See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/280934099
HSDPA: An Overview
Article · June 2014
CITATION READS
1 297
2 authors, including:
Alagan Anpalagan
Ryerson University
162 PUBLICATIONS 1,149 CITATIONS
SEE PROFILE
All content following this page was uploaded by Alagan Anpalagan on 13 August 2015.
The user has requested enhancement of the downloaded file.
Telecommunications / Télécommunications
HSDPA: An Overview
1.0 Introduction
he requirements for future wireless communication systems by: Ryan Caldwell and Alagan Anpalagan
T are rapidly evolving as applications increase in complexity.
Many users require access to real-time applications, the
internet, and high speed file transfer on a regular basis. To
Ryerson University, Toronto, ON
meet the demanding throughput and delay requirements of
these services, the 3rd Generation Partnership Project (3GPP) has devel-
oped a new high speed data transfer protocol named High-Speed Abstract
Downlink Packet Access (HSDPA). High Speed Downlink Packet Access (HSDPA) is a packet-based
In the 3GPP standards, Release 4 specifications of this protocol provide data service in W-CDMA downlink over a 5 MHz bandwidth. It is
efficient IP support enabling provision of services through an all-IP core intended to provide higher capacity, reduced delay and sign ifi-
network. Release 5 specifications focus on HSDPA to provide data rates cantly higher data rates. This article first describes its physical
to support packet-based multimedia services. HSDPA is evolved from layer and then discusses four important implementations of
and is backward compatible with Release 99 WCDMA systems. Fol- HSDPA. They are: Adaptive Modulation and Coding (AMC),
lowing feasibility studies in 2002, the wireless industry is in trial mode Hybrid Automatic Request (HARQ), Fast Scheduling and Fast Cell
to implement HSDPA in the near future. Selection (FCS).
HSDPA is expected to provide a significant performance increase over
the 2 Mbps data transfer capabilities of WCDMA. Throughput is
expected in excess of 10 Mbps, increasing to nearly 20 Mbps when Sommaire
combined with a Multiple-Input Multiple-Output (MIMO) antenna HSDPA est un service de données qui est basé sur la transmission
structure. The substantial increase in throughput is realized by means of de paquets en downlink W-CDMA sur bande passante de 5 MHz.
a fast link adaptation scheme that will utilize Adaptive Modulation and Ce service a pour objectifs d’offrir une plus grande capacité, une
Coding (AMC). This protocol operates by transmitting with a constant réduction des délais et des taux de transmission des données plus
power, while the Modulation and Coding Scheme (MCS) is altered to élevés. Cet article décrit les couches physiques et les quatre impor-
match the instantaneous channel conditions experienced by the User tantes implantations du HSDPA. Les implantations du HSDPA
Equipment (UE). If errors occur in the transmission process, Hybrid sont les suivantes: modulation adaptative et code (AMC), demande
Automatic Repeat reQuest (HARQ) is employed to quickly retransmit automatique hybride (HARQ), ordonnancement rapide et sélection
packets at the link layer. Finally, fast scheduling is used to quickly cellulaire rapide (FCS).
schedule users when experiencing a constructive fade, maximizing
throughput and reliability to these users.
2.0 Physical Layer Within each 2 ms TTI, a constant spreading factor of 16 is used for code
To facilitate operation of HSDPA, three new channels have been added multiplexing, providing a maximum of 15 parallel channels for the HS-
to the WCDMA platform. These channels include the High-Speed DSCH. These channels may all be assigned to one user during the TTI,
Downlink Shared Channel (HS-DSCH), High-Speed Shared Control or may be split amongst several users. The number of parallel channels
Channel (HS-SCCH) and the Uplink Dedicated Physical Control Chan- allocated to each user depends on cell loading, QoS (Quality of Ser-
nel (DPCCH). The following sections discuss these channels in detail. vice) requirements, and the capabilities of the UE.
