Professional Documents
Culture Documents
net/publication/301403003
A case study on protocol stack integration for 3GPP LTE evolved node B
CITATIONS READS
0 2,149
4 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Felipe Augusto Pereira de Figueiredo on 14 July 2016.
Fabbryccio A. C. M. Cardoso, Felipe A. P. Figueiredo, Rafael Vilela and João Paulo Miranda
Centre for Research and Development in Telecommunications (CPqD), Campinas/SP, Brazil
{fcardoso, felipep, rvilela, jmiranda}@cpqd.com.br
Abstract—This paper presents a case study of the integration based on the micro telecommunications computing archi-
of the Long Term Evolution (LTE) physical layer into the radio tecture (µTCA). The physical interface between devices is
protocol subsystem of a macrocell base station. We conduct tests raw gigabit ethernet (GbE), used for validation purposes.
in the laboratory using a modular experimental setup aimed The proposed validation tests focus on the physical layer.
at validating this protocol stack integration. Measurements and Basically, the tests consist of capturing in real time the
analysis are carried out with the aid of Wireshark, a well-known
exchange of message sequences required to establish a radio
protocol analyzer tool. This choice allows the inspection of data
packets at the application program interface between PHY and resource control (RRC) connection with a user equipment (UE)
MAC subsystems. Our results suggest that the LTE protocol stack terminal emulator [4]. Besides exercising all physical channels,
can operate synchronously, thus supporting message sequences this approach exploits layer 2-3 interactions regarding transport
conveyed within both random access and radio-resource-control and logical channels, as well as transparent and acknowledged
connection procedures. The proposed experimental setup and service modes. It is worth emphasizing that the objective of our
test methodology make it possible to validate the control plane experiment is not to verify conformance to the LTE standard.
operation of the protocol stack without the need of a core network Instead, the experiment was designed to support debugging of
interface. the physical layer and its corresponding application program
I. I NTRODUCTION interface (API), driven by standard protocol procedures.
The Long Term Evolution (LTE) technology was standard- The remainder of the paper is as follows. Sections II and
ized in 2008 as the successor of third generation (3G) cellular III offer some background on the baseband unit in terms of the
systems. For the first time, we have technology convergence LTE protocol stack, layer interaction, and connection establish-
in one generation of cellular communication systems. This has ment. Our base station architecture is then presented in Section
already been affecting market scale in a number of aspects, e.g. IV. We carry on with Sections V, VI and VII describing our
the migration to LTE is taking place at a pace considerably test methodology, setup, and results, respectively. Limitations
faster than that from earlier generation networks to 3G [1]. of our approach are given in Section VIII, alongside with ways
Higher data rate and growth saturation in existing networks to overcome them. Section IX closes the paper.
drive the demand for LTE infrastructure, which is currently
being installed so as to upgrade the entire ecosystem. This II. T HE LTE BASEBAND U NIT
includes base stations, backhaul, and the core network. At the The physical layer supports single and multiple transmit
base station, in particular, the radio protocol stack is typically antennas, using transmission modes TM1-TM4, with up to
specialized in physical layer and protocol subsystems, each 20 MHz bandwidth, frequency division duplexing and normal
requiring dedicated devices. Examples of such devices include, cyclic prefix [11]. For simplicity, packet data convergence pro-
but are not limited to, multicore general purpose processor tocol (PDCP) functions [5], e.g. header compression, cypher
(GPP), for the protocol subsystem, while application specific and integrity protections, are turned off at the layer 2.
integrated circuit (ASIC), digital signal processor (DSP) and The UE establishes a connection to the base station,
field programmable gate array (FPGA) are some of the options referred as evolved node B (eNodeB) in the context of LTE,
available when it comes to the physical layer subsystem. using procedures of random access [6] and RRC connection
However, even when highly integrated system on a chip establishment [7]. The former is used to achieve time syn-
(SoC) devices are used, system designers face the task of op- chronization in the uplink and, if it is the first access, to
timizing the interface between subsystems. The synchronized acquire a 16-bit layer 2 address known as cell radio net-
operation and tight timing constraints of LTE require hardware work temporary identity (RNTI). Dedicated control messages
and software optimizations, followed by a careful validation are allowed after the UE has established a RRC connection
of the radio protocol stack in terms of real time operation. through the signaling radio bearer (SRB) type SRB1, including
This paper deals with the latter subject in that it presents initial non-access stratum (NAS) protocol transmission [8],
a case study of a complete radio protocol integration for which triggers security activation and network attachment.
