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

net/publication/282909463

Timing over a 4G (LTE) mobile network

Conference Paper · January 2014

DOI: 10.13140/RG.2.1.1339.2081

CITATIONS READS
5 2,842

7 authors, including:

Rimantas Miškinis Domas Jokubauskis

Center for Physical Sciences and Technology Center for Physical Sciences and Technology
9 PUBLICATIONS   23 CITATIONS    6 PUBLICATIONS   10 CITATIONS   

SEE PROFILE SEE PROFILE

Emilis Urba
Semiconductor Physics Institute
18 PUBLICATIONS   43 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Domas Jokubauskis on 17 October 2015.

The user has requested enhancement of the downloaded file.

 Timing over a 4G (LTE) mobile network

Rimantas Miškinis, Domas Jokubauskis, Dmitrij Bronislovas Dzindzelėta, Kęstutis Svirskas

Smirnov, Emilis Urba, Bogdan Malyško Department of Radiophysics, Vilnius University
Time and Frequency Standard Laboratory Vilnius, Lithuania
Center for Physical Sciences and Technology
Vilnius, Lithuania
Email: rimantas.miskinis@ftmc.lt

Abstract— Various characteristics of time transfer over the

Fourth Generation (4G), or Long Term Evolution (LTE), mobile
telecommunication networks are investigated. The investigation
aims at properties of the time transfer over the LTE networks
using commercially available equipment. Two operating modes
of the network have been detected, and the operating features,
characteristic of the two modes, are reported. Potentially
attainable accuracy of the time transfer over the LTE networks
is discussed.

I. INTRODUCTION
Emerging technologies of mobile communications allow
drastic increase of the data transfer rate over a mobile
network. Features of mobile device synchronization and time
transfer first of all depends on the data transfer characteristics
of mobile network. In our previous works [1, 2], various
characteristics of time transfer over mobile
telecommunication networks were investigated. Using the
NTP (Network Time Protocol) technology, it was
demonstrated that the most important role in computer (PC)
clock synchronization over a 3G network is played by
network asymmetry, not the absolute delay time. In the
present work, we focus on the investigation of timing and
synchronization of a PC clock over a 4G (LTE) mobile
network.
II. MEASURING TECHNIQUE
Figure 1. Device setup for investigation of time transfer via the
LTE mobile network
the UTC(LT) clock using NTP (daemon v.4.2.6p5) technology
via the LTE mobile networks. The MTS created features low
power consumption (< 5 W), is small and light, and can be
adapted for field applications.
The idea behind the experiments is the following: MTS
periodically sends requests via the mobile internet connection
provided by a mobile cellular communications operator to an
1E-6
Overlapping Allan Deviation

Here we describe our experiments with a specially 1E-7

designed Mobile Testing System (MTS) consisting of
appropriate hardware and running necessary software.
Experimental setup, similar to that used in our previous work 1E-8

[1, 2], is revealed in Fig. 1.

The MTS has been constructed using a single-board 1E-9

computer “Raspberry Pi” Model B (RPi) and “Huawei E398” 1 10 100 1000
Averaging Time, s
10000 100000

modem connected to the RPi via an USB interface. The

“Raspbian Wheezy” operating system, with Linux (kernel
v.3.6) and LinuxPPS kernel module (PPS API (RFC 2783)
implementation) is running on RPi. Second pulses (PPS)
generated by UTC(LT) traceable clock are time-stamped
according to the internal RPi clock, which is synchronized to
Figure 2. Stability characteristics of the MTS system clock, while
synchronization takes place every 8 seconds by the NTP protocol in
the local network

978-1-4799-5252-6/14/$31.00 ©2014 IEEE 491

NTP server of the Time and Frequency Standard Laboratory
(TFSL). The server, which is synchronized to the time scale 0,005
UTC(LT), the Lithuanian representation of the Universal
Coordinated Time realized by TFSL, sends its NTP responses Continuous Uplink Mode
in the same way as they come. MTS, while receiving the 0,000
responses, processes the data obtained, taking into account the

Offset from UTC(LT), s

time marks generated both by the MTS and the NTP server.
Provided that travelling forward and backwards takes the same -0,005

time, such a technique allows obtaining both internet delay

(travelling time) and exact difference (offset) between the time
of MTS clock and the NTP server. Finally, using a common -0,010

algorithm, MTS synchronizes its clock according to the results Switching Uplink Mode 3a
obtained in order to set the offset to zero. At last, on the other
-0,015
hand, MTS receives a pulse per second (PPS) signal 0 20000 40000 60000 80000

representing UTC(LT) directly and records the time indicated Measurement time, s

by the MTS’ clock when the pulses come.

Thus, keeping in mind that both PPS coming directly and
the NTP server’s responses represent the same time scale – 0,006
UTC(LT), deviation of the recorded time values from exact Continuous Uplink Mode
integer values and its scattering reflects the accuracy at which 0,004

a mobile device is synchronized to a time scale generated by a

NTPd time offset,s

0,002
remote source and disseminated by the NTP server.
0,000

To estimate the possibilities for time synchronization in a

-0,002
mobile communication network, we use non-standardized
measuring equipment. Therefore, before making any -0,004

