Removing the PDCCH Bottleneck and Enhancing

the Capacity of 4G Massive MIMO Systems

Pavan Reddy M., Harish Kumar D., Saidhiraj Amuru, Kiran Kuchi
Department of Electrical Engineering, Indian Institute of Technology Hyderabad, India 502285
Email:{ee14resch11005, ee14mtech11003, asaidhiraj, kkuchi}@iith.ac.in

Abstract—The Full Dimension-MIMO (FD-MIMO) technol-

ogy is capable of achieving huge improvements in network
throughput with simultaneous connectivity of a large number
of mobile wireless devices, unmanned aerial vehicles, and the
Internet of Things (IoT). With large antennae at the base station,
an enhanced multi-user transmission is performed to increase
the capacity of physical downlink shared channel (PDSCH).
However, the current 3rd Generation Partnership Project (3GPP)
specifications of Long Term Evolution (LTE) do not allow
the base station to perform any beamforming techniques for
the physical downlink control channel (PDCCH). The PDCCH
capacity dictates the number of users scheduled in a subframe,
and thus, the PDCCH limits the cell capacity. Motivated by this,
we propose a novel beamformed PDCCH (BF-PDCCH) design
for LTE which strictly adheres to the 3GPP specifications. With
just software-based changes at the base station, the proposed
design significantly enhances the PDCCH capacity. The proposed
design uses the sounding reference signals (SRS) transmitted in Fig. 1: EPDCCH configuration in a subframe
the uplink to decide the best beam for a user, and then ingeniously
schedules the users in PDCCH. Since the performance of the
proposed BF-PDCCH heavily relies on this SRS reception, we
model the SRS channel estimation errors while evaluating the kind of beamforming, 2-3.6 times gain in the cell throughput
performance of the proposed BF-PDCCH. Through system-level is achieved [1].
simulations, we show that the proposed design achieves significant
network throughput, and outperforms the current state of art In Long Term Evolution (LTE), the downlink physical layer
algorithms and also the existing 3GPP schemes. has five channels [3–5]. They are physical broadcast channel
(PBCH) for broadcasting the system information, physical
Keywords: Beamforming, blind decoding, control channel, PDCCH, control format indicator channel (PCFICH) for defining the
search space, SRS.
structure of the control channel, physical HARQ indicator
channel (PHICH) for conveying the ack/nack, physical down-
I. I NTRODUCTION link control channel (PDCCH) for carrying the control infor-
Full Dimension-Multi Input Multi Output (FD-MIMO) is mation and physical downlink shared channel (PDSCH) for
a key technology in achieving larger network throughputs by transmitting the user intended data. In this paper, we focus
simultaneously connecting a large number of devices. This on PDCCH, which carries the downlink control information
has been an active topic in the standardization activities of (DCI). DCI conveys the information required to decode the
3rd Generation Partnership Project (3GPP). In FD-MIMO, a user intended data. The PDCCH region in any subframe is
two-dimensional antenna array structure is used that helps limited to 3 symbols [5] and hence, can accommodate a
in beamforming along both elevation and azimuth directions. limited number of DCIs in a transmission time interval (TTI).
With this kind of beamforming, an enhanced multi-user MIMO Thus, the PDCCH effectively indicates the number of users
transmission can be done at the base station to achieve a scheduled in any TTI.
multi-fold enhancement in the network throughput [1]. From In Release 8 3GPP specifications, PDCCH and PDSCH rely
Release 8, 3GPP has continuously evolved its specifications to on cell-specific reference signals (CRS) for the channel esti-
enhance the multi-user MIMO feature and thus, enable a large mation. From Release 13, the PDSCH supports beamforming
number of users to be supported by the base station [2]. From for specific users, and whenever it is beamformed, PDSCH
Release 13, 3GPP specifications support both the azimuth and carries DMRS. Both the data and the DMRS observe the same
elevation beamforming for the data channel. Based on a newly beamforming, and hence the user observes a minimal impact
introduced channel state information-reference signals (CSI- in channel estimation. However, PDCCH is not beamformed
RS) and the demodulation reference signals (DMRS), the base and has to be decoded based on the CRS. CRS is common
station performs beamforming on the data channel. With this for all the users, and beamforming CRS would impact the
performance of cell search and synchronization. Thus, with
the current 3GPP specifications of LTE, the control channel
does not possess the benefits of beamforming. Note that a
user can decode the data channel only after decoding a DCI.