2.1 High Speed Downlink Shared Channel (HS-DSCH) 2.2 High Speed Shared Control Channel (HS-SCCH)
The HS-DSCH is the primary radio bearer for this technology. As the The HS-SCCH is used to signal a variety of information to the UE
name suggests, this channel is provisioned as a shared resource for all before the beginning of each scheduled TTI. This includes the channel-
users in a particular sector. During each transmission slot, users are ization-code set, modulation scheme, transport block size, and HARQ
assigned an MCS level that will maximize throughput, while maintain- protocol information. Although all of this information is required to suc-
ing a low probability of retransmissions. cessfully decode the data on the HS-DSCH, the channelization-code set
and the modulation scheme are time critical parameters. These identify
Primary channel multiplexing occurs in the time domain, where each which parallel codes on the HS-DSCH contain data for the UE, and
Transmission Time Interval (TTI) consists of three timeslots, totaling 2 whether QPSK or 16-QAM is used in the upcoming transmission. If this
ms. The selection of a 2 ms TTI has resulted in several benefits over the data is not received before the beginning of the HS-DSCH TTI, the data
10 ms TTI used for data transfer with WCDMA. This includes reduced will be buffered until the UE is aware of these parameters. This
round trip delay and higher validity of the channel estimation mecha- increases the delay and buffering requirements considerably. One might
nism. For instance, if a longer TTI is combined with AMC, more imagine that this could be solved simply by sending all this information
channel variations will occur during the transmission, possibly corrupt- prior to the TTI. This, however, will increase the delay between the
ing the data packet. channel quality feedback and the transmission, reducing the validity of
Figure 1: Timing Relation
22 IEEE Canadian Review - Spring / Printemps 2004
the channel estimation process. As a result, the 3GPP has elected to sep- net and file transfers, which already rely on TCP/IP for higher layer
arately interleave the time critical data, and send it in the first 0.667 ms error correction and flow control.
slot of the HS-SCCH. Figure 1 shows this timing relation. To limit the complexity of the retransmission process, the 3GPP has
Since this channel is only required prior to the HS-DSCH TTI, it is selected the Stop-and-Wait (SAW) protocol. This operates by transmit-
favorable to provide a shared resource for this purpose. For this reason, ting one packet and awaiting a response from the UE. The problem,
the UE must monitor between one and four HS-SCCH channels simul- however, is that the system is idle while awaiting acknowledgements.
taneously. An HS-DSCH Indicator (HI) is then sent on the users To reduce this inefficiency, the 3GPP has selected an N-Channel SAW
Dedicated Channel (DCH) to indicate the control channel that contains protocol. While one channel is awaiting an ACK or NACK, the other
data for the UE. By sharing this resource, less power is required for sig- (N - 1) channels continue to transmit. Although the value for N is still
naling, increasing the system’s efficiency considerably. being evaluated for performance and complexity, it is expected to be
between two and four.
2.3 Uplink Dedicated Physical Control Channel (DPCCH) Another benefit of HARQ is that it uses a three stage virtual buffer to
store a soft copy of the previously transmitted packet. When a retrans-
The third new channel required to implement HSDPA is the Uplink mission occurs, the data is combined with data stored in the soft buffer
DPCCH. This channel is responsible for uplink signaling of Acknowl- to effectively increase the coding gain. This enables the retransmission
edgements (ACK) and Negative Acknowledgements (NACK) to process to require fewer transmissions, and increases the average
indicate the status of the previous packet. The Channel Quality Informa- throughput as a result.
tion (CQI) is also sent on this channel. A five bit value is used to
indicate which modulation and coding schemes are suitable for the The method of combining packets with those stored in the soft buffer is
upcoming transmission. Different codes indicate MCS levels ranging very important to the performance and complexity of the retransmis-
from QPSK using R = turbo coding, to 16-QAM with R = turbo sion process. Two fun damental schemes have been proposed to
coding. One state is also reserved to indicate that no transmission accomplish this task. These include Chase Combining (CC) and Incre-
should be made in the event of extremely poor channel conditions. mental Redundancy (IR).
