Anda di halaman 1dari 6

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/319164296

Performance Analysis of the Random Access


Channel in NB-IoT

Conference Paper · September 2017

CITATIONS READS

0 213

3 authors, including:

Ruki Harwahyu Ray Cheng


University of Indonesia National Taiwan University of Science and Tec…
26 PUBLICATIONS 38 CITATIONS 95 PUBLICATIONS 967 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ruki Harwahyu on 18 August 2017.

The user has requested enhancement of the downloaded file.


This is draft version for personal use. The original manuscript will be published in the proceeding of IEEE VTC Fall 2017.

Performance Analysis of the Random Access Channel


in NB-IoT
Ruki Harwahyu, Ray-Guang Cheng Chia-Hung Wei
National Taiwan University of Science and Technology Foxconn Advanced Communication Academy (FACA)
Taipei, Taiwan Taipei, Taiwan
crg@mail.ntust.edu.tw d9902208@mail.ntust.edu.tw

Abstract— Narrowband Internet of Things (NB-IoT) is a new preamble consists of four symbol groups. Each symbol group
access technology introduced by 3GPP. It aims to support contains one cyclic prefix plus five symbols. The four symbol
massive machine-type communications services in wide area. groups are transmitted at four different sub-carriers following
This work presents an analytical model to estimate access success
probability and average access delay of random access (RA)
a pre-defined frequency hopping pattern. The same preamble
procedure for three coverage enhancement (CE) levels in NB- is repeatedly transmitted within the same NPRACH. A base
IoT. The behavior of RA procedure in NB-IoT is analyzed and station chooses a proper number of repetitions for different CE
accuracy of the analysis is verified by computer simulations. levels to ensure good signal quality at the receiver. The initial
sub-carrier selected for transmitting the first symbol group in
Index Terms—NB-IoT, multi-channel slotted ALOHA the first preamble is randomly chosen from a set of sub-
carriers dedicatedly allocated to the CE level. In our
I. INTRODUCTION subsequent discussion, preamble is said to be transmitted “in
Narrowband Internet-of things (NB-IoT) is a new 3GPP CE level c” if the first symbol group of the first repetition of
standard aiming to support a massive number of low-cost, low- the preamble is transmitted in a sub-carrier which is allocated
power, and low data rate devices operated in an enhanced to CE level c.
coverage area. NB-IoT adopts existing LTE-A OFDMA Different to LTE, there is only one preamble to be used by
technology but each carrier only requires 180 kHz bandwidth. all devices. In this system, collision occurs if two or more
An NB-IoT cell can be configured with one or more carriers in devices choose the same initial sub-carrier. Collided devices
which one of them is assigned as the anchor carrier [1]. NB-IoT shall perform random backoff and retransmit a new attempt in
supports three different operation modes. An operator can use a newly chosen initial sub-carrier in the next available
an NB-IoT carrier to replace a GSM carrier (200 kHz) in NPRACH. The process repeats until the maximum number of
„stand-alone‟ mode; to utilize a physical resource block (PRB) attempts in this CE level is reached. Different to LTE, the
of 180 kHz within a normal LTE carrier in „in-band‟ mode; or
failed device can re-start the whole RA procedure in the next
to occupy an unused resource block within the guard-band of a
higher CE level. The device will declare a RA failure if RA
LTE carrier in „guard band‟ mode.
procedures still fail in the highest CE level.
In NB-IoT, all devices receive synchronization signals and
The access mechanism in NPRACH in each CE level is
broadcasting messages from an anchor carrier. As in LTE, a
based on multi-channel slotted ALOHA system. In such a
device performs random access (RA) through narrowband
system, concurrent channel access from massive devices may
physical RA channel (NPRACH) [1] from the anchor carrier to congest the RA channel. The congestion becomes even worse
acquire uplink resource for data transmission. NPRACH is a in NB-IoT due to the limited resource resulted from narrow
set of consecutive sub-carriers in a time interval specially bandwidth and repetition. The access mechanism in each
reserved for preamble transmission. The sub-carrier spacing of NPRACH can be modelled by a multi-channel slotted ALOHA
NPRACH is 3.75 kHz. Up to forty-eight sub-carriers can be system. An analytical model was presented in [2] to
allocated to an NPRACH. These sub-carriers are exclusively investigate the behavior of multi-channel slotted ALOHA
shared by up to three coverage enhancement (CE) levels with system under congestion. A more comprehensive analytical
a basic sub-carrier allocation unit of 12 sub-carriers [1]. The model is also presented in [3] to deal with performance
three CE levels are defined based on the reference signal evaluation of RA in OFDMA-based network such as
received power (RSRP) measured by the device. Devices in LTE/LTE-A. This work extends the model proposed in [2] to
the lowest CE level have the highest RSRP while devices in evaluate the performance of RA procedure for three CE levels
the highest CE level have the lowest RSRP. in NB-IoT. During evaluation, we consider a one-shot access
A device starts an RA procedure by transmitting a with simplified system by applying common settings for all
preamble in an NPRACH at its initial CE level. The initial CE CE levels and disabling the random backoff. In IoT
level is determined based on the RSRP measured at the device. environment, one-shot access request can be generated by a
Different to LTE, there is only one preamble available. The service with synchronized data transmission. One of them can
This is draft version for personal use. The original manuscript will be published in the proceeding of IEEE VTC Fall 2017.