Thus, even though the beamforming allows to schedule more
users in PDSCH, the PDCCH has a limited capacity and has
become a bottleneck in increasing the network throughput. In
Release 11, to enhance the PDCCH capacity, 3GPP introduced
enhanced PDCCH (EPDCCH) design which uses the concepts
of beamforming. However, the EPDCCH has to be transmitted
in the resources of the data channel, as shown in Fig. 1.
Further, the location of the EPDCCH has to be conveyed
apriori to the user.
The availability of the large antennae structure with the
FD-MIMO is never exploited in the context of the PDCCH.
This is because beamforming requires some feedback from
the user. However, the control channel itself is the first
communication link where a user performs blind decoding
for the DCI. Improving the PDCCH capacity by exploiting
the large antennae structure has a high impact on network
throughput and has never been considered in the literature.
Hence, in this paper, we propose a novel beamformed PDCCH
(BF-PDCCH) design for LTE, which addresses all the above-
said issues.
The rest of the paper is organized as follows. In Section II,
we present the motivation and summary of contributions of
our work. Section III presents some of the related work in
the literature. In Section IV, we explain the current physical
downlink control channel structure and the enhancements as
per 3GPP specifications. Section V presents the proposed BF-
PDCCH design . In Section VI, we explain the algorithms for
the implementation and evaluation of the proposed scheme.
In Section VII, we present the simulation model and discuss
the numerical results. Some concluding remarks and possible
future work are presented in Section VIII.
II. M OTIVATION AND C ONTRIBUTIONS
Currently, there has been a wide commercial deployment of
massive MIMO across the world [6], [7]. As reported in [7],
the LTE massive MIMO achieves 2-3 times gain in the network
capacity on the field. Removing the control channel bottleneck
and enhancing the control channel capacity could thus further
maximize the network capacity. Motivated by this, we propose
a novel beamformed PDCCH design in this paper.
The main contributions of this paper are as follows:
• We propose the BF-PDCCH design for LTE massive
MIMO and present the implementation procedure for the
same. The features of the proposed design are as follows.
– The proposed design strictly adheres to the 3GPP
specifications and requires no changes at the user.
– Unlike EPDCCH, the proposed design does not use
the resources from the data channel.
• Through system-level simulations, we show that the
proposed BF-PDCCH outperforms all the current 3GPP
mechanisms.
• Further, we model channel estimation errors in the sim-
ulator. We show that the proposed BF-PDCCH still
achieves significant gains compared to the current state
of art algorithms and the existing 3GPP schemes.
A. Difference from Release-15
5G-New Radio (5G-NR) is a new technology introduced by
3GPP in Release 15 specifications. In the Release 15, 5G-NR
supports beamforming of the synchronisation, the broadcast,
and the control channels. During the initial cell search, the
user gets locked to a specific beam from the base station.
This locked beam information is used by the base station to
spatially multiplex the subsequent control and data channel
information of different users across various beams.
In LTE, the 3GPP specifications do not allow the base
station to perform beamforming for the synchronisation, the
broadcast, and the control channels. With the availability of
large antenna structure and the beamforming features, the base
station is capable of scheduling more users in the data channel.
However, the base station has to convey control information
for all these users, and note that the control channel is not
beamformed in LTE. Thus, the control channel limits the
cell capacity in LTE massive MIMO. With just software
changes to the existing 4G-LTE technologies and deployments,
our proposed BF-PDCCH could remove the control channel
bottleneck and maximize the massive MIMO capacity.
Next, we present some related work in the literature.
III. R ELATED W ORK
In [8], authors have presented an algorithm to optimally
schedule the users in PDCCH and thus, increase the control
channel capacity. In [9], authors have proposed a novel method
of allocation for cell radio network temporary identifiers and
increase the control channel capacity. In [10], authors propose
power allocation techniques to improve the control channel
capacity. However, none of the above papers discussed the
beamforming and exploited the large antennae structure for
increasing the control channel capacity. Further, we have
implemented the optimal LTE-PDCCH scheduling algorithm
presented in [8] and compared its performance against the
proposed BF-PDCCH design in Section VII.