modular LTE baseband units [2]. Taking into consideration System information acquisition, random access and RRC
the high capacity often envisioned for macro cellular base connection establishment procedures involve practically all
stations, the physical layer subsystem is implemented “in- interactions between layers in the protocol stack. These pro-
house”, at CPqD, using a Xilinx Virtex 6 FPGA device. In cedures can exercise physical, transport and logical channels,
contrast, the protocol subsystem is a commercial software as well as NAS and signaling radio bearers SRB0 and SRB1.
component from Aricent [3] that is optimized for use with We therefore deem these procedures as appropriate to validate
Intel core i7 technology. Each device runs on a separated the interactions between the physical layer subsystem (signal
radio advanced mezzanine card (AMC) plugged into a chassis processing) and the protocol subsystem.
NAS (Non Acces Stratum) IP (Internet Protocol) S1
X2
L2/L3
running
on
GPP
Layer 3
User Traffic
FAPI
Baseband
RRC (Radio Resource Control)
RRC PDUs L1
running
on
FPGA
PDCP Control (Control Traffic)
…
PDCP (Packet Data Convervence L1
Downlink
L1
Uplink
Protocol) TX
RX
Radio
RLC Control SRB1, SRB2, DRB Bearers
… Port
0
Port
1
Port
Layer 2
Antennas
MAC Control CCCH, DCCH, DTCH, BCCH, PCCH
Channels
MAC (Medium Access Control)
Transport
LTE-‐Uu
L1 Config DL-SCH, PCH, BCH, UL-SCH, RACH
Channels
Layer 1
Figure 1: Layer interactions of LTE radio protocol stack. Figure 2: Base station functional architecture.
RA Procedure
(FMC) connector between RF card and baseband unit. RRCConnectionRequest/ SRB0/ CCCH/ UL-SCH/ PUSCH
The logical interface between the protocol and the
ACK/ HI/ PHICH
physical layer subsystems is based on FAPI over raw
GbE [12]. This approach provides a baseline designed CR/ DL-SCH/ PDSCH (TC-RNTI)
to evaluate the correct operation of the baseband unit, ACK/ UCI/ PUCCH
integrating protocol and physical layer subsystems.
The experiments described later on in this paper focus on RRCConnectionSetup/ SRB0/ CCCH/ DL-SCH/ PDSCH
RRC Connection
Establishment
SR/ UCI/ PUCCH
protocol interface. The protocol subsystem was previously
tested using a MAC-to-MAC communication approach, while DCI0/ PDCCH
the physical layer was validated in terms of the physical RRCConnectionSetupComplete/ SRB1/ DCCH/ UL-SCH/
channel using a Signalium UE Emulator (SORBAS) and an PUSCH
Agilent Infinium running VSA89600 [4]. Detailed information ACK/ HI/ PHICH
Protocol Analyzer
(Wireshark)
Fig. 4. Experimental Setup.
Figure 4: Block diagram of the experimental setup. L2/L3 L1
The next protocol procedures after connection establishment L2/L3 L1
would be initial analysis
The protocol security considers
activation and
the radio bearerof
sequence UE
establishment. However, these require integration to the core UE
messages exchanged, along with their timestamps. In
network through the S1 interface.
fact, the occurrence of a subsequent message in the eNodeB
call The
flowvalidation
is an of the eNodeB
important radio protocol
indication that stack
the isprevious
based
Fig. 5. The setupeNodeB
used in the experiment.
on laboratory
message tests, which
has correctly beenwerereceived
designed and with decoded.
the aim of
analyzing
The ratemessage exchanges
of messages between atUEtheandFAPI
exchanged eNodeB.level composedFigure
from left
5: toActual
right by an Intel
setup usedcorein i7
theboard for L2/L3,
experiment.
Message sequences,
generates a myriad asofshown in Fig.for
packets 3, are filteredindication
timing and analyzedand an FPGA AMC board with FMC radio for the physical layer,
to check if of
configuration the physical
initial connection proceduresMessages,
layer resources. are correctly
such and a carrier hub for switching capabilities. Detailed information
asimplemented
SUB FRAME in accordance
INDICATION, with LTE. DLThe experimental
CONFIG setup
REQUEST is provided in Table
VII.I. Wireshark runs in a notebook
E XPERIMENTAL R ESULTSwith Fedora
described in Fig. 4 involves protocol analysis software, uTCA OS, which is not messages
Exemplary shown in Fig. 5 but and
captured can filtered
capture using
packetWireshark
data
and UL CONFIG REQUEST, are exchanged every 1
hardware platform and test equipment, as described in Table I. through a switch
are shown connected
in Figure to athis
6. For carrier hub. example,
specific The eNodeBTable IV
ms. This is a clear indication that it is not possible to
TX/RX is cable
adds timing connected
information to the UE
concerning RF unit
subframe through
number (SF) and
analyzeThe message
protocol analyzer
sequencesWireshark
withoutalready comes with
filtering built-
capabilities attenuators, and operates
system frame numberwith singleThis
(SFN). antenna mode in downlink.