Fig. 1. Timing diagram of the random access channel in NB-IoT.

be implemented by activating a group of IoT devices with a access success probability for CE level c is the success
group paging message, similar to the one that has been probability for devices initially in CE level c. The average
envisioned since the uprising of machine-type communication access delay for CE level c is the average time required by a
service in LTE [6]. The usage of one-shot access in this paper device initially in CE level c to complete the RA procedure.
allows readers to understand the underlying concept and the Hence, the performance metrics will be derived based on
intrinsic behavior of RA procedure in NB-IoT easier. estimated average number of contending, successful, and
The rest of the paper is organized as follow. Section II collided devices in each NPRACH for all CE levels.
presents the system model for NB-IoT. The proposed analytical Let M cb [i] be the average number of contending devices
model and numerical results are given in Sections III and IV,
initially in CE level c which transmit preambles at the
respectively. Section VI summarizes the conclusion and future
beginning of the ith NPRACH in CE level b (0≤c≤2, c≤b≤2).
work.
Among these M cb [i] contending devices, Scb [i] and Ccb [i ] of
II. SYSTEM MODEL them are identified as successful and collided at the end of the
This work considers multiple stationary IoT devices NPRACH, respectively. From [2], the number of success
simultaneously start their RA procedure after receiving a devices, Scb [i] , can be approximated by
group paging message [2] in an NB-IoT system containing