measurements, it is necessary to explore its metrological -0,006

characteristics. The calibration setup and procedure employed
is the same as in [1].
To this end, we have performed the experiment with the
same MTS, by connecting it to internet via the wires. The
deviations and their scattering in this case are much less than
those obtained using a mobile connection and fall in the sub-
millisecond range. This implies that the results obtained with a
mobile connection can be reasonably attributed to the mobile
internet connection, not to the properties of our MTS.
From the measurements carried out, one can conclude that
MTS system clock has a constant error of -47 µs, while the
standard deviation is 39 µs. Stability characteristics of the
MTS system clock are shown in Fig. 2.
III. RESULTS
Here we review only the results obtained while the mobile
terminal operates with the base stations belonging to the same
Radio Network Controller (RNC) and the signal level of a
base station higher than -90 dBm. The time transfer is carried
out by connecting to the same base station. Our measurements
include two different configurations of the LTE network:
Continuous Uplink Mode and Switched Uplink Mode, or the
Discontinuous Reception (DRX) mode [3]. Operating in the
Continuous Uplink Mode implies the network configuration
such that continuous uplink and downlink connection between
the modem and the base station is maintained. This mode
requires large power consumption by the modem because it is
communicating with the network continuously. Generally, the
operators don’t use this mode. Here we just demonstrate the
possibilities of the LTE technology. Usually, the Switched
Uplink Mode, which implies that the uplink channel is
switched off if there is no data transmission for about 10 s, is
Switching Uplink Mode
3b
0,010 0,015 0,020 0,025 0,030
NTP Packet round-trip Delay, s
Figure 3. Timing characteristics of the MTS system clock
synchronized over the LTE network operating in different modes
used. When the uplink channel is switched on, large delays
(up to 0.5 s) occur in the network. However, about 1 s after the
uplink channel is switched on, the network delay returns to its
normal value. It is why we do synchronization every 8 s and
skip the first packet after the network wake-up.
As the results show (see Table 1), the time offset from
UTC(LT) in the 4G network, while operated in the Continuous
Uplink Mode (see Fig. 3a), is smaller than that in the Switched
Uplink Mode. The same applies to the delay of NTP packets
and related offset scattering. It is shown in Fig. 3b.
The results in Fig. 3b are obtained in the following way.
As described above, MTS is sending NTP requests via the
mobile internet connection, receiving responses, calculating
offset of its own time from the NTP time, and applying
corrections to its own time according to the algorithm. With
every subsequent NTP request/response, MTS calculates the
round-trip delay of the packet and offset of the NTP time from
the current MTS time. Every pair of the values obtained is
denoted by a point on Fig. 3b. The results imply that the NTP
system is operating as expected, and the change of the time
offset shown in Fig. 3a is caused rather by properties of the
cellular network system.

492
 300

Uplink
Downlink

200

Packet delay, ms
100

1 2 3

0
0,0 0,5 1,0 1,5 2,0
Figure 4. Device setup for asymmetry measurement in the LTE Packets session start time, s
network
Figure 6. Dynamics of the LTE link in the DRX mode after an idle
Table 1. The results of the MTS synchronization period of 20 seconds. There are 3 distinct phases: 1) link establishment
(uplink delays up to 300 ms), 2) fine operation (uplink delays up to 25
Mean of offset Standard ms), 3) normal operation (uplink delays up to 50 ms)
Configuration
from UTC(LT) deviation of offset
Calibration session start time. Then, MTS sends a response in the exactly
-0.047 ms 0.039 ms opposite way: the response, having passed via the Ethernet
Mode
Continuous connection and internet, the radio segment, and the modem,
1.85 ms 0.67 ms finally comes back to MTS. Once again, the times of packet
Uplink Mode
Switched departure and arrival are recorded.
6.40 ms 0.87 ms
Uplink Mode The network asymmetry, or the difference between the
traveling times in the two directions, is displayed in Fig. 5.
Inaccuracy of synchronization over a communication The results agree well with the offset considered before. We
network, which results in the offset of the MTS clock from suppose that the main contribution to the asymmetry is made
UTC(LT), is caused by the network asymmetry, i.e., different by the radio segment.
propagation delays when an NTP packet travels in the up-link Three distinct connection initiation phases are observed
and down-link directions. To explore the effect, we have made after a wake-up of the LTE link in the DRX mode (see Fig. 6).
an asymmetry measurement. MTS, which clock is The link establishment phase should be avoided for timing-
synchronized to UTC(LT) directly, sends a packet to the sensitive communication. It may be possible to selectively
modem; the packet travels via the radio segment and internet utilize Phase (2) for NTP synchronization to achieve the best
and returns back to MTS via an Ethernet connection (see Fig. time transfer accuracy.
4). The times of packet departure and arrival are recorded. We
define the time of packet departure from the modem as the IV. CONCLUSIONS
60
1 2 3 2 While synchronizing a computer clock over a mobile
telecommunication network, its operation in the Continuous
Base Station Uplink
50 Cell ID: Downlink Uplink Mode yields far better results than that in the Switched
1. 2C1071F Asymmetry Uplink Mode. The measurements reported can be used as a
2. 2C7840B
40 3. 2C7100B means for low-level monitoring of the mobile communication
network. The accuracy of the synchronization over a LTE
Packet delay, ms

network demonstrated is better than 10 ms.

30

V. REFERENCES
20
[1] R. Miškinis, D. Smirnov, E. Urba, B. Dzindzelėta, “Timing and
synchronization in mobile telecommunication networks,” Frequency
10 Control and the European Frequency and Time Forum (FCS), 2011
Joint Conference of the IEEE International, pp. 665-669, San
Francisco, CA, 2-5 May 2011.
0
0 600 1200
[2] R. Miškinis, D. Smirnov E. Urba, B. Dzindzelėta. Improving Timing
Capabilities in 3G Mobile Networks // European Frequency and Time
Packets session start time, s Forum 2012. – P. 368 – 370.
[3] ETSI TS 136 321 V9.4.0 (2011-11), LTE; Evolved Universal
Figure 5. Uplink and Downlink packet traveling time in LTE network Terrestrial Radio Access (E-UTRA); Medium Access Control (MAC)
while modem switches between base stations protocol specification, (3GPP TS 36.321 version 9.4.0 Release 9)

493

View publication stats