Giving the instantaneous channel conditions (from the feedback of UE) and currentavailable resources, a system with AMC can choose the appropriate downlinkmodulation and coding format. When UE is in favorable positions (e.g. close to basestation or LOS link), higher order modulation and higher code rates are assigned, suchas 16QAM and 3/4 rate, to achieve higher peak rates. On the contrary, for UE inunfavorable positions (e.g. cell border), lower order modulation and lower code ratessuch as QPSK and 1/4 rate are assigned to ensure the quality of communication.
In case of WCDMA, the system adjusts transmit power to avoid the effects of pathfading in order tomaintain a relatively stable data rates for the communication. Moderately,in HSDPA the fast link adaption methods take the place of power control mechanism,AMC technique will select corresponding data rate based on channel condition, henceforth the interference variation is reduced.
Hybrid Automatic Repeat request
A system with HSDPA chooses rude data rates through AMC, and then uses a newerror control scheme, Hybrid Automatic Repeat request, HARQ, technique to adjustthe coding rates precisely, thus improves the link adaption accuracy and the efficiencyof channel utilization.
Under the radio environment, the fading effects, the interference from other users andchannel noise have significant influence to the transmission quality. Thus, the radiolinks has higher and time varying error rates compared to the wired links. That's why a proper error control mechanism is badly required.
There are basically two error control algorithms, namely Automatic Repeat request which is ARQ and Forward Error Correction normally called as FEC. ARQ is often applied on different wired links, which should have a receiver in order to request for the retransmission of the lost or corrupted data packets.
Comparatively, FEC is adopted mainly in wireless communications, which transmits some redundant data, called parities, together with original data to allow reconstruction of lost/corrupted packets at the receiver. Nevertheless too much FECredundancy will reduce the transmit efficiency, in that the scheme of Hybrid ARQ isintroduced to combine FEC and ARQ.
A general ARQ algorithm says that if a received packet cannot be decoded effectively, itwill be discarded directly and retransmission is requested. No knowledge of theprevious transmission will be used when the retransmitted packet is received anddecoded. Comparatively, in a HARQ scheme, the corrupted packet is not discardedbut stored in the buffer of the receiver instead. When the retransmitted packet isreceived, it will be combined with the previous transmission of the same informationbits, this process is called soft combining. Then this combined signal is puttedinto the decoder and if again it fails in decoding then the further retransmissions (up to a preset number defined by the present system) will occur and is soft combined till the packet is decoded effectively. Thesoft combining process of HARQ enhances the possibility of a successful decoding ofthe information bits, that's why increases in the transmit efficiency will be achieved.
Mainly two different types of HARQ schemes are defined in the 3GPP specifications:
First one is the Incremental Redundancy and the second one is the ï€ Chase combining.
During the simulations of the present thesis, the HARQ scheme of chase combining is implemented.
This scheduler controls the whole distribution of shared resources to users and it determines theoverall behavior of the system to a large extent. Fast scheduling is applied mainlybased on channel conditions in order to exploit AMC and HARQ to their maximum potential and should also concern the amount of data waiting for transit and the priorities ofservices at the same time. This scheduler tells about the multi-user diversity and strives for transmitting to the users when radio conditions will permit high data rates; italso maintains a certain degree of fairness.
To fit the rapid variation of radio link channel conditions better, the packet scheduleris moved from the radio network controller (RNC) to the Node B where it has easyaccess to air interface measurements in Release 5. And a shorter 2ms transmissiontime interval (TTI), which represents the periodicity at which a data block set istransferred by the physical layer on the radio interface, is also introduced instead of 10ms TTI, to get the instantaneous channel conditions accurately.
The Scheduling algorithms for WCDMA systems range from max C/I basedscheduling, which provides the best efficient use of resources, to Round-Robinscheduling which keeps the best fairness for different users. Between these twoextreme methods, a relative C/I based scheduling, which is called FairChannel-Dependent scheduling is also used base on different system servicesrequirement. Descriptions of different types of scheduler are stated below:
Round-robin scheduler (RR) is the scheduler which selects the user that has not been served for the longest time period. The difference and variations of channelconditions for each user is not taken into consideration.