 c 0 M cb [i ] 
b
three CE levels. Let Mc be the number of contending devices 
 
initially in CE level c (i.e., 0≤c≤2); Rc be the number of non-  Rb 
Scb [i]  M cb [i]e  
(1)
overlapping sub-carriers reserved by the base station for CE
level c (i.e., R0+R1+R2≤48); Tc be the NPRACH periodicity [5] The average number of collided devices, Ccb [i], is given by
in CE level c; NPTmax,c be the maximum number of
transmission attempts of a device in CE level c; and NPTmax,G Ccb [i]  M cb [i]  Scb [i]. (2)
be the maximum number of transmission attempts of a device Let I min (c, b) and I max (c, b) be the starting and ending
in all CE levels. For the ease of explanation, it is assumed that NPRACHs in CE level b for a device initially in CE level c to
Tc is long enough and the backoff window is set to 0 ms. Thus, transmit its attempts. The collided devices in NPRACH i may
a device can transmit preamble and receive acknowledgement start new attempts in NPRACH (i+1) in the same CE level
within Tc and a collided device will immediately retransmit a (i.e., CE level b) or in the starting NPRACH I min (c, b  1) in
new attempt in the next NPRACH. Notice that despite of our the next higher CE level (i.e., CE level b+1). In the former
simplification, this assumption is valid according to [5]. case,
The relationship among Rc, Tc, Mc, NPTmax,c and NPTmax,G in
M cb [i  1]  Ccb [i] for 1  I min (c, b)  i  I max (c, b)  1. (3)
NB-IoT with three CE levels are illustrated in Fig. 1. In this
example, each arrow represents a transmission attempt; R2 = In the latter case,
2R1 = 2R0, T1 = 2T0, T2 = 3T0, NPTmax,0 = NPTmax,1 = NPTmax,2 = 2, M cb 1[ I min (c, b  1)]  Ccb [I max (c, b)] for 0  c  b  1, (4)
and NPTmax,G = 5. Hence, devices initially in CE level 0 conduct where the attempt in NPRACH i in CE level b reaches the
2 (i.e., NPTmax,0) attempts in CE level 0, 2 (i.e., NPTmax,1) constraints of I max (c, b) and the new attempts in CE level
attempts in CE level 1, and 1 (i.e., NPTmax,G-NPTmax,0-NPTmax,1) (b+1) start from NPRACH I min (c, b  1) . The valid range of i
attempt in CE level 2.
depend on b and c. Take the case of c=0 as an example, the
III. ANALYTICAL MODEL devices transmit from NPRACH 1 to NPRACH min(NPTmax,0,
This paper presents an analytical model to estimate the NPTmax,G) in CE level 0 (i.e., b=0). Thus, M 00 [i  1]  C00 [i] for
performance metrics of access success probability and average i  (1, min(NPTmax,0, NPTmax,G)-1) and M 0 [i  1]  0 for i ≥
0

access delay [6] for devices in each initial CE level. The


min(NPTmax,0, NPTmax,G). The devices failed in CE level 0 can
This is draft version for personal use. The original manuscript will be published in the proceeding of IEEE VTC Fall 2017.

still transmit new attempts in CE level 1 if NPTmax,G > NPTmax,0. 2 I max ( c , b )

In this case, the last attempt in CE level 0 is transmitted at its  


b  c i  I min ( c , b )
iTb Scb [i]
NPTmax,0th NPRACH, and the first attempt in CE level 1 is Dc  .
  
2 I max ( c , b )

transmitted at its
T0
T1
NPTmax,0   1 th NPRACH. That is,  
b  c i  I min ( c , b )
b
S [i]
c
(6)

M 01   T N PTmax,0   1  C00  N PTmax,0  . The value of I min (c, b) and


T0

1 IV. NUMERICAL RESULT


I max (c, b) for different c and b are summarized in Table I. Computer simulations were conducted on top of a C
platform to verify the accuracy of the proposed analytical
TABLE I. VALUES OF I min (c, b) AND I max (c, b ) model. In the simulation, NPRACH Repetition for CE levels 0,
1 and 2 are set to be 8, 32 and 128, respectively [7]. The
c b I min (c, b) I max (c, b) analytical model is proven to be accurate by testing various
0 0 1 min( NPTmax,0 , NPTmax,G ) settings. For the ease of demonstration, we only demonstrate
0 1  TT0 NPTmax,0   1  TT0 N PTmax,0  the case with a common configuration for all three CE levels.
1  1 
 min( N PTmax,1 , N PTmax,G  N PTmax,0 )
That is, Mc = M, Tc = 1280ms, Rc = 12 sub-carriers, NPTmax,c =
5, and NPTmax,G =15, for 0≤c≤2, where M is an integer ranging
0 2

