPSK, 78, 0.1523},
{2, Modulation_QPSK, 193, 0.3770},
{3, Modulation_QPSK, 449, 0.8770},
{4, Modulation_16_QAM, 378, 1.4766},
{5, Modulation_16_QAM, 490, 1.9141},
{6, Modulation_16_QAM, 616, 2.4063},
{7, Modulation_64_QAM, 466, 2.7305},
{8, Modulation_64_QAM, 567, 3.3223},
{9, Modulation_64_QAM, 666, 3.9023},
{10, Modulation_64_QAM, 772, 4.5234},
{11, Modulation_64_QAM, 873, 5.1152},
{12, Modulation_256_QAM, 711, 5.5547},
{13, Modulation_256_QAM, 797, 6.2266},
{14, Modulation_256_QAM, 885, 6.9141},
{15, Modulation_256_QAM, 948, 7.4063},

Since the Spectral Efficiency is 0.927, from the CQI Table, 𝐶𝑄𝐼 𝐼𝑛𝑑𝑒𝑥 = 3 is chosen.
8. Similarly, the MCS Index is taken from the respective MCS Table with respect to the
spectral efficiency from the CQI Table:

{0, 2, Modulation_QPSK, 120, 0.2344},

 {1, 2, Modulation_QPSK, 193, 0.3770},
{2, 2, Modulation_QPSK, 308, 0.6016},
{3, 2, Modulation_QPSK, 449, 0.8770},
{4, 2, Modulation_QPSK, 602, 1.1758},
{5, 4, Modulation_16_QAM, 378, 1.4766},
{6, 4, Modulation_16_QAM, 434, 1.6953},
{7, 4, Modulation_16_QAM, 490, 1.9141},
{8, 4, Modulation_16_QAM, 553, 2.1602},
{9, 4, Modulation_16_QAM, 616, 2.4063},
{10, 4, Modulation_16_QAM, 658, 2.5703},
{11, 6, Modulation_64_QAM, 466, 2.7305},
{12, 6, Modulation_64_QAM, 517, 3.0293},
{13, 6, Modulation_64_QAM, 567, 3.3223},
{14, 6, Modulation_64_QAM, 616, 3.6094},
{15, 6, Modulation_64_QAM, 666, 3.9023},
{16, 6, Modulation_64_QAM, 719, 4.2129},
{17, 6, Modulation_64_QAM, 772, 4.5234},
{18, 6, Modulation_64_QAM, 822, 4.8164},
{19, 6, Modulation_64_QAM, 873, 5.1152},
{20, 8, Modulation_256_QAM, 682.5, 5.3320},
{21, 8, Modulation_256_QAM, 711, 5.5547},
{22, 8, Modulation_256_QAM, 754, 5.8906},
{23, 8, Modulation_256_QAM, 797, 6.2266},
{24, 8, Modulation_256_QAM, 841, 6.5703},
{25, 8, Modulation_256_QAM, 885, 6.9141},
{26, 8, Modulation_256_QAM, 916.5, 7.1602},
{27, 8, Modulation_256_QAM, 948, 7.4063},

Since the Spectral Efficiency is 0.927, MCS Index 3 which corresponds to this
spectral efficiency is chosen and the 𝑀𝑜𝑑𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑂𝑟𝑑𝑒𝑟 = 2
9. The TBS size is then determined using the Modulation Order and code rate.
10. Determination of number of Resource Elements within the slot.