In [11], [12], novel search space designs of EPDDCH have
been presented to improve the capacity of the control channel.
In [13], [14], authors have presented the performance and
analysis of the EPDCCH design. In [15], the authors have
proposed an algorithm to improve the channel estimation
accuracy and thus, in turn, improve the performance of the
EPDCCH. However, as per the 3GPP specifcations [3–5], the
EPDCCH uses the resource elements from the data channel
for beamforming. To the best of our knowledge, none of the
papers in the literature have addressed the issue of increasing
the control channel capacity by exploiting the large antennae
structure.
Next, we explain the current control channel design and its
enhancements as per 3GPP specifications.

Fig. 2: Search space design of the current 3GPP PDCCH
IV. 3GPP P HYSICAL D OWNLINK C ONTROL C HANNEL
The PDCCH is present in the first few orthogonal frequency
division multiplexing (OFDM) symbols of every subframe.
The first OFDM symbol has PCFICH, PHICH, and PDCCH
multiplexed in it. PCFICH defines the number of symbols
for PDCCH. DCI is the payload transmitted in PDCCH. DCI
carries the information required for decoding the user data,
location of uplink scheduling, modulation and coding scheme,
and random access responses. There are various DCI formats
for each purpose. Before transmission, the DCI payload along
with cyclic redundancy check parity bits is convolution en-
coded. The convolution coded data is then rate-matched to a
certain number of bits called aggregation level (AL). These
rate-matched bits are QPSK modulated and multiplexed in the
radio frame.
The multiplexing of PDCCH data in the radio frame is
performed as follows. Resource element (RE) is the smallest
unit of time-frequency resource in LTE. Excluding the PC-
FICH and PHICH resource elements, the remaining resource
elements available are grouped in the number of four (in a
frequency-first and time-next manner) and called as Resource
Element Groups (REGs). A collection of nine such REGs
forms one control channel element (CCE). In LTE PDCCH,
the base station schedules the DCIs in the units of CCEs.
Based on the channel conditions of the user, the payload is
rate matched to an aggregation level (AL). The data in one
AL can fit into one CCE. In LTE PDCCH, the AL∈{1,2,4,8}
and thus, AL=2 requires 2 CCEs and so on.
The user does not know the exact location of the DCI
in PDCCH and hence, has to perform blind decoding over
the PDCCH region. To reduce these blind decoding attempts,
3GPP limits the region (or CCEs) over which the user has
to perform blind decodings. This region (or CCEs) is termed
as search space. There are two search spaces in the control
channel region.
• Common Search Space (CSS): The base station schedules
the DCIs which address more than one user in a specific
region called CSS. These DCIs carry system information,
paging, and group scheduling related information.
• UE specific Search Space (USS): The base station sched-
ules the DCIs intended for a particular user in USS.
The number of CCEs in a PDCCH varies with the number
of symbols over which PDCCH is present and the operating
bandwidth. In any subframe, irrespective of the number of
CCEs available, the CSS is present only in the first 16 CCEs.
The search space design as per the current 3GPP specifications
is depicted in Fig. 2.
As per 3GPP specifications, the base station can schedule
the user in limited fixed CCEs. This limitation ensures a
further reduction in the blind attempts within the search space.
The base station calculates the indices for scheduling a DCI
based on the AL and user identity using (1). The base station
can transmit the DCI for a user only in those possible CCE
locations [3].
CCEindex = L{(Yk + m)modbNCCE,k /Lc} + i (1)
where, i = 0 . . . L − 1, L is aggregation level, NCCE,k is
number of CCEs in the subframe k, Yk and m are the
constants defined by the higher layer parameters in 3GPP
specification [16]. Note that the user has no information about
the location and the aggregation level of the DCI. Hence, the
user calculates all the possible indices and blindly performs the
search only at these locations. The user repeats this procedure
for all the aggregation levels until it decodes a DCI.
In any subframe, based on the available bandwidth, there
are a limited number of CCEs in PDCCH. This limitation has
an impact on multi-user scheduling in PDSCH. In Release 11,
3GPP has introduced enhanced PDCCH design to increase
the PDCCH capacity. The EPDCCH is transmitted in the data
channel region, as shown in Fig. 1. The search space region
for monitoring the DCI in EPDCCH is conveyed to the user
prior through higher layer signaling. The advantage of the
EPDCCH is that it can use beamforming concepts like data
channel and thus, schedule more number of DCIs. However,
this comes at the cost of sacrificing the PDSCH resources.