information was collected
in inasupport
call for
flowLTEwith,
protocols.
for However,
instance,as 90 the FAPI
ms protocol
duration.is Messages between L1 and L2/L3unit
wereincrement
opened andinanalyzed at
not supported,isa third-party
Wireshark an excellent FAPItool
dissector needs to be
to capture andinstalled.
analyze at the FAPI level. Every SF is equiva-
FAPI, MAC and RRC levels. The eNodeB could be also
In thisflows
packet case,through
Aricent easily
dissector was installed
configured filters.toWeguarantee
employ lent to 1 ms, while in SFN it is equivalent to 10 ms.
connected to an EPC through the GbE interface of the L2/L3
theinteroperability
filters describedof theinL1Table
developed
II andin this
III project with Aricent
to validate the call Note the second random access response (RAR) in packet
board. This interface, however, is not used in the experiment.
L2/L3.
flow Recall2. that
in Figure FAPI is
Wireshark alsonotenables
a standard, but only a
color representation number 50288. This is the result of an erroneous preamble
recommendation from the Small Cell Forum. The
according to the identified message, or to the values assumed complete list detection. This VII. EXPERIMENTAL
false RESULTS because it generates
alarm is inconvenient
byofspecified
protocols required for the analysis is: SLL (TCPDump),
message fields. The set of rules used to color anExemplary
overhead messages
of radio captured
resourceand usage andusing
filtered eNodeB processing
Wireshark
Ethernet, Ipv4, UDP, FAPI, MAC-LTE, RLC-LTE, PDCP-
packets according to the filters are based on the same rules
LTE, LTE-RRC. aretime.
shownHowever,
in Fig. 6.it For
is not
thiscritical
specificinexample,
terms ofTable
system function-
IV adds
used in the filters. ality, since there is no UE access attempt
timing information concerning subframe number (SN) and associated to that
The protocol analysis considers the sequence of messages RAR.frame
system The number
correct (SFN).
RAR Thiscaused the UE was
information to respond
collectedwith
at an
exchanged, alongVI.withEtheir timestamps. SIn
XPERIMENTAL fact, the occurrence
ETUP theRRCConnectionRequest,
FAPI level. Every unit increment which isin correctly detected
SN is equivalent to by
1 the
of a subsequent message in the call flow is an important ms, while ininSFN
eNodeB it is 50326
packet equivalent
(5 msto 10 ms.the corresponding RAR).
after
The setup
indication thatused to run the
the previous experiment
message follows
has correctly thereceived
been scheme Receipt confirmation with ACK is sent 4 ms after the receipt
depicted in
and decoded.Figure 4. The eNodeB is modular, as shown in ofNote the second random access response (RAR) in packet
the RRCConnectionRequest in packet 50328. As the second
Figure 5, composed from left to right by an Intel core i7 board number 50288. This is the result of an erroneous preamble
The rateanofFPGA
messages exchanged RAR is in error, the eNodeB does not receive any connection
for L2/L3, AMC board atwith
the FAPI
FMClevel radiogenerates
for the detection. This false alarm is inconvenient because it generates
a myriad
physical of packets
layer, and a for timing hub
network indication and configuration
for switching of
capabilities. an overhead of radio resource usage andframe,
request within the expected time eNodeBso processing
it sends back
NACK as response in packet 50336. This ends the message
Detailed information is provided in Table I. Wireshark runsasin
physical layer resources. Messages, such time. However, it is not critical in terms of system functionality,
S UB_FRAME_INDICATION,
a notebook with Fedora OS, which DL_CON is notFIG_RE
shown Q UinE SFigure
T and 5 sequence
since there isstarted by theattempt
no UE access false associated
preamble todetection.
