On the System-level Performance of Coordinated Multi-point

Transmission Schemes in 5G NR Deployment Scenarios

Siva Muruganathan, Sebastian Faxér, Simon Järmyr, Shiwei Gao, and Mattias Frenne
Ericsson
Email: {siva.muruganathan, sebastian.faxer, simon.jarmyr, shiwei.gao, mattias.frenne}@ericsson.com

Abstract—This paper investigates the system-level perfor-

mance of dynamic point selection (DPS) and non-coherent joint
transmission (NC-JT) coordinated multi-point transmission
(CoMP) schemes under different 5G NR deployment scenarios
using state-of-the-art system-level simulations. It is observed
that at a mid-band carrier frequency, NC-JT does not provide
performance gains over DPS or single transmission/reception
point (TRP) transmission unless the channel from a TRP is rank
deficient. Therefore, benefits with NC-JT are more likely to be
found in indoor deployment scenarios where the TRPs are typi-
cally equipped with only 2 transmit antenna ports, whereas ben-
efits are less likely to be observed in macro-cell deployments
where TRPs typically have a larger number of antennas ports.
It is further observed that NC-JT gains tend to diminish with
increasing system load, partly due to increased interference,
which typically lowers the transmission rank. Figure 1: Illustration of single-TRP transmission

I. INTRODUCTION interval (TTI) basis. This enables the network to serve the UE
In recent years, the demand for mobile data traffic has by the TRP which momentarily offers the best channel con-
grown at an unprecedented and accelerating rate, approach- ditions to the UE, or to perform dynamic load balancing be-
ing a year-on-year growth in data traffic of almost 100% [1]. tween the TRPs. Support for DPS was enabled already in
This increasing demand for traffic has been one of the key LTE Rel-11 with the introduction of the quasi-co-location
drivers for the development of fifth-generation (5G) cellular (QCL) property between reference signals and multiple chan-
networks and in particular the 3GPP 5G New Radio (NR) nel state information (CSI) processes [5].
standard, with its first release (Rel-15) completed in June Another related CoMP feature is dynamic point blanking
2018 [2]. Compared to the 4G LTE-A standard, 5G NR brings (DPB), where a joint multi-TRP scheduler can make a dy-
improvements in user experienced data rate, network capac- namic decision to not schedule any UEs from one or more
ity, latency, energy efficiency and reliability [3]. One princi- TRPs, thereby reducing interference to the UEs served by the
pal component of the 5G NR standard is the use of multi-in- remaining TRPs, as is illustrated in Figure 2b. Typically, DPS
put-multi-output (MIMO) transmission techniques for im- and DPB are operated simultaneously and correspond to dif-
proved coverage and spectral efficiency, both in traditional ferent scheduling hypotheses of a single joint scheduler.
sub-6 GHz and in millimeter wave (mmWave) deployments.
While for DPS/DPB, the UE is only receiving transmis-
The 5G NR standard will continue to evolve and improve. sions from a single TRP at a given time, in joint transmission
In the ongoing 3GPP Rel-16 work, enhancements to NR (JT), the UE can receive transmissions from multiple TRPs
MIMO functionality is considered [4]. One candidate scheme simultaneously. JT can further be broken down into two fla-
for improving spectral efficiently is the use of coordinated vors, non-coherent JT (NC-JT) and coherent JT (C-JT). For
multi-point (CoMP) transmission. In traditional cellular net- C-JT, the same precoding layer(s) are transmitted from the
works, the user equipment (UE) is only connected to a single multiple cooperating TRPs with the intention for the trans-
transmission/reception point (TRP) of the network at a time missions to coherently add up at the receiving UE, as is illus-
and each TRP makes independent scheduling, precoding and trated in Figure 2d. In order to accomplish this, tight synchro-
resource allocation decision, as is illustrated in Figure 1. For nization is required between the TRPs and a high CSI accu-
CoMP, on the other hand, multiple TRPs cooperate and coor- racy is needed in order to design the precoding weights,
dinate their transmissions such that a UE can receive trans- which makes C-JT difficult to practically implement and re-
missions from multiple TRPs simultaneously. quires large CSI feedback overhead.
Several different CoMP schemes are being considered, In NC-JT, on the other hand, the requirement for synchro-
each with different characteristics in terms of achievable per- nization and CSI accuracy is comparably lower, since each
formance and demand for coordination. The baseline Rel-15 TRP is transmitting different layers, as is illustrated in Figure
NR scheme uses dynamic point selection (DPS), where a UE 2c. Limited support for NC-JT was introduced in the feCoMP
can be dynamically scheduled to be served by one of multiple work item (WI) for LTE Rel-15 [6] and is further considered
TRPs in a coordination cluster, as illustrated in Figure 2a. also for NR in the Rel-16 eMIMO WI. Currently, there are no
Such dynamic scheduling can be on a per transmission time plans in 3GPP to introduce support for C-JT for LTE or NR.
 a) Dynamic Point Selection (DPB) b) Dynamic Point Blanking (DPB)

c) Non-coherent Joint Transmission (NC-JT) d) Coherent Joint Transmission (C-JT)