Note that the EPDCCH has demodulation reference signals to
decode the beamformed data. In this paper, we propose a novel
BF-PDCCH design which does not use any PDSCH resources
and yet achieves improvement in network capacity. Next, we
present the design of the proposed BF-PDCCH scheme.
V. P ROPOSED B EAMFORMED PDCCH D ESIGN
In [17], we present a detailed explanation on implemen-
tation of beams. Readers unfamiliar with the beam design
and implementation are suggested to read the article [17]. In
multi-user MIMO, the transmission is done simultaneously in
each beam to achieve multi-fold improvement in the network
throughput. Since these beams are spatially well separated,
there is minimal interference between each other. This way,
multi-fold improvement is achieved in the network throughput.
In this paper, we consider P beams are active in the sector
all the time to implement multi-user MIMO. The primary
synchronization signal (PSS), the secondary synchronization
signal (SSS), PBCH, and PDCCH assume same channel
characteristics as they rely on CRS. Hence, all of these have to
behave similarly in terms of spatial configuration [3, Section
6.8.4]. Thus, if PDCCH has to be beamformed, then it forces

Fig. 3: Search space design of the proposed BF-PDCCH
CRS, PBCH and PSS/SSS also to be beamformed. Considering
all these constraints, we propose a PDCCH beamforming
design which can improve the PDCCH capacity.
A. Proposed Design
The design of the proposed beamformed PDCCH is as
follows. Consider P beams active in the sector all the time.
All the common signals PSS, SSS, PBCH, PCFICH, PHICH,
and CRS, are transmitted in all the beams. Thus, both the CRS
and the common signals observe a similar channel (a detailed
mathematical explanation for the same is presented in [17] and
is shown with system-level simulations in the Section VII).
Hence, there is no impact on CRS channel estimation based
reception for common signals. As mentioned earlier, PDCCH
has two search spaces, CSS and USS. CSS has to address
all the users present in the sector. Hence, CSS is transmitted
exactly at same CCEs across all the beams. Since USS is
specific to a particular user, the USS for each user can be
beamformed. To beamform USS, a mechanism is needed to
identify the best beam for each user. We identify the best beam
based on the sounding reference signal (SRS) transmitted by
the user in the uplink. The implementation procedure for
finding the best beams is presented in Section VI-A. Now
with the best beam assigned to each user, the USS is scheduled
differently in each beam. However, note that some users are
at the cell edge/beam edge and experience a large interference
with this dynamic scheduling of different data streams in each
beam. Also, during initial connection set-up with the base
station, the users need to decode the PDCCH before they can
even send the SRS. Hence, in all these scenarios, the USS is
not beamformed and is transmitted in all the beams. The USS
is beamformed only for the users whose SRS are available and
are at the boresight of the beam. Therefore, we split the USS
into two search spaces, USS - 1 and USS - 2. Thus, we divide
the search space for BF-PDCCH into three regions:
1) CSS: Irrespective of the available bandwidth CSS is
present in the first 16 CCEs of the PDCCH as in current 3GPP
specifications [5]. Since CSS is common to all the users in the
sector, it is present in all the beams.
2) USS-1: USS - 1 has DCIs addressing the users at cell
edge/beam-edge and hence, is present in all the beams.
3) USS-2: USS - 2 has the DCIs addressing the users at
boresight. The base station schedules different DCIs in each
beam, and thus, USS - 2 enhances the PDCCH capacity.
An illustration of the proposed BF-PDCCH design is shown
in Fig. 3.
B. Analysis of the Proposed Design
The main characteristics of the proposed BF-PDCCH are as
follows.
1) Beamformed Common Signals: To adhere to the 3GPP
specifications; if PDCCH is beamformed, it forces all the other
common channels to be beamformed. However, the users are
not aware of PSS/SSS, common channels, and the PDCCH
being beamformed. They decode all the channels as per the
3GPP specifications.
2) Impact of beamforming: There are P beams active in the
sector all the time, and these P beams are designed to cover
the entire sector. Note that the same data is present in all the
beams for the common channels. Thus, there is minimal effect
on the performance of the common channels because of the
beamforming. Detailed mathematical reasoning for the same
is present in [17].