Maximum C/I Scheduler (Max C/I): This scheduler essentially ranks all the usersaccording to their instantaneous carrier-to-interference (C/I) ratios. The present scheduleris most favorable for obtaining the maximum network throughput. In the present situation, theUE in favorable positions will have the highest throughput, but the system servicesmay be unavailable for the users in inauspicious positions. Max C/I are a kind ofchannel-dependent scheduler since the variations of radio channel condition isused for scheduling.
Fair Channel-Dependent Scheduler (FCDS): It is a more practical schedulerwhich has a strategy that incorporates the RR method and the Max C/Imethod, i.e. it uses variations of the radio channel conditions to improve systemcapacity while implementing a degree of equality. Hence forth thismay be taken as atrade-off in between the two extreme scheduling methods.
We introduce the following algorithm to explain their relationship:
In the presentalgorithm the performance of the three scheduling methods used in HSDPA istested and compared.
HSDPA relies on a new transport channel which is called High Speed Downlinkshared channel (HS-DSCH). It can be understood that HS-DSCH can only be implicated on packet switching domain and only for HSDPA is a packet-based data service.
The transport channel which carries data transfer services is the Dedicated Channel (DCH) in 3GPP. Compared with the DCH, the HS-DSCH has the followingfeatures:
First one is higher capacity, second one is ï€ Reduced delay and third one is ï€ Higher peak data rates.
According to the architecture shown in figure, we can see theRLC and MAC-d layers remain unchanged. The mostdistinct difference is that HS-DSCH is controlled by the MAC-hs which located inNode B. When the data is transferred over HS-DSCH, the corresponding RLC entitycan operate in either AM or UM mode, but not in TM mode.
HS-DSCH is a pure downlink channel, therefore an associated uplink DCH is alwayscreated alongside the HS-DSCH to transmit related data and control information.
In HSDPA, a new MAC entity with additional functionality is added and is placed in
Node B, hence it is called MAC-hs. Two new sub-entities are included in MAC-hs,one is for scheduling, common functions of scheduling and priority handling in
Release 99 is also included in this entity. The other sub-entity of MAC-hs is for
HARQ, the MAC level retransmissions are hence operated between Node B and UE.
In Release 99, the retransmissions are handled only in RLC level, thus if a receivedpacket is corrupted, the retransmission is sent from UE to the RNC, via Node B,whereas after MAC-hs is added in Release 5, the RLC retransmission only happenswhen the limit of HARQ retransmission times is reached. This provides the advantageof reducing the delay for successful decoding of packet by providing a shorterround-trip of retransmission, since the MAC-hs is allocated in Node B.
In this section the process of packet delivery is introduced, detailed views of MAC-hsarchitecture on both UTRAN and UE side are then given respectively.
3.3.1 MAC-hs Packet Delivery
Before discussing the packet delivery, the definitions of Service Data Unit and Packet
Data Units are given below:
A packet transmitted to a higher layer from current layer or received from a higher layerto current layer is called Service Data Unit, or SDU.
ï€ A packet transmitted to a lower layer from current layer or received from a lower layerto current layer is called Protocol Data Unit, or PDU.
If an HS-DSCH is allotted, the packets are generated and transported with the followingsteps:
Initially the MAC SDU will be condensed into MAC-d PDU in MAC-d entity.
ï€ These MAC-d PDUs are preserved as MAC-hs SDU and then are transmitted fromMAC-d to Mac-hs.
ï€ One or more then one MAC-hs SDU hence buildup a MAC-hs PDU with necessary headerinMAC-hs entity.
ï€ To conclude the MAC-hs PDU are distributed from MAC-hs to physical layer, one
MAC-hs PDU can be transmitted in a TTI per UE.
A MAC-hs PDU is known as a transfer block (TB) also. The transfer block size, i.e. howmany MAC-hs SDU are included in MAC-hs PDU is varied base on the value of
Channel Quality Indicator, CQI, received instantaneously at Node B, which is sent byUE to indicate the estimated channel condition. A high value of CQI represents goodchannel condition, so the transfer block mighthave extra MAC-hs SDU,whereas a low value of CQI signifies bad channel situations and the transfer blockwill contain less SDU.
MAC Packet Formats
After the addition of MAC-hs entity Release 5, the packet formats is more complicated thenin Release 99, a fresh packet type MAC-hs PDU is introduced plus to the former
MAC-d PDU. The packet formats for non-high-speed and high-speed mode areexplained below in the figures.