Figure 2: Illustration of four CoMP transmission schemes

their small form factor. At the UE side, each UE is assumed

In this paper, the performance of DPS and NC-JT CoMP to be equipped with 4 receive antennas and an MMSE-IRC
schemes are evaluated and analyzed in a 5G NR context with receiver. Note that with 4 receive antenna ports, the UE is
realistic system-level simulations using state-of-the-art 5G capable of receiving up to 4 MIMO layers, i.e., a maximal
channel models and deployment scenarios. The remainder of rank of 4. With NTx = 2, a TRP can only transmit up to 2
the paper is organized as follows: Section II gives a descrip- MIMO layers or maximal rank 2, thus there is a rank gap be-
tion of the investigated scenarios, Section III presents evalu- tween what the UE is capable of receiving and what a single
ation results, and Section IV provides concluding remarks. TRP can transmit. This case is referred to as “rank deficient”
induced by available number of transmit antennas. Note that
the channel can also be rank deficient due to the property of
II. 5G NR DEPLOYMENT SCENARIOS the channel, in, e.g., line-of-sight conditions.
This section describes four different 5G NR deployment
For evaluations involving DPS and NC-JT, the TRPs are
scenarios with different characteristics for which evaluation
partitioned into different coordination clusters wherein two
results are presented in Section III.
adjacent TRPs form one coordination cluster. It is assumed
A. Indoor Hotspot Scenario at 4 GHz Carrier Frequency that each UE is associated with one coordinated cluster. The
First, consider the indoor hotspot (InH) scenario at 4 GHz TRPs belonging to one coordination cluster are connected via
carrier frequency as defined in 3GPP technical report ideal backhaul (i.e., with zero latency). This enables coordi-
TR38.802, section A.2 [7]. This scenario consists of a de- nated scheduling with a single scheduler per coordination
ployment with 12 ceiling mounted TRPs that are placed in a cluster. A UE can hence be scheduled from either one of the
rectangular area of 120m × 50m. TRPs (using DPS) or both TRPs (using NC-JT) in the coor-
dination cluster based on, e.g., proportional fairness (PF)
Each TRP is assumed to be equipped with NTx transmit scheduling metrics.
antenna ports that are facing down from the ceiling. For the
evaluation results in Section III, results for NTx = 2 and NTx =
4 transmit antenna ports per TRP are provided. Note that
these numbers of antenna ports are typical for InH TRPs in
mid-band carrier frequencies (for example, 4 GHz) due to
B. Dense Urban Macro Scenario at 4 GHz Carrier The coordination cluster setup is like Section IIB where
Frequency three macro TRPs belonging to the same macro site form one
Consider next the dense urban (DU) macro scenario at 4 coordination cluster.
GHz carrier frequency as defined in 3GPP technical report
TR38.802, section A.2 [7] although here only the macro layer
III. SYSTEM LEVEL EVALUATON RESULTS
is modelled. In our evaluations, 19 macro sites were assumed
with each site consisting of 3 macro sectors (i.e., macro Evaluation results for the scenarios described in Section
TRPs). II are provided in this section. These results are generated us-
ing a system-level simulator where the traffic is generated us-
For the DU scenario, two different numbers of transmit ing 3GPP FTP Model 1 [8] with a packet size of 0.5 Mega-
antenna ports were considered: NTx = 2, and NTx = 4. At the bytes.
UE side, each UE is assumed to be equipped with 4 receive The performance is measured in terms of mean and cell
antennas and an MMSE-IRC receiver. edge user-perceived throughput (UPT), where cell edge UPT
When evaluating the DPS and NC-JT schemes, it was as- is defined as the 5th percentile UPT, and is presented as rela-
sumed that the macro TRPs belonging to the same macro site tive gains in mean and cell edge UPT with respect to a single
are connected via ideal backhaul and comprise one coordina- TRP baseline (i.e., no CoMP). The performance is evaluated
tion cluster, and that each UE is associated with one coordi- for different system loads in the baseline system. The system
nation cluster. Due to the ideal backhaul assumption, it is loads are characterized by average percentage of system re-
possible to use a single scheduler per coordination cluster, source utilization (RU) across all TRPs. In these evaluations,
where the scheduler is responsible scheduling data to the UE 10%, 20%, and 40% system RU of the baseline scheme are
using either DPS or NC-JT. considered.