3) 3GPP compliant: The CSS and the USS regions of the
proposed BF-PDCCH do not deviate from the current 3GPP
specifications.
4) Multi-user scheduling: The base station can schedule a
user in USS - 2 only after it receives the SRS from that user.
Until then the base station schedules the DCI for the user in
USS - 1.
5) No extra signaling: Unlike EPDCCH, there is no extra
signaling for the user to indicate where it has to look for the
DCI. Only the base station has the notion of USS - 1 and
USS - 2, and it intelligently schedules the users in USS - 1
and USS - 2.
6) Users are unaware of the USS split: The users are not
aware of the PDCCH being beamformed. They consider the
region of USS - 1 and USS - 2 as a regular USS in the legacy
LTE PDCCH and do the blind decoding similarly to what they
do in the legacy LTE PDCCH.
Next, we present procedures for the implementation and
evaluation of the proposed BF-PDCCH design.
VI. P ROCEDURES FOR I MPLEMENTATION AND
EVALUATION OF BF-PDCCH
We initially present the algorithm for the implementation of
proposed BF-PDCCH. The SRS channel estimates in the real-
time are not perfect and have an impact on the performance
of the proposed design. Hence, we consider these estimation
errors during the evaluation of the proposed design. For this
purpose, we present the procedure for the abstraction of the
SRS channel estimation errors to use them in system-level
simulations.
A. Algorithm for implementation of the proposed BF-PDCCH
The user transmits SRS in the uplink. The base station (with
M antenna elements in horizontal and N antenna elements in
 Fig. 5: Pictorial representation of user scheduling in 1 and 2
best beams respectively

in legacy LTE PDCCH, the users need 8 CCEs. In the case of

BF-PDCCH, they require 4 CCEs each when scheduled in a
single best beam. However, the base station schedules the users
across 8 beams using just 4 CCEs in total. Note that if the users
require AL=1, 2 with BF-PDCCH, then the multiplexing gain
Fig. 4: Flow chart depicting the implementation of the pro- increases further, and when the users require AL=8 with BF-
posed scheme PDCCH, the gain completely disappears. Hence, with eight
beams active in the sector, we consider the users with a
maximum of AL=4 for USS - 2.
vertical) receives the channel coefficients H (of size M × N ) 2) Identifying best 2 beams: If the SIN Rj1 < SIN RAL=4 ,
from each user. We then calculate the beam weights (Wi ) for then we check the two best beams for the user. We define
each active beam, as mentioned in [17]. SIN Rj2 as the signal to interference-plus-noise ratio when
1) Identifying single best beam: We define SIN Rj1 as the the base station schedules the user in two beams ([j, j + 1]).
signal to interference-plus-noise ratio when the base station Note that the user observes interference from all the remaining
schedules the user in j th beam. Note that when a base station beams in the sector.
schedules a user in one beam, it observes the interference from
P
2
all the other beams. Thus, we formulate SIN Rj1 as follows.
j+1

k=j HWk



2
2
SIN Rj =
P
2 , ∀j = 1, . . . , P −1
kHWj k
P
2
k=1,k6=[j,j+1] HWk
+ σn2


SIN Rj1 =
P
2 , ∀j = 1, . . . , P (2)

P (3)
2
k=1,k6=j HWk
+ σn2


We follow a similar procedure like earlier and choose best
For each user, we calculate the SIN Rj1 across all the active beams [j, j + 1] such that SIN Rj2 is maximum. The user
beams. Then we choose the j for which SIN Rj1 is maximum is then scheduled in those two best beams if SIN Rj2 >
as the best beam for the user. For this chosen j, SIN Rj1 is SIN RAL=4 .
compared against the minimum SINR required for AL = 4 Note that in both the above cases, the base station schedules
in legacy LTE case. If SIN Rj1 > SIN RAL=4 , then user is the users in USS - 2. If the SIN Rj2 ≤ SIN RAL=4 , then the
scheduled only in j th beam. The reason for this is explained user is assumed to be in cell-edge/beam-edge. For these users,
with an example as follows. Consider the following worst- the base station schedules the data in all the beams in USS - 1.
case scenario where eight users require AL=1 in legacy LTE All the above procedure is presented as a flow chart in Fig. 4,
PDCCH scenario, and the same eight users need AL=4 in BF- and a pictorial representation of user scheduling is shown in
PDCCH scenario with eight beams (P = 8). When scheduled Fig. 5.