′ 𝑅𝐵 𝑆ℎ 𝑃𝑅𝐵 𝑃𝑅𝐵
𝑁𝑅𝐸 = 𝑁𝑆𝐶 × 𝑁𝑆𝑦𝑚𝑏 − 𝑁𝐷𝑀𝑅𝑆 − 𝑁𝑂𝐻

𝑅𝐵
𝑁𝑆𝐶 = 12. 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑢𝑏𝑐𝑎𝑟𝑟𝑖𝑒𝑟𝑠 𝑖𝑛 𝑃ℎ𝑦𝑠𝑖𝑐𝑎𝑙 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 𝐵𝑙𝑜𝑐𝑘

𝑆ℎ
𝑁𝑆𝑦𝑚𝑏 = 14. 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑆𝑦𝑚𝑏𝑜𝑙𝑠 𝑝𝑒𝑟 𝑆𝑙𝑜𝑡

𝑃𝑅𝐵
𝑁𝐷𝑀𝑅𝑆 = 0 → 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 𝐸𝑙𝑒𝑚𝑒𝑛𝑡𝑠 𝑓𝑜𝑟 𝐷𝑀 − 𝑅𝑆 𝑝𝑒𝑟 𝑃𝑅𝐵

𝑃𝑅𝐵
𝑁𝑂𝐻 = 0. 𝑃𝐷𝑆𝐶𝐻 𝑜𝑣𝑒𝑟ℎ𝑒𝑎𝑑

′
𝑁𝑅𝐸 = 12 × 14 − 0 − 0 = 168

11. Total number of Resource Elements allocated for PDSCH

′ )
𝑁𝑅𝐸 = min(156, 𝑁𝑅𝐸 × 𝑛𝑃𝑅𝐵

𝑛𝑃𝑅𝐵 = 1. 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑒𝑑 𝑃𝑅𝐵𝑠 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑈𝐸

 𝑁𝑅𝐸 = min(156, 168) × 1 = 156 × 1 = 156

12. Intermediate number of information bits is calculated.

𝑁𝑖𝑛𝑓𝑜 = 𝑁𝑅𝐸 × 𝑅 × 𝑄𝑚

𝑀𝐶𝑆𝐶𝑜𝑑𝑒𝑅𝑎𝑡𝑒 449
𝑅= 1024
= 1024 = 0.438 and 𝑄𝑚 = 2 (Modulation order)

𝑁𝑖𝑛𝑓𝑜 = 156 × 0.438 × 2 = 136.65

13. Since 𝑁𝑖𝑛𝑓𝑜 ≤ 3824 , the TBS Size is calculated

′
𝑁𝑖𝑛𝑓𝑜
𝑁𝑖𝑛𝑓𝑜 = max (24, 2𝑛 × 𝑓𝑙𝑜𝑜𝑟 ( ))
2𝑛

Where 𝑛 = max (3, 𝑓𝑙𝑜𝑜𝑟(log 2(𝑁𝑖𝑛𝑓𝑜 )) − 6)

= 𝑚𝑎𝑥 (3, 𝑓𝑙𝑜𝑜𝑟(𝑙𝑜𝑔2 (136.65)) − 6) = 𝑚𝑎𝑥 (3, 1) = 3

′ 136.65
𝑁𝑖𝑛𝑓𝑜 = 𝑚𝑎𝑥 (24, 23 × 𝑓𝑙𝑜𝑜𝑟 ( )) = 𝑚𝑎𝑥 (24, 136) = 136
23

Hence, the TBS size will be 136, i.e., index 15.