 T1  T0 N   N PTmax,1 
 T2  T1 PTmax,0    
 T1  T0 N   N PTmax,1 
 T2  T1 PTmax,0    from 10 to 100. In the following figures, each point
1  N PTmax,G  N PTmax,0  N PTmax,1 represented an average value of 107 repeated simulations. In
1 1 1 the following figures, lines and symbols are used to denote
min( NPTmax,1 , NPTmax,G )
analytical and simulation results, respectively.
1 2  T1
NPTmax,1   1  T1
NPTmax,1   NPTmax,G  N PTmax,1
 T2  T2
Figures 2, 3 and 4 show M cb [i] , Scb [i] and Ccb [i ] for M=40
2 2 1 NPTmax,G
and 0≤c≤2. It can be found that the conditions of the three CE
levels are the same during the first five NPRACHs, i.e.
M cb [i], Scb [i] and Ccb [i ] can be iteratively derived from the M 0b [i]=M1b [i]=M 2b [i] , S0b [i]=S1b [i ]=S2b [i ] and
first NPRACH in the three CE levels using (1)-(4). Without
C0b [i]=C1b [i]=C2b [i ]
for 1≤i≤5. It is because the number of
loss of generality, it is assumed that the starting time of the
first NPRACH in the three CE levels are synchronized initial contending devices, Mc, and the number of sub-carriers,
immediately after receiving a group paging message. Hence, Rc, are identical for all three CE levels. In the 6th NPRACH,
M cc [1]=M c and M cb [1]=0 for b≠c. The boundary condition the collided devices initially in CE levels 0 and 1 (from the 5th
NPRACH) jump to CE levels 1 and 2, respectively, to transmit
is M cb [i]=Scb [i]=Ccb [i]=0 for b<c since retransmissions can their attempts. In this case, M 01 [6]=C00 [5] and M12 [6]=C11[5] .
only be conducted in device‟s initial CE level or the higher CE In contrast, the failed devices initially in CE level 2 can only
level. The relationship among M cb [i] and Ccb [i ] is illustrated stay in CE level 2 to retransmit their attempts, i.e.
in Fig. 1. In this example, M0, M1, and M2 devices start their M 22 [6]=C22 [5] . During the 6th to 10th NPRACHs, the average
first attempt in the first NPRACH of the three CE level, number of devices initially in CE level 0 and contend in CE
respectively. The collided devices of C00 [1], C11[1] and C22 [1] 1
level 1 is gradually decreased because S0 [i] gradually
will become M 00 [2], M11[2] and M 22 [2] , respectively. The increases. In contrast, the average numbers of successful
collided devices which has transmitted NPTmax,c times in their devices initially in CE levels 1 and 2 drop almost to zero
initial CE level will jump to a higher CE level and start a new because the number of contending devices in the 6th
attempt in the next available NPRACH, such that NPRACH are doubled (compared to those which contend in
CE level 1). The devices initially in CE level 0 and failed in
M 01 [3]=C00 [4] and M12 [3]=C11[3] .
CE level 1 will move to CE level 2 to transmit new attempts in
Access success probability of CE level c, Pc, is the ratio the 11th NPRACH. The newly joined contending devices
between average number of successful devices initially in CE further congest the channels in CE level 2 and thus, no device
level c and total number of devices initially in CE level c. That
is, is successful (i.e., S02 [i]=S12 [i]=S22 [i]=0 11≤i≤15). As a result,
2 I max ( c , b )
M c2 [i] for 0≤c≤2 are almost unchanged.
 
b  c i  I min ( c , b )
Scb [i]
Pc  . (5)
Mc
Average access delay for CE level c, Dc, is the average time
required by a successful device initially in CE level c to
complete its RA procedure. That is,
This is draft version for personal use. The original manuscript will be published in the proceeding of IEEE VTC Fall 2017.

M=40 sim. c=0


ana. c=0
1
sim. c=0
sim. c=1
number of transmitting devices, Mb [i]

ana. c=0
c

ana. c=1

c
access success probability, P
0.8 sim. c=1
40 sim. c=2
ana. c=1
ana. c=2
30 sim. c=2
0.6 ana. c=2
20

10 0.4

0
1 2 0.2
5
10 1
CE level where the 0
15 0 attempt is conducted, b 10 20 30 40 50 60 70 80 90 100
NPRACH index, i number of UEs per CE level, M
Fig. 2. Number of contending devices when M=40. Fig. 5. Access success probability.

M=40 sim. c=0 Figure 6 shows the average access delay, Dc, for the three
ana. c=0
CE levels with different number of M. It is found that the
sim. c=1
average access delay for devices initially in CE levels 1 and 2
number of successful devices,Sb [i]

ana. c=1
c

5 sim. c=2 are quite similar since they have almost the same conditions.
ana. c=2 The average access delay for devices initially in CE levels 1
4
and 2 is higher than that in CE level 0 when M is less than 30
3
because the devices in CE level 0 success with fewer attempts.
2 The average access delay for devices initially in CE levels 1
1 and 2 reaches its maximum at M = 34 and then greatly
0
decreases because their success probabilities quickly decrease
1 2 when M increases. It can be found that the successful devices
5 are those initially from lower CE level since the number of
1
10 contending devices in higher CE level increases rapidly.
CE level where the
15 0
NPRACH index, i attempt is conducted, b 12
sim. c=0
Fig. 3. Number of successful devices when M=40.
ana. c=0
average access delay, D (sec.)