A MAC-d PDU consists of an MAC header and a MAC Service Data Unit (MACSDU).
Each MAC-hs PDU contains a header and one or several MAC-hs SDU.
MAC-hs Architecture, UTRAN side
The MAC-hs will handle the data transmitted on the HS-DSCH.
Moreover it has to manage all the physical fundsassignedto HS-DSCH. From Figure it is evident that the MAC-hs on the UTRAN sidecomprises of four functional entities as described below:
The flow control unitin MAC-hs is assigned between Node B and RNC, so as toguarantee that the MAC-hs buffers always compriseof enough data to utilize the providedphysical layer resources efficiently, while evading buffer overflow as well.
In the Scheduling/Priority Handling unit, the data of MAC-d flows are arranged intolines by the priority and the MAC-d flow it belongs to. The scheduling entity couldthen make use of these priority queues when making a scheduling decision. Under thecontrol of the scheduler, one or numerous MAC-d PDU from one among the priority queuesisamassed into a transfer block. A MAC-hs header, which containssuch things like thequeue identity and transmission sequence number (TSN), is added to form atransport block, or MAC-hs PDU. This transport block is again promoted to thephysical layer for further processing.
One HARQ object handles the hybrid ARQ functionality for one user, and there shallbe one HARQ process per TTI.
The Transport Format Resource Combination (TFRC) selection objecthas to choose an suitable transport format and resource combination for the data tobe transmitted on HS-DSCH.
MAC-hs Architecture, UE side
Figure following shows a detailed view of the two functional entities that are found in theMAC-hs at the UE: HARQ and Reordering:
The HARQ objecthas to handle the HARQ process. Simply one HARQprocess will exist for each HS-DSCH per TTI and will handle all the tasks requiredfor hybrid ARQ, e.g. generating ACK/NACK packets.
The Re-ordering Queue distribution objectlines the successfully received datablocks as per their Transmission Sequence Number (TSN) and their priorityclass. One re-ordering queue objectis present for each priority class. Then the data blockde-assembly entity will generates the appropriate MAC-d PDU flows from theRe-ordering queues.
To assess communication system performance, we choose ns-2 as our simulator. Itpreviously supported some wireless modules, but the functionality provided by thosemodules was not sufficient to simulate a radio link with UMTS HSDPA, so weutilize a contribution module EURANE of ns-2 to implement the simulation ofHSDPA.
The UC Berkeley people have designed an event driven network simulator called Ns-2. Currently it isa portion of VINT project and implements traffic behaviors, network protocols,routing, etc. for simulation. Because it is open source software, during thedevelopment many contributions has been included from other researchers it hasbecome a mutual tool for network researchers to simulate and evaluate networkrelated project.
In this paper Ns-2 simulator has been implemented and runs with two programming languages: C++ forthe object oriented simulator, and OTcl (an object oriented extension for Tcl)interpreter for performing user configuration scripts before the simulation begins.
There are all the time two corresponding hierarchies for every protocol or network objectsimplemented in ns-2, the compiled C++ hierarchy and the interpreted OTcl hierarchy. The C++ hierarchy allows user to achieve efficiency and faster running time inthe simulation especially for the detailed definition and operation of protocols since itcan reduce packet and event processing time.
Through the friendly interface of OTcl language, users can define a particular networktopology, the protocols and applications that they want to simulate and the form of theoutput that they want to obtain from the simulator quickly and clearly in an OTcl script.
Ns-2 is a discrete event simulator. The timing of events is maintained by the scheduler.
In ns-2 an event is an object, e.g. a packet in the C++ hierarchy with anexclusive ID, ascheduled time and the pointer to an object handles the event. The scheduler queuesthe events to be executed in an ordered structure and triggers them one by one,together with the handler of the event.
NS network components mainly have two types of elements, the compound networkcomponents and the basic network components. Figure shows a partial OTcl classhierarchy of basic network component in NS. A compound component containsseveral basic components with different functions, to play a complicated role duringthe simulation.
On the root, NsObject class is the superclass of entirelyall rudimentary network component objectswhich handle packets that mightcomprise compound network objects such as nodesand links. Depending on the amount of the possible output data paths, the basic networkcomponents can be further separated into two subclasses,
Connector and Classifier.