14. The bits per PRB per layer is determined based on the TBS size.
15. For Layer 1,
𝑏𝑖𝑡𝑠𝑝𝑒𝑟𝑃𝑅𝐵 = 𝑏𝑖𝑡𝑠 𝑝𝑒𝑟 𝑃𝑅𝐵 (𝑖𝑛𝑖𝑡𝑖𝑎𝑙) + 𝑇𝐵𝑆 𝑆𝑖𝑧𝑒
= 0 + 136 = 136
For Layer 2,
𝑏𝑖𝑡𝑠𝑝𝑒𝑟𝑃𝑅𝐵𝐿2 = 𝑏𝑖𝑡𝑠𝑝𝑒𝑟𝑃𝑅𝐵𝐿1 + 𝑇𝐵𝑆 𝑆𝑖𝑧𝑒
= 136 + 136 = 272
16. The Total PRB available is dependent on bandwidth and 𝜇 and is shown in the GUI.
In this example 𝑃𝑅𝐵 𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 = 𝑃𝑅𝐵 𝐶𝑜𝑢𝑛𝑡 = 66
17. The total PRB available is calculated.
𝑇𝑜𝑡𝑎𝑙 𝑃𝑅𝐵 𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 = 𝑃𝑅𝐵𝐶𝑜𝑢𝑛𝑡 − 𝑐𝑒𝑖𝑙(𝑃𝑅𝐵 𝑐𝑜𝑢𝑛𝑡 × 𝑂𝐻𝐷𝑜𝑤𝑛𝑙𝑖𝑛𝑘 )
𝑃𝑅𝐵 𝐶𝑜𝑢𝑛𝑡 = 66, 𝑂𝐻𝐷𝑜𝑤𝑛𝑙𝑖𝑛𝑘 = 0.18 (Per standard)
𝑇𝑜𝑡𝑎𝑙 𝑃𝑅𝐵 𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 = 66 − 𝑐𝑒𝑖𝑙(66 × 0.18)
= 66 − 11 = 54
18. Number of PRBs allocated is 54.
19. The slot allocation will then take place and the bits per slot is assigned.
𝑏𝑖𝑡𝑠 𝑝𝑒𝑟 𝑆𝑙𝑜𝑡 = 𝑏𝑖𝑡𝑠 𝑝𝑒𝑟 𝑃𝑅𝐵 × 𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑒𝑑 𝑃𝑅𝐵
= 272 × 54 = 14688 𝑏𝑖𝑡𝑠 = 1836 𝐵𝑦𝑡𝑒𝑠
𝑖. 𝑒. , 𝑎 𝑠𝑙𝑜𝑡 𝑐𝑎𝑛 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡 𝑎 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑜𝑓 1836 𝐵𝑦𝑡𝑒𝑠
 20. Throughput estimation. 𝐷𝐿 𝑈𝐿 𝑅𝑎𝑡𝑖𝑜 = 4: 1, implies a DL fraction of 0.8. Since 𝜇 = 3,
1
slot time 23 ms.
1836 × 8 × 0.8
𝐷𝐿 𝑀𝐴𝐶 𝑇ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 = = 94 𝑀𝑏𝑝𝑠
1
( 3 ) × 10−3
2

𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑃𝑎𝑐𝑘𝑒𝑡𝑆𝑖𝑧𝑒
𝐷𝐿 𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑇ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 = 𝐷𝐿 𝑀𝐴𝐶 𝑇ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 × ( )
𝑀𝐴𝐶𝑃𝑎𝑐𝑘𝑒𝑡𝑆𝑖𝑧𝑒

1460
= 94 × ( ) = 92.23 𝑀𝑏𝑝𝑠
1488

(Matches with NetSim’s Result of 92.21 Mbps, which can be seen in Table 2 entry pertaining
to MIMO 2*2, Bandwidth 100MHz and Tx Power 31 dBm. The entry is marked in green.)

Results
The following throughputs were obtained after running simulations with different Antenna
counts (MIMO layers), Bandwidths and Transmit Power values.

MIMO Total Tx Peak Application Throughput (Mbps)

Power2 (dBm) Bandwidth (MHz)
for various
50 100 200 400
31 49.58 75.92 92.21 75.92
34 70.48 103.07 151.89 184.42
1*1 37 99.22 146.46 206.15 303.79
40 114.93 206.15 292.93 412.30
43 144.94 238.68 412.30 585.92

31 73.11 92.21 75.92 86.78

34 99.22 151.89 184.42 151.89
2*2 37 141.03 206.15 303.79 368.91
40 198.50 292.93 412.30 607.65
43 229.86 412.30 585.92 824.66

31 88.76 75.92 86.78 173.56

34 146.23 184.42 151.89 173.56
4*4 37 198.50 303.79 368.91 303.79
40 282.07 412.30 607.65 737.82
43 397.06 585.92 824.66 1215.30