C. Indoor Hotspot Scenario at 30 GHz Carrier Frequency A. Indoor Hotspot Results at 4 GHz Carrier Frequency
The third scenario considered in our evaluations is the The results for InH at 4 GHz carrier frequency are given
InH scenario at 30 GHz carrier frequency where analog, or in Figure 3 and Figure 4. From Figure 3, it is observed that
time domain, beamforming is assumed at the transmitter and with NTx = 2 transmit antenna ports per TRP, NC-JT can
the receiver. The placement of TRPs for this scenario is like achieve notable mean throughputs gains over both single TRP
that described in Section IIA. transmission and DPS schemes at 10% RU. The mean
throughput gain of NC-JT is moderate at 20% RU, and be-
Each TRP is assumed to be equipped with a 4x4 cross- yond 20% RU NC-JT performs poorly when compared to
polarized antenna array capable of generating 16 analog both single-TRP transmission and DPS.
beams for transmission. It is assumed that NTx = 2 transmit
antenna ports are used per analog beam. The UEs are as- In contrast to the results in Figure 3, the results from Fig-
sumed to be equipped with two panels facing in opposite di- ure 4 with NTx = 4 transmit antenna ports per TRP indicate
rections. Each of the panels at the UE is assumed to contain that NC-JT performs poorly when compared to both DPS and
a 2x4 cross polarized antenna array with 8 analog beams. single-TRP transmission for all RU values.
Each of the analog beams associated with one panel has 2 re- The main reason for poor performance of NC-JT is likely
ceive antenna ports. Hence, there are 4 receive antenna ports that with NTx = 4 transmit antenna ports per TRP, a maximum
in total across two analog beams received using the two pan- rank (or number of MIMO layers) of 4 can already be
els. In addition, the UE is assumed to be equipped with a achieved from a single TRP and using NC-JT does not pro-
MMSE-IRC receiver. vide any benefits. Conversely, using NC-JT results in data
The coordination cluster setup is similar to Section IIA being transmitted from two TRPs which increases the inter-
where two adjacent TRPs form one coordination cluster. ference in the system which leads to poor performance of NC-
JT. With NTx = 2 transmit antenna ports per TRP, however,

D. Dense Urban Macro Scenario at 30 GHz Carrier

Frequency
The last scenario considered in our evaluations is the DU
scenario at 30 GHz carrier frequency where analog beam-
forming is assumed at the transmitter and the receiver. The
placement of TRPs for this scenario is like that described in
Section IIB, with the exception that 7 sites are considered in-
stead of 19.
In this scenario, each TRP is assumed to be equipped with
an 8x8 cross-polarized antenna array with 32 analog beams
available for transmission. Each analog beam is assumed to
contain NTx = 2 transmit antenna ports. The UEs are assumed
to be equipped with two panels facing in opposite directions
with a setup like that described in Section IIC. Figure 3. Results for InH at 4 GHz carrier frequency with NTx = 2
transmit antenna ports per TRP and 4 receive antenna ports at UE.
Figure 4. Results for InH at 4 GHz carrier frequency with NTx = 4
Figure 6. Results for DU at 4 GHz carrier frequency with NTx = 4
transmit antenna ports per TRP and 4 receive antenna ports at UE.
transmit antenna ports per TRP and 4 receive antenna ports at UE.

the transmission from a single TRP is rank deficient and is C. Indoor Hotspot Results at 30 GHz Carrier Frequency
limited to a maximum rank of 2. In this case, transmission
The results for InH at 30 GHz carrier frequency are given
from two TRPs using NC-JT can increase the maximum rank
in Figure 7. From these results, it is noted that NC-JT pro-
to 4 which can be beneficial at low loads where the interfer-
vides similar cell-edge performance to that of DPS. However,
ence tends to be low.
NC-JT provides a fair mean throughput gain over DPS. The
reason is that DPS is limited to a maximum rank of 2 since
B. Dense Urban Results at 4 GHz Carrier Frequency only 2 transmit antenna ports are possible per analog beam
The results for DU at 4 GHz carrier frequency are given transmitted from a single TRP. On the other hand, by trans-
in Figure 5 and Figure 6. mitting from two TRPs, NC-JT is able to deliver up to a max-
imum rank of 4 to a UE which results in the mean throughput
Similar to the observations in Section IIIA, with NTx = 2 gain of NC-JT over DPS. It should however be emphasized
transmit antenna ports per TRP in the DU scenario, NC-JT that NC-JT gains tend to diminish with increasing RU. This
has some mean throughput gain over single TRP transmission is because a higher RU leads to more interference in the sys-
and DPS at 10% RU. However, above 20% RU, NC-JT per- tem which typically lowers the desired transmission rank, and
forms poorly compared to both single TRP transmission and hence also lowers the potential benefit of NC-JT.
DPS.
With NTx = 4 transmit antenna ports per TRP in the DU
scenario, NC-JT does not provide any mean throughput gains
over DPS and single TRP transmission. This result is also
consistent with the observation in Section IIIA.