 TABLE I: System level simulation parameters
Parameter Value
Cell layout 7 cell sites, 3 sectors/site
Inter-site distance 500 m
BS antenna height 25m
UE antenna height 1.5m
Carrier frequency 2.4 GHz
BS transmit power 44 dBm
2)
SRS estimation error (σE 5 dB
64, 2 (BF-PDCCH),
Number of antennae (BS, UE)
1, 2 (LTE PDCCH)
Bandwidth 20 MHz
Channel model 3D UMa in TR 36.873 [19]
−3π −π π 3π
Azimuth:[ , , , ]
Direction of selective beams 16 16 16 16
9π 11π Fig. 6: Variation of SINR of users with proposed and the
Elevation:[ , ]
16 16 existing 3GPP schemes
Available DL CCEs 42 (with CFI=3 OFDM symbols)
Aggregation levels {1, 2, 4, 8}
Extra PRBs for EPDCCH 4 PRBs
BS antenna radiation pattern According to TR 36.873 [19] Table I. We have modeled the link abstractions for allocating
AL based on the SINR of each user. We have also considered
the SINR degradation happening because of the CRS re-use.
B. SRS Channel Estimation errors In our simulations, we implement all these link and channel
In reality, the SRS channel estimations may differ from the abstractions, as per the procedures mentioned in [17]. For
actual channel observed by the base station. This difference comparing the performance of the proposed BF-PDCCH, we
can be due to the poor signal to noise power ratio of the implement the current 3GPP LTE PDCCH and EPDCCH
user, channel aging, and the estimation algorithm used by the schemes. For the EPDCCH case, we consider an extra four
base station. These incorrect channel estimates have an impact PRBs of resources available from the data channel. Further, we
on the beam selection procedure explained in Section VI-A. have implemented the optimal scheduling procedure presented
Hence, we have to consider these channel estimation errors in [8] for comparing the performance of BF-PDCCH.
in the simulator for observing the real-time gains with the The performance evaluation of all these schemes is carried
BF-PDCCH. We follow the abstraction procedure mentioned out as follows. The users are dropped randomly in a cell site.
in [18] for including the SRS estimation errors as follows. For the given parameters in the Table I, the SINR is calculated
for each user. While calculating the SINR in each scheme,
H̃ = α(H + E) (4) the interference from other sectors (Iwrap ) is modeled using
s wrap around algorithm mentioned in [20]. For the proposed
2 1
E ∼ N (0, σE ), α = 2
BF-PDCCH, when the base station schedules the data in
1 + σE one best beam, the user observes interference from all the
where H is the actual channel observed by the base station remaining P − 1 beams. Similarly, when the base station
(on M × N antenna elements) when the user transmits SRS schedules the data in two adjacent beams, the user observes
in uplink, H̃ is the channel estimated by the base station, E is interference from all the remaining P − 2 beams. Apart from
the additive white complex Gaussian noise with zero mean and the interference from the other sectors and other beams, the
σE2
variance. To include the SRS channel estimation errors, we user observes errors in the channel estimation because of CRS
use H̃ instead of H in the beam selection procedure explained re-use happening across the beams. For each scheme, using
in (2)-(3). the link abstractions [17], we map the ALs to each user based
Since the CRS is the only reference signal present across the on the calculated SINR. For a given bandwidth and number
beams, the effect of the channel re-use also has to be included of OFDM symbols, the available CCEs vary. We assume half
in the simulator for proper evaluation of the proposed design. the available CCEs carry downlink DCIs. Within this limited
In [17], authors present the modeling of these channel re-use number of CCEs, we schedule the users with different ALs.