that RAR.Note
The that
UL_C ON FIG_RE QU E S T , are exchanged every 1 ms. This is a
butclear
can indication
capture packet data through a switch connected to a
that it is not possible to analyze message
network
sequences without filteringTX/RX
hub. The eNodeB connects
capabilities to the
in a call flowUE RF for
with, unit TABLE
Table III: II CAPTURE
Capture FILTERS
Filters for FOR L2/L3L2
Basic MESSAGES
and FAPI Messages
through attenuators,
instance, 90 ms duration.and operates with single antenna mode in !"##$%"& '()*"+&
Message Filter
downlink. Messages between L1 and L2/L3 were opened and !"#$
Wireshark is anMAC
excellent L1.lte phy%&'())*+#,,-.#,-.!'//01'$
analyzed at FAPI, and tool
RRC to capture and analyze
levels. The eNodeBpacketcould PDSCH HARQ
2"#$PDSCH HARQ
header.msgId == 0x85
L1.lte phy30*(%&'+)4&5$66$789999$
header.msgId == 0x85 and
beflows through easily
also connected to anconfigured
EPC through filters.the
WeGbEemploy the filters
interface of the (ACK)
2"#:$ L1.FAPI fddHarqPduIndication st.harqTB1 == 0x01
%&'())*+/;/&'3"49<)30&5<4#%<*=>;?':$
described
L2/L3 board.in This
Tableinterface
II and III isto not
validate
usedthe in call
the flow in Fig. 3.
experiment. PDSCH HARQ L1.lte phy header.msgId == 0x85 and
2"#$@2"#AB$2"#CB$+++D$ %&'())*+E**F.,<4951$
Wireshark also enables color representation according to the (NACK) L1.FAPI fddHarqPduIndication st.harqTB1 == 0x02
identified message, or to the values assumed by specified L1.lte phyI:+%&'.?F;.F'0J')+3/1"J$
header.msgId == 0x83 and 66$ 78KK$ 04J$
Table II: Capture Filters for L2/L3 Messages PUSCH HARQ
GH,-$@?)'03E%'D$$ I:+LHM".)0*F"4J5*0&5<4./&+4N3O9M)'0
L1.FAPI dlHiDCIPduInfo st.numOfHI != 0x00
message fields. The set of rules used to color packets according L1.lte phy3E%'$P6$787777$
header.msgId == 0x83 and
to the filters are based on the same
Message Filterrules used in the filters. PUSCH HARQ
GHG$ 30*(%&'+)0)$04J$30*(%&'+)0)+&$66$787:$
L1.FAPI dlHiDCIPduInfo st.numOfHI != 0x00 and
(ACK)
MIB lte-rrc.BCCH BCH Message L1.FAPI dlHiPduInfo st.hiValue == 0x01
GG,,<44'*&5<4G'QN'/&$ %&'())*+))*,<44'*&5<4G'QN'/&$
SIB VI. EXPERIMENTAL SETUP
mac-lte.rnti == 0xffff L1.lte phy header.msgId == 0x83 and
lte-rrc.systemInformationBlock ,G$PUSCH HARQ 30*(%&'+J%/*F+%*5J$66$78:*$
L1.FAPI dlHiDCIPduInfo st.numOfHI != 0x00 and
SIB1 The setup used to run the experiment follows the scheme (NACK)
Type1 GG,,<44'*&5<42'&N?$ L1.FAPI dlHiPduInfo
%&'())*+))*,<44'*&5<42'&N?$
st.hiValue == 0x00
depicted
SIB in Fig.
(SIB2, SIB3, ...) 4. The eNodeB is modular,
lte-rrc.bcch as shown in Fig. 5,
Config L1.lte phy%&'())*+))*,<44'*&5<42'&N?,<3?%'&'$
GG,,<44'*&5<42'&N?,<3?%'&'$
SR
header.msgId == 0x8a and
L1.lte phy header.msgId == 0x88 and L1.FAPI rxSRIndication st.numOfSr != 0x0000
RACH (preamble) L1.FAPI rachIndication UL Scheduling L1.lte phy header.msgId == 0x83 and
st.numOfPreamble != 0x0000 Grant L1.FAPI dlHiDCIPduInfo st.numOfDCI != 0x00
RAR mac-lte.rar and mac-lte.rar.t == 0x01 UL CRC L1.lte phy header.msgId == 0x86
RRCConnectionRequest lte-rrc.rrcConnectionRequest UL CRC L1.lte phy header.msgId == 0x86 and
CR mac-lte.dlsch.lcid == 0x1c (OK) L1.FAPI crcPduIndication st.crcFlag == 0x00
RRCConnectionSetup lte-rrc.rrcConnectionSetup UL CRC L1.lte phy header.msgId == 0x86 and
RRCConnectionSetupComplete lte-rrc.rrcConnectionSetupComplete (NOK) L1.FAPI crcPduIndication st.crcFlag == 0x01
Figure 6: Exemplary screenshot showing packets captured using filters defined as in Tables II and III.