10 sim. c=1
M=40 sim. c=0 ana. c=1
c

ana. c=0 sim. c=2


sim. c=1 8 ana. c=2
number of collided devices, Cb [i]

ana. c=1
c

40 sim. c=2
ana. c=2 6
30

20 4

10
2
0 10 20 30 40 50 60 70 80 90 100
1 2 number of UEs per CE level, M
5 Fig. 6. Average access delay.
1
10
CE level where the
15 0
NPRACH index, i attempt is conducted, b V. CONCLUSION
Fig. 4. Number of collided devices when M=40. NB-IoT is designed to support massive machine devices in
wide area. The RA procedure in NB-IoT introduces a concept
Figure 5 shows the access success probability, Pc, for the of CE level to support devices with different signal qualities.
three CE levels based on (5). It is found in Fig. 5 that devices The configuration in each CE level can be independently
initially in CE level 0 have a higher access success probability designed. The flexible configuration allows operator to specify
than that in CE levels 1 and 2. The access success probabilities customized settings for each base station in order to meet
for devices initially in CE level 1 and 2 are similar because the requirements set by environment (e.g., target coverage area),
average number of success devices from CE levels 1 and 2 are services (e.g., traffic characteristics, power consumption), and
the same. network performance metrics (e.g., access success probability,
This is draft version for personal use. The original manuscript will be published in the proceeding of IEEE VTC Fall 2017.

average access delay), etc. However, interaction among CE REFERENCES


levels makes the configuration of the RA procedure much [1] 3GPP TS 36.321, “Medium access control (MAC) protocol
more complicated and challenging than expected. Hence, a specification,” V13.2.0, Jun. 2016.
joint optimization is expected to enhance the performance. [2] C. H. Wei, R. G. Cheng, S. L. Tsao, “Modeling and estimation of
This paper presents an analytical model to estimate the one-shot random access for finite-user multichannel slotted
access success probability and average access delay during RA ALOHA systems,” IEEE Commun. Lett., vol. 16, pp. 1196-1199,
procedure in NB-IoT for devices from three initial CE levels. Aug. 2012.
The accuracy of the analytical model is verified by computer [3] C. H. Wei, G. Bianchi, R. G. Cheng, “Modeling and analysis of
random access channels with bursty arrivals in OFDMA wireless
simulations. With the proposed model, the contentions among
networks,” IEEE Trans. Wir. Commun., vol. 14, no. 4, pp. 1940-
devices initially in different CE levels at each NPRACH in 1953, April 2015.
each CE level can be easily observed. The proposed model can [4] C. H. Wei, R. G. Cheng, S. L. Tsao, “performance analysis of
be used to optimize the setting such as number of supported group paging for machine-type communications in LTE
CE levels and number of sub-carriers, NPRACH periodicity, networks,” IEEE Trans. Veh. Technol., vol. 62, no. 7, pp. 3371-
and the maximum number of transmissions for each CE level. 3382, Sept. 2013.
[5] 3GPP TS 36.331, “Radio resource control (RRC) protocol
ACKNOWLEDGMENT specification,” V14.1.0, Dec. 2016.
This work was supported in part by the National Science [6] 3GPP TR 37.868, “Study on RAN improvements for machine-
type communications,” V11.0.0, Sep. 2011.
Council, Taiwan, under contracts MOST 102-2219-E-011- 001
[7] X. Lin, A. Adhikary, Y. P. Eric Wang, “Random access preamble
and 102-2221-E-011-003-MY3. design and detection for 3GPP narrowband IoT systems,” IEEE
Commun. Lett., vol. 5, no. 6, pp. 640–643, Dec. 2016.

View publication stats