abstractions along with various other link level abstractions. In Fig. 6, the SINR distribution of users in BF-PDCCH
We consider all these abstractions in our simulations while is shown and compared against that of the legacy LTE. The
evaluating the performance of the proposed BF-PDCCH. Next, highlighted region in Fig. 6 has the feasible SINR values where
we present the simulation results. the users require AL < 8. This region is calculated assuming
a maximum block error rate (BLER) of 0.01. If a user has
VII. S IMULATION R ESULTS AND D ISCUSSION the SINR in this highlighted region, then the base station can
The parameters for the system-level simulations are consid- spatially multiplex the other users in the other beams. Thus,
ered as per the 3GPP specifications [19] and are presented in we achieve the spatial multiplexing gain in BF-PDCCH. Note

Fig. 7: CDF plot of number of users scheduled/CCE
that, when we consider the SRS channel estimate errors, the
beam selection is not perfect. If the base station schedules the
users based on this imperfect beam selection, the user observes
a degradation in the SINR. Hence, when we include the SRS
estimation errors, the SINRs are low (black curves in Fig. 6)
compared to that without SRS errors (red curves in Fig. 6).
Further, even with SRS estimate errors, the base station can
schedule more than 40% of the users in a single beam and
more than 50% of the users in 2 best beams. In all these
cases, the BF-PDCCH gives spatial multiplexing gain.
In Fig. 7, the CDF of the number of users scheduled
per CCE is presented for various schemes. The number of
CCEs available is more in enhanced PDCCH because of the
extra four PRB resources. Hence, for a fair comparison, the
normalization is carried out with the available CCEs for each
scheme. We calculate the LTE minimum bound by scheduling
each user with the maximum AL(= 8) all the time. For
the LTE maximum bound, we schedule each user with the
minimum AL(= 1) all the time. The 3GPP LTE PDCCH
has the least performance among all the schemes. In optimal
scheduling algorithm [8], the user scheduling is done based
on a set-packing problem, and thus, the base station can
accommodate more users in the control channel region. Hence,
the optimal LTE-PDCCH scheduler has comparatively more
number of users scheduled per CCE. The EPDCCH has a
better performance compared to the 3GPP LTE PDCCH and
optimal LTE-PDCCH scheduler. However, note that EPDCCH
uses an additional four PRBs from the PDSCH resources
and also requires additional control signaling. The proposed
BF-PDCCH has better performance than all these schemes.
Further, when we consider channel estimation errors, there is a
drop in the number of users scheduled. However, even after the
inclusion of the channel estimation errors, the proposed BF-
PDCCH outperforms all the existing mechanisms and achieves
more gains than the LTE maximum bound.
The average number of users scheduled in a TTI and
the average BLER for each scheduling algorithm are pre-
sented in the Table II. Excluding the LTE minimum and
the maximum bounds, all the scheduling algorithms have an
TABLE II: Average users scheduled in a TTI and average
BLER with the proposed and current 3GPP mechanisms
Average Average
Scheme
users/TTI BLER
3GPP LTE PDCCH 22.1 ≈1%
Optimal LTE-PDCCH Scheduler [8] 22.6 ≈1%
Enhanced PDCCH 26.6 ≈1%
Proposed BF-PDCCH without SRS errors 50.3 ≈1%
Proposed BF-PDCCH with SRS errors 34.4 ≈1%
LTE Maximum Bound 42 ≈20%
LTE Minimum Bound 5 ≈ 0.1%
average BLER ≈ 1%. This is because, in all these schemes,
the base station schedules the users with an AL such that
BLER ≤ 0.01. However, for the LTE maximum bound, the
base station schedules each user with AL = 1, and hence
with a very high chance, the error occurs for the users
operating in poor signal coverage. In the LTE minimum bound,
the base station schedules each user with the best possible
AL(= 8). Thus average BLER is approximately 0.1% for the
LTE minimum bound and is approximately 20% for the LTE
maximum bound. In the proposed BF-PDCCH scheme, the
users experience a poor operating SINR and require large AL.
However, because of the multi-user MIMO, more number of
users can be scheduled. From the Table II and Fig. 7, it can
be observed that the proposed BF-PDCCH achieves significant
improvement in the network capacity.
VIII. C ONCLUSION AND F UTURE W ORK
We proposed a novel beamforming design for the con-
trol channel of LTE, which is aligned to the current 3GPP
specifications and requires no changes at the user end. The
proposed design removes the control channel bottleneck and
maximizes the network capacity. Unlike the current 3GPP
mechanisms of enhancing the capacity, the proposed scheme
does not use additional resources from the data channel. We
efficiently use the large antenna structure available at the base
station and schedule more users in the PDCCH with FD-
MIMO. For this, we rely on the sounding reference signals
transmitted in the uplink to decide the best beam for a user.