Figure 5. Results for DU at 4 GHz carrier frequency with NTx = 2

Figure 7. Results for InH at 30 GHz carrier frequency.
transmit antenna ports per TRP and 4 receive antenna ports at UE.
D. Dense Urban Results at 30 GHz Carrier Frequency 10% RU 20% RU 40% RU
80
The results for DU at 30 GHz carrier frequency are given

Performance Gain over single TRP [%]

in Figure 8. Similar to the results in Section IIIC, it is ob- 70

served that NC-JT provides a fair mean throughput gain over 60

DPS. This again is due to NC-JT being able to deliver up to a 50

maximum rank of 4 to a UE while DPS is limited to a maxi- 40
mum rank of 2.
30

20
IV. CONCLUSIONS 10

In this paper, the performance of NC-JT under different 0

5G NR deployment scenarios were evaluated. From the eval- Cell-edge Throughput gain of DPS Cell-edge throughput gain of NC-JT
uation results, the following are observed: Mean throughput gain of DPS Mean throughput gain of NC-JT
At a mid-band carrier frequency of 4 GHz, NC-JT does not Figure 8. Results for DU at 30 GHz carrier frequency.
provide performance gains over DPS and single-TRP trans-
mission when the single TRP channel is not rank deficient,
i.e., each TRP can provide the maximum rank that a UE can
receive. However, if each TRP is equipped with a smaller REFERENCES
number of transmit antenna ports than the number receive an- [1] “Ericsson mobility report November 2018”. Ericsson
tennas ports available at the UE, or if line-of-sight probability
[2] 3GPP TS38.300 “3rd Generation Partnership ProjectTechnical
is high, some benefits may be possible for NC-JT over DPS Specification Group Radio Access Network; NR; NR and NG-
and single-TRP transmission. RAN Overall Description; Stage 2 (Release 15)”
At a mid-band carrier frequency of 4 GHz, most of the NC-JT [3] R1-1806431, An overview of the IMT-2020 Evaluations,
Ericsson, 3GPP TSG-RAN WG1 Meeting #93, Busan, Korea,
benefit for NC-JT are found in indoor scenarios where the May 21st – 25th 2018
TRPs are most likely to be equipped with 2 transmit antenna [4] RP-182067, “Revised WID: Enhancements on MIMO for NR”,
ports while the UEs are equipped with 4 receive antenna ports. Sasmung, 3GPP TSG RAN Meeting #81, Gold Coast,
Australlia, Sept 10 - 13, 2018
At a high-band carrier frequency of 30 GHz, NC-JT outper-
[5] RP-101425, “Revised SID Proposal: Coordinated Multi-Point
forms both DPS and single-TRP transmission as both of these Operation for LTE”, Samsung, 3GPP TSG RAN#50
schemes are limited to a maximum transmission rank of 2. On , Istanbul, Turkey, 7th – 10th December, 2010
the other hand, by transmitting from two TRPs, NC-JT is able [6] RP-170750, “New WID: Further enhancements to Coordinated
of delivering up to a maximum rank of 4. Multi-Point (CoMP) Operation for LTE”, Intel, ZTE, 3GPP
TSG RAN Meeting #75, Dubrovnik, Croatia, March 6 - 9, 2017
NC-JT gains tend to diminish with increasing system load as [7] TR38.802. “3rd Generation Partnership Project;Technical
higher system load leads to higher interference which typi- Specification Group Radio Access Network;Study on New
cally lowers the desired transmission rank. Radio Access TechnologyPhysical Layer Aspects (Release
14)”
[8] 3GPP TR 36.814, “3rd Generation Partnership
Project;Technical Specification Group Radio Access
Network;Evolved Universal Terrestrial Radio Access (E-
UTRA);Further advancements for E-UTRA physical layer
aspects(Release 9)”