Objects belonging to Connector class have merely one output data path, correspondingly,while in the Classifier class, the objects that have possible multiple output data paths are categorized.
Node is one of the primary objects in ns-2. A node consists of a node entry object and
classifiers as shown. Unicast node is the default node type in ns-2, which has an
AddrClassifier that does unicast routing and a port classifier. A structure of unicast
node is shown in figure 4.3. Another type of node is multicast node, which added a
McastClassifier to unicast node, for handling multicast packets.
Link is another essential object to establish a network scenario for simulation. In ns alink object represents a single direction, simplex link. To create a duplex-link forsimulation, two simplex links in both direction. Figure 4.4 shows how a packet ishandled in a simplex link:
An agent in a node can be either an active or a reactive component during networkimplementation. Links and nodes are reactive components that react to incomingpackets and apply their behaviors to the packet. Compared to them, agents can be anactive component. For instance, when transport agents TCP and UDP react to externalorders for sending data, they may generate necessary packets for data transmission orgenerate connection control packets. Another example is routing agents which canperform dynamic route calculations. They actively generate packets and send them toother agents in same type to determine routes within the network. In the case oftransport agents, they require external components to generate data to send.The function of traffic generators is to generate data in the network. It could be eithera data source that sends data considering a traffic distribution, or a source thatsimulates an application. Two examples of the former are constant bit rate (CBR) andexponentially distributed traffic in terms of active and inactive sending. Two examplesof the latter are FTP-transfer and a Telnet-connection. Traffic generators use transportagent to introduce data into simulated network.
To analyze the result of the simulation, the traffic must be traced during the process.
When a link is sketched, the trace objects (EnqT, DeqT, DrpT and RecvT) are slotted inwithin the link objects as presented in Figure. For every packet that passes a traceobject, information about the packet is written to the specified trace file.
Even though ns-2 doesn't offerfitting modules for UTRAN/UMTS, a number ofcontributed works was developed such as Pablo Martin and Paula Ballester'scontributed modules, and the ns and UMTS model by Alfredo Todini and Francesco Vacirca's. Within those modules EURANE is the only one which providesthe support of HS-DSCH.
The simulator EURANE (Enhanced UMTS Radio Access Network Extensions forns-2) is one of the main of SEACORN (Simulation of Enhanced UMTS Access andCore Networks), which investigates enhancements to UMTS for RAN and CNthrough simulation. EURANE is an end to end extension which adds three extra radiolink nodes to ns-2, the Radio Network Controller (RNC), Base station (BS) and theUser Equipment (UE). These nodes have provision for four kinds of transport channels;these arededicated Channel DCH,common channels FACH and RACH, and high speedchannel HS-DSCH. In EURANE, R99 channels, i.e. DCH, RACH and FACH, usestandard error model provided by ns-2, for HS-DSCH, a pre-computed input powertrace file and a block error rate (BLER) performance curve are introduced to generatethe error model of high-speed channel.
Nodes & Link
The Node RNC, BS and UE are all implemented from new object class, namely
UMTSNode class. Based on different configurations, different kinds of classifiers andlink objects are used to compose different nodes. For instance, when creating a BSnode, bandwidth and TTI of both uplink and downlink can be set respectively. Themost important parameter should be determined first is the node type, after that, othercharacters of this node type can be configured further.
Each UMTS node has zero or more UMTS network interfaces (NIF) stacks,composed of objects representing different layers in the stack, the major componentsbeing RLC, MAC and physical layer objects. Channels are connected to the physicallayer object in the stack. NIFs are also important for packet tracing since the commonmethods in ns-2 can not trace the traffic within radio links.
Typically, NIF stacks at the UE will have all of these objects but those at the BS willonly have MAC layer and physical/channel layer objects. At the RNC, each NIF stackwill only be composed of one RLC layer object.
The main functionality additions to ns-2 come in the form of the RLC AcknowledgedMode (AM) and Unacknowledged Mode (UM), which is implemented at RNC andUE. The RLC entity AM is used for the Dedicated Channel (DCH) and the two
Common channels (RACH and FACH) in acknowledged mode, correspondinglyanother RLC entity AM-hs is developed to support HS-DSCH. Unacknowledgedmode is sustained for DCH and HS-DSCH as well,by the subsets of AM and AM-hs,namely UM and UM-hs, correspondingly. Without indication, all simulations in this thesisare taken under AM.