31 73.11 86.78 173.56 0

34 177.59 151.89 173.56 347.18
8*8 37 292.52 368.91 303.79 347.18
40 397.06 607.65 737.82 607.65
43 564.20 824.66 1215.30 1475.59

Table 2: Saturation throughput obtained for n261 band (gNB-UE distance of 100m, Rural macro pathloss) for
various Bandwidth-MIMO-TxPower combinations. The blue entries show the doubling of throughput when power
and BW is doubled. Red shows examples where throughput decreases with increase in bandwidth for fixed
power and MIMO layers. Green entries are where throughput decreases with increase in MIMO layers, for fixed
BW and power.

2 This is the transmit power summed over entire BW and summed over all MIMO streams.
Discussion
In Table 2 we observe entries marked in:

• Blue: When both the bandwidth and the power are doubled, with MIMO count kept
constant, the peak throughput doubles. This is along expected lines.
• Red: In the high bandwidth and low power regime, when the bandwidth is doubled with
the transmit power and MIMO count held constant, the peak throughput does not
increase but rather decreases.
• Green: At low power when the MIMO layers are increased with fixed transmit power
and bandwidth, the peak throughput surprisingly decreases.

Let us understand the red entries, i.e., throughput of 1*1 MIMO, 31 dBm Tx power for
bandwidths of 200 and 400 MHz. We can simplify the PHY rate as equal to 𝑘 × 𝐿 × 𝑄 × 𝐵 × 𝑅
where 𝑘 is some constant, 𝐿 is the number of layers (set to 2 here), 𝑄 is the modulation order
(2 in this case), 𝑅 is the code rate and 𝐵 is the bandwidth. From Table 3 we see that when the
bandwidth increases the spectral efficiency decreases due to an increase in thermal noise at
higher bandwidths. The received power is constant since the transmit power is fixed. Since
the drop in the MCS3 (due to the reduced spectral efficiency) is larger than the bandwidth
increase - 0.483 × 200 vs. 0.188 × 400 - the net effect is a decline in the throughput.

Rx Spec Eff MCS R

BW Noise Spectral MCS Throughput
Power SNR Table cut Code (Code
(MHz) (KTB) Efficiency Index (Mbps)
(dB) off Rate rate/1024)
200 -91.26 -90.81 -0.44 0.927 0.8770 3 449 0.438 92.21
400 -91.26 -87.80 -3.45 0.537 0.3770 1 193 0.188 75.92
Table 3: Rx Power, Noise, SNR, Spectral Efficiency obtained for different bandwidths with Tx power at 31 dBm.

Next, we turn to the green entries. In Table 4, notice that when the MIMO layer count is
increased from 4 to 8, the received power (per layer) decreases. This happens because the
transmit power is equally divided among all the layers. As SNR reduces, the spectral efficiency
per layer decreases. Since the MCS drop (due to lower spectral efficiency) is larger than the
multiplexing gain got from multiple MIMO streams - 4 × 0.438 vs 8 × 0.188 - the consequence
is a decrease in throughput.

MIMO R
Rx Spec Eff MCS
Layers BW Noise Spec MCS (Code Throughput
Power SNR Table cut Code
(MHz) (KTB) Efficiency Index rate/ (Mbps)
(dB) off Rate
1024)
4 50 -97.28 -96.83 -0.44 0.927 0.8770 3 449 0.438 88.76
8 50 -100.29 -96.83 -3.45 0.537 0.3770 1 193 0.188 73.11
Table 4: Rx Power, Noise, SNR, Spectral Efficiency obtained for each MIMO layer with Tx power set to 31 dBm.

3Refer Steps 5 through 8 in the section Analytical estimation of Data throughput, to understand how spectral
efficiency is got from SNR, and then how MCS is set from spectral efficiency.
Exercises
1. Estimate the data throughput analytically for different values of Transmit power,
Bandwidth and MIMO layer count (Each student can be given a personalized
experiment)