Chase Combining is the least complex combining scheme; however, it
3.0 Adaptive Modulation and Coding provides a lower coding gain than IR. The Node B simply retransmits
the original packet, and the UE combines this with data stored in the
Adaptive Modulation and Coding is the fundamental technology that soft buffer.
allows HSDPA to surpass the data rates of its predecessors. Tradition-
ally, systems that utilize Code Division Multiple Access (CDMA) have The IR schemes, however, provide additional coding gain by transmit-
used a constant modulation scheme (usually M-PSK), with fast power ting parity bits in the retransmission. The result is that fewer
control to adapt to changes in channel conditions. Instead, AMC trans- retransmissions are necessary to successfully retransmit packets. This is
mits with a constant power while the Modulation and Coding Scheme is especially true under poor channel conditions, or when the user is trav-
altered to adapt to these variations. This results in higher average eling at a high velocity.
throughput because higher order MCS levels are assigned to users expe-
riencing favorable conditions. Spectral efficiency is also increased 5.0 Fast Scheduling
because the highest possible MCS level is utilized during each
transmission. The fast scheduling entity is also very important in the operation of
The selection of the MCS level is done to maximize throughput, while HSDPA. One primary change from the previous implementations is that
maintaining a low probability of retransmission. In effect, if a user is the scheduler is located at the Node B. This enables the scheduler to
experiencing favorable channel conditions, a high order modulation quickly respond the changes in the channel conditions, and ensures that
scheme such as 16-QAM with R = turbo coding may be used to max- the UE is served while on a constructive fade. There are three main
imize throughput for this user. Conversely, if the channel conditions are types of schedulers that have been proposed for HSDPA. These include
poor, QPSK with R = turbo coding can be used to provide higher reli- Round Robin (RR), Maximum C/I, and Proportional Fair (PF).
ability in the transmission. Further granularity is accomplished by using The RR scheduler operates by scheduling users based upon their posi-
code rates from to with the previously mentioned modulation tion in a first-in first-out queue. Although this scheduler provides the
schemes. least complex operation and the most fairness between users, the UE’s
The MCS level used in the upcoming transmission is selected based channel conditions are not taken into consideration. As a result, users
upon the power measured on the Common Pilot Channel (CPICH). The may be scheduled when experiencing a destructive fade, causing the
UE then encodes the data and transmits this CQI on the DPCCH. The packet to be corrupted.
UE is then scheduled by the Node B (base station controller), where the As an alternative, the Maximum C/I algorithm schedules users when
final selection of the MCS level and channelization-code set is per- their instantaneous SIR is the highest amongst all users at the respec-
formed. This functionality is located at the Node B to ensure that tive base station. This scheduling algorithm ensures that all users are
fairness is retained between users, and the majority of QoS require- served on a constructive fade, and as a result, has a higher percentage of
ments can be satisfied. successful transmissions. Also, the throughput and spectral efficiency is
maximized because the highest possible MCS level is used during each
4.0 Hybrid Automatic Repeat Request transmission. The disadvantage, however, is the lack of fairness
between users in the sector. In normal circumstances, the geometry of
Although the MCS level is selected to ensure a reasonable probability radio propagation causes nearly 50% of users to be located near the cell
of a successful transmission, errors do occur in any wireless system. border. This indicates that nearly half of the users may receive inade-
This is a result of highly variable channel conditions caused by interfer- quate service as a result.
ence from other users and base stations. Under normal circumstances,
approximately 10 - 30% of first transmissions must be resent to be suc- A compromise between these two schedulers is the Proportional Fair
cessful. For this reason, the choice of the retransmission protocol is vital algorithm. This schedules users based on the offset between the long
to the performance of any wireless communication system. term average SIR and the instantaneous SIR. The result is that each user
is served while on a constructive fade, while fairness is maintained
The 3GPP selected HARQ for retransmissions because of its ability to because instantaneous channel conditions will exceed the long term
quickly retransmit packets. HARQ functionality is implemented at the average at some instant. Figure 2 shows the operation of the previously
MAC (Media Access Control) layer, as an alternative to the RLC discussed scheduling schemes.
(Radio Link Control) layer used for many other data transfer protocols.