We then ingeniously schedule the users and enhance the
PDCCH capacity. We abstracted the channel estimation errors
and used them in the system-level simulations to evaluate the
performance of the proposed scheme. We have shown that the
proposed design always performs better than the state of art
algorithms and the existing 3GPP schemes. In future, we will
validate the performance of the proposed BF-PDCCH design
by implementing it on the hardware test-beds.
R EFERENCES
[1] Y. Nam et al., “Full-dimension MIMO (FD-MIMO) for next generation
cellular technology,” IEEE Commun. Mag., vol. 51, no. 6, pp. 172–179,
June 2013.
[2] J. Lee et al., “LTE-advanced in 3GPP Rel -13/14: an evolution toward
5G,” IEEE Commun. Mag., vol. 54, no. 3, pp. 36–42, March 2016.
[3] Physical Channels and Modulation, 3GPP TS 36.211, Jul. 2016.
[4] Multiplexing and Channel Coding, 3GPP TS 36.212, Jun. 2016.
 [5] Physical Layer Procedures, 3GPP TS 36.213, Jun. 2016.
[6] “Samsung 5G-Ready LTE Massive MIMO Solutions Advance Sprint’s
Network,” Los Angeles, Samsung, 2018. [Online]. Available: https:
//news.samsung.com/us/5g-ready-mimo-sprints-network
[7] “LTE Evolution to Build All Business Foundation,” Shen-
zen, China, Huawei, White Paper, 2019. [Online]. Avail-
able: http://www-file.huawei.com/-/media/CORPORATE/PDF/white%
20paper/2019/lte evolution to build all business foundation.pdf
[8] Balamurali, “Optimal downlink control channel resource allocation
for LTE systems,” in Int. Conf. Signal Process. Commun. (SPCOM),
Bangalore, India, July 2010, pp. 1–5.
[9] S. Kucera and D. Lopez-Perez, “Expanding Coverage Range and Control
Channel Capacity of Co-Channel LTE Small Cells by Using PDCCH
Orthogonalization,” in Proc. IEEE 80th Veh. Technol. Conf. (VTC),
Vancouver, BC, 2014, pp. 1–5.
[10] D. G. Gonzalez et al., “Power allocation for the PDCCH in LTE: A
way to increase its capacity in realistic deployments,” in Proc. WPMC,
Atlantic City, 2013, pp. 1–5.
[11] S. Hirayama et al., “Performance of downlink control information
signals using decision-feedback channel estimation for EPDCCH,” in
IEEE Int. Conf. Commun. Syst., Macau, 2014, pp. 625–629.
[12] F. Nabil and Y. A. Fahmy, “Scattered search space scheme for PDCCH
and ePDCCH in LTE-advanced,” in 7th Int. Congress Ultra Modern
Telecomm. Control Syst. Workshops (ICUMT), Brno, 2015, pp. 172–177.
[13] S. Ye et al., “Enhanced Physical Downlink Control Channel in LTE
Advanced Release 11,” IEEE Commun. Mag., vol. 51 no. 2, pp. 82–89,
February 2013.
[14] A. K. Talukdar, “Performance Evaluation of the Enhanced Physical
Downlink Control Channel in a LTE Network,” in IEEE PIMRC,
London, 2013, pp. 987–991.
[15] L. Liu et al., “Search space design of enhanced physical downlink
control channel for long term evolution advanced system,” in IEEE
WCNC, Shanghai, 2013, pp. 3323–3328.
[16] Radio Resource Control (RRC), 3GPP TS 36.331, Jan. 2016.
[17] M. P. Reddy et al., “A Novel Beamformed Control Channel Design for
LTE with Full Dimension-MIMO,” arXiv:1905.05520 [eess.SP], 2019.
[Online]. Available: https://arxiv.org/abs/1905.05520
[18] WF on SRS estimation error modeling, 3GPP TR R1-150867, Feb. 2015.
[19] Study on 3D channel model for LTE, 3GPP TR 36.331, Dec. 2017.
[20] Clarification on the Wrap-Around Hexagon Network Structure, IEEE
802.20 C802.20-05/15, Mar. 2005.