4.2.3 MAC Object
Also on MAC layer, the implementation of MAC architectures is separated to supportnormal channels and high speed channel. The basic MAC utilized for the DCH andcommon channels (RACH and FACH), and the furthercomplex MAC-hs used forthe HS-DSCH.
Scheduling plays an important role in HSDPA, and is implemented on MAC-hs entityin EURANE. At every TTI the MAC-hs checks the status of the flow/priority buffers,and depending on the scheduling algorithm, determines which packets (and theamount) will be used to construct a MAC-hs PDU for transmission.
We introduce two key parameters of scheduler settings here: one is the parameter"scheduler type" which defines the type of scheduling performed at the MAC-hs. Avalue of 1 configures Round Robin Scheduling to be used, while a value of 2 wouldconfigure the scheduling to be Maximum C/I. A value of 3 would set the scheduler tooperate as Fair Channel-Dependent Scheduling (FCDS). Another parameter "alpha"comes from the formula in section 3.1.3. It is used when the scheduler is set to FCDSwhich defines the amount of weighting used in the algorithm. A value of 0.0 wouldequate to the Round Robin case, while a value of 1.0 would equate to the Maximum
C/I case .
Error Model of HS-DSCH
To apply the physical layer, EURANE utilizes a standard ns-2 channel object tolink the BS and UE. This is combined with the attachment of an error model.
For DCH and common channels, a standard ns-2 error model which defines the packetloss rate directly is used, a slight difference with original ns-2 is the error model isadded to the specified interface on the given radio node.
A further complicated transmission error model is executed for HSDPA, this errormodel is the pre-processed from ns-2 and contains two parts. The first part is a physicallayer simulator which will be used to produce a BLER performance curve, and the other one is an input tracefile of received powers and CQI produced from Matlab scripts.
Trace Format of EURANE
General method of tracing is not compatible with new UMTSnodes, therefore a special UmtsTrace objects (class UmtsTrace derives from class
Trace)  is used to trace RLC PDU (i.e. MAC-d SDU) inside the UTRAN and theUE. The trace format for UMTS nodes is like the figure:
ïƒ˜ï€ Event: Each trace line starts with an event (+, -, d, r) descriptor, which representthe action of enqueue, dequeue, drop and receive of packets, respectively.
ïƒ˜ï€ Time: the simulation time (in seconds) of that event.
ïƒ˜ï€ From node and to node: These two fields identify the link on which the eventoccurred.
ïƒ˜ï€ Pkt type and pkt size: the packet type and size in Bytes. One moreamendment of
EURANE trace file format is the "pkt type" which possessesseveral additionalpresentation todefine the behavior of radio link.
ïƒ˜ï€ Fid is the flow id of IPv6 that a user can set for each flow at the input OTcl script.
ïƒ˜ï€ Srcaddrand dstaddrare source and destination address in forms of "node.port".
ïƒ˜ï€ Seqnum, the network layer protocol's packet sequence number. It is mentionable thatalthough UDP implementations do not utilize sequence number, ns-2 takes record ofUDPpacket sequence number for analysis purposes.
ïƒ˜ï€ Pkt id shows the unique id of the packet which is the key point to trace a packet.
RLC seqrepresents RLC sequence number is an extra object which EURANEadds to trace RLC packets.
To be able to test the features of HS-DSCH, additions to EURANE extension of ns-2are needed. This partdefines the design and implementation of the model of
DCH power consumption of EURANE
5.1 Design Decisions
UMTS is a system where the voice and data services are integrated. Generally the basestation (Node B) will contain a set of different transport channels, e.g. commonchannel (FACH and RACH) for voice services, dedicated channel (DCH) andHS-DSCH simultaneously. Besides the power provided to common channels andDCH channels, the rest part can be all allocated to HS-DSCH, as shown in figure 5.1,thus to keep the maximum efficiency of utilization to the power of base station. In thiscase, the existing power of HS-DSCH is not a constant value and differsdepending on theamount and channel conditions of other channels within base station.
In EURANE, the input power trace files are pre-computed from Matlab scriptstogenerate the Channel Quality Indicator, a pre-requisite parameter P_outis calculatedwith the following equation.