the correct RAR has assigned the UE with RNTI equal to 58. the correct receipt of that message in Packet 50679, after
The eNodeB has transmitted a contention resolution mes- four subframes, thus finalizing the connection procedure.
sage in packet number 50338, five subframes after RRC- The periodic broadcasting of SIB and MIB messages that
ConnectionRequest and one subframe after ACK confir- support the UE procedure of system acquisition is shown in
mation to the UE. Note that Wireshark does not cor- Figure 7. Note that SIB is transmitted every 4 SFN, which
rectly detect the contention resolution message, which is corresponds to 40 ms. The SIB1 message is sent every 2
instead identified as malformed. However, this message is SFN (20 ms) and SIB2, along with the remainder SIBs, is
confirmed by the UE with an ACK in packet number transmitted in SystemInformation with periodicity configured
50369 after four subframes. This issue is caused at the by SIB1 of 16 SFN (160 ms).
FAPI level due to some behavior discrepancies with re-
VIII. L IMITATIONS AND D ISCUSSIONS
spect to the reserved bits in the installed FAPI dissector.
Whenever such issue occurs, the “info” field at Wireshark The analysis of only a few procedures is obviously not
display becomes inconsistent and should be disregarded. enough to evaluate the full real-time operation of the ra-
Following the message sequence, notice that in Packet dio protocol stack. Besides, the test methodology has taken
Number 50558 FAPI, message TXRequest is carrying two into account only radio protocol messages from the control
RRC messages, one for SIB and another to RRCConnec- plane. Although the physical channel PDSCH has been tested
tionSetup. The latter RRC message is confirmed by the with this approach, further tests are necessary to evaluate
UE within the expected time frame (after four subframes). PDSCH performance from the user data plane, e.g. to check
However, this analysis is not straightforward because of is- if the protocol stack operates correctly for this channel un-
sues in the FAPI dissector regarding TXRequest messages, der higher data rate scenario. In this case, full experimental
which can erroneously treat this message as malformed. setup considering also EPC core network are required for
After receiving message RRCConnectionSetup, the UE tests involving network attachment and data radio bearers.
requests resources to transmit the message RRCConnection- The limitation of not testing the radio protocol stack from
SetupComplete after 15 ms. The message Scheduling Request the user plane perspective was an option to avoid potential
is correctly detected in Packet 50638 and it is answered with issues that might emerge from core network interactions. A
a transmission grant in Packet 50640 after four subframes. second validation phase is then required to validate these
The transmission grant is received by the UE in Downlink interactions, which include, among other procedures, initial
Control Information Format 0 (DCI0) of PDCCH, which security activation, radio bearer establishment and handover.
allows the transmission of RRCConnectionSetupComplete. At However, recall that the protocol subsystem already is a
the eNodeB, this message is received in Packet 50677, four
subframes after sending the grant. The eNodeB confirms
IX. C ONCLUSIONS
This paper has described the experimental valida-
tion of a radio protocol stack designed for the base-
band unit of LTE base stations. Procedures for system
acquisition, random access and RRC connection estab-
lishment were considered as a way to demonstrate the
joint operation between physical layer and the remain-
ing protocol stack without core network interactions.
Our results have also confirmed that the UE emu-
lator SORBAS successfully executes essential tasks dur-
ing these procedures, e.g. to configure its protocol stack
on the basis of received system information. RNTI
identification was obtained during random access pro-
cedure as expected. Further, the radio bearers between
UE and eNodeB were also successfully established.
Finally, the proposed test methodology has allowed the
complete radio protocol validation from the control plane
perspective. This approach exercises all physical channels,
transport channels, logic channels, RLC services, as well as
layer 3 control tasks, without involving any interfaces with
the core network.
ACKNOWLEDGMENT
This work was supported by Funttel project “RASFA-4G”
at CPqD Telecom and IT Solutions.
R EFERENCES
[1] GSMA Intelligence, “Global LTE network forecasts and assumptions,
2013–17”, Nov. 2013.
[2] J. Perala, P. Jurmu and J. Pinola , “Hybrid Approach for Protocol Testing
of LTE System”, In Proc. IEEE Int. Conf. on Advances in System
Testing and Validation Lifecycle, pp. 32-36, Aug. 2010.
[3] The Aricent Group, “eNodeB Protocol Stacks and Software
Frameworks”, [Online] Available: http://www.aricent.com/software/lte-
enodeb-software-framework.html.
[4] Signalion GmbH. [Online] Available: http://www.signalion.com. (This
product was discontinued).