In this equation,Iinteris the inter-cell interference,Iintrais the intra-cellinterference,andPTxis overall transmitted power of base station, and. The last parameter Plossis the totalpath loss which depends on the user's configuration, containing the fast fading, slow(i.e. shadow) fading, distance loss and base station antenna gain. All terms inthis equation are logarithmic. SoP_outcan be considered to be instantaneous estimatedsignal to noise ratio, SNR, of the channel, and can be further mapped to an appropriateCQI.
The limitation of this formula is, however, that the base station is supposed to operateHS-DSCH only. Hence the available transmit power is a constant (set to 38dBm inEURANE Mat lab scripts).
It is interesting and important to explore the behavior of HS-DSCH under thevariation of the transmit power, which is more reasonable than a constant value, andto observe the effects of other transport channels to HS-DSCH. Our aim is to add apower consumption models of DCH channels to the original Matlab process, thereforethe available power of HS-DSCH is given by DCH power subtracting from the totaltransmit power PTx.
5.2 DCH Power Model
To introduce the DCH power consumption in EURANE, two primary parts ofmodification should be applied to the original Matlab scripts, one is the user interfaceto configure related parameters, and the other part is the calculation of DCH powerconsumption per TTI and eventually the calculation of the available power forHS-DSCH, using the equation (5-1).
During the implementation, a new script was created to provide an interface to definethe input DCH parameters, e.g. the environment, number of DCH users, starting andending time etc, and to generate DCH power trace by the given parameters. To createthe existing power trace of base station, selected code in the original scripts has beenchanged.
The derivation of mathematical relation to create DCH power consumption forsimulation is discussed below:
Suppose that a DCH channel is distributed to a user i with a certaincarrier-to-interference ratio (CIR), ãi. In a simplistic model, ãican be calculated bydividing the universal signal-to-interference ratio (SIR), Ã0 by the spreading factor SF,which is often referred to as the processing gain PGi.
A concept of a geometry factor GFiis introduced to indicate the relation betweeninterference in the own cell, Iintra,i, the external interference other base stations,Iinter,i,and thermal noise Ni. It is essentially the signal to noise ration between own andexternal interference
The link attenuations consist of three components: path gain, shadow fading andmultipath fading in increasing order of frequency content (i.e. the latter comprise mostspatial variations of the three). When the path gain and shadow fading are given gpsand themultipath fading is gmf, if the total base station power is P, then the intra-cellinterference is
This is the aggregate received power from the own base station. Since the downlink codesare orthogonal in an ideal situation, but affected by the receiver performance, thefraction a of this interference is perceived as effective interference.
Concerning the average situation, assume that the geometry factor equation still holds
if we only considering the shadow fading and path gain for the interference of own
cell, Iintra,i. This yields
Finally, CIR of a user i using the power pi can be expressed as
The service requires a certain CIR. Combining equation (5-2) and (5-6), we can solvethe required power pi as
Each value of the equation is in linear scale, the universal signal-to-interference ratio
Ã0 is given the fixed value of 2. Corresponding to PTxmentioned in 5.1, which is setto 43 dBm , the total transmit power P has the value of 20 (Watt). The effectiveinterference a is also set to the fix value 0.64  since it is enough to have aappropriate result for simulation though it varies in practice.
The geometry factor GFiis configurable in the integer rangeof 1 to 1000.
Where 1 represents the worst condition, e.g. the interference power of other base
stations is close to the interference of own cell, and 1000 represents the best condition,
e.g. the interference power of other base stations is far less than that of own cell.
The processing gain PGi, i.e. the spreading factor, of DCH channel can be associated
with data rate Ribased on a simple model of Release 99.
Hence the spreading factor is determined by the choice of data rate, e.g. given the
default value of 64 kbps, the spreading factor will be 32. Higher request of data rateresults in lower processing gain and therefore the increase the power pi, based onequation.
According to previous description, within the parameters in (6), PGiand GFiisconfigured by users, Ã0, P, a are constants, the method to generate multipath fadinggmf, which is related with the environment chosen by user, is given by original Matlabprocess. Thus the power of each pi is computed by this equation, and the sum of pifinally buildup the DCH power consumption. Hence the available power of the basestation can be computed:
Therefore P_outin equation (5-1) can be calculated using the logarithmic form of
availableinstead of PTx.