This decreases delay considerably because this entity is located at the 6.0 Fast Cell Selection
radio interface. In normal instances, a negative acknowledgement may
require less than 10 ms at the MAC layer, while this process may take In any cellular comm unication system, handoffs are necessary to
between 80 and 100 ms at the RLC layer while the information is sent accommodate users that are in motion, or those located near the cell
over the network interfaces [4]. border. Traditionally, CDMA systems have utilized a soft handoff pro-
By decreasing the delay associated with retransmissions, protocols such cedure to seamlessly switch between base stations. The scheduled
as TCP/IP can be easily implemented into higher layers of the system. nature of the HS-DSCH, however, makes it impossible to use a soft
This will allow support for a variety of applications, such as the inter- handoff mechanism with HSDPA. Instead, a fast, hard handoff algo-
IEEE Canadian Review - Spring / Printemps 2004 23
European Wireless. February 2002. pp 451 - 457.
[5]. A. Das, F Khan, A. Sampath, and H. Jung Su. “Design and Perfor-
mance of Downlink Shared Control Channel for HSDPA”, Proc. of
13th IEEE PIMRC, Sept. 2002. pp 1088 - 1091.
[6]. M. Dottling, J Michel and B. Raaf. “Hybrid ARQ and Adaptive
Modulation and Coding Schemes for High Speed Downlink Packet
Access”. Proc. of 13th IEEE PIMRC, Sept. 2002. pp 1073 - 1077.
[7]. Y. Ofuji, A. Morimoto, S. Abeta, and M. Sawahashi. “Compari-
son of Packet Scheduling Algorithms Focusing on User
Throughput in High Speed Downlink Packet Access”, The 13th
annual symposium on Personal, Indoor and Mobile Radio Commu-
nications, 2002. Sept. 2002. pp 1462 - 1466
9.0 Acknowledgement
This work was supported in part by the NSERC USRA and NSERC
Figure 2: Scheduling Schemes and Operations (Adapted from [7] Discovery Grant.
10.0 Acronyms
3G - Third Generation
rithm has been proposed to quickly switch between base stations. 3GPP - Third Generation Partnership Project
This technology, named Fast Cell Selection (FCS), operates by monitor- ACK - Acknowledgement
ing the SIR level of all the base stations in the UE’s active set. When a AMC - Adaptive Modulation and Coding
different base station in this set can provide a higher SIR (higher CDMA - Code Division Multiple Access
CPICH power), the user is transferred to the respective base station.
CPICH - Common Pilot Channel
Both Internode and Intranode handoffs can be supported with FCS. DPCCH - Dedicated Physical Control Channel
When utilizing Intranode B FCS, the fast handoff mechanism is FCS - Fast Cell Selection
restricted to base stations that are a subset of the current Node B. If a
sector outside of the current Node B can provide better channel condi- HARQ - Hybrid Automatic Repeat Request
tions, the RNC (Radio Network Controller) is responsible for the HSDPA - High Speed Downlink Packet Access
handoff procedure. This increases the handoff delay considerably, and HS-DSCH - High Speed Downlink Shared Channel
may result in QoS requirements that cannot be satisfied. The alternative HS-SCCH - High Speed Shared Control Channel
to this restriction is to utilize Internode handoffs. Internode B FCS MAC - Media Access Control
allows the UE to quickly change base stations regardless of the respec- MCS - Modulation and Coding Scheme
tive Node B. This decreases the delay in the handoff procedure;
however, complexity is increased considerably. When a handoff occurs, NACK - Negative Acknowledgement
the UE must use over-the-air signaling to quickly signal its status to the PSK - Phase Shifting Key
new Node B. QAM - Quadrature Amplitude Modulation
RLC - Radio Link Control
This technology has proven to be beneficial in providing seamless cov-
erage to mobile users. It also increases performance to users located RNC - Radio Network Controller
near the cell border, especially when combined with the RR scheduler. TTI - Transmission Time Interval
Less benefit is realized, however, when implemented with a Maximum UE - User Equipment
C/I scheduler because users located near the cell border are rarely WCDMA - Wideband Code Division Multiple Access
serviced.
7.0 Conclusion
The future of cellular communication is an ever expanding market-
About the Authors
place. As applications increase in complexity, the resources that support Ryan Caldwell is in his fourth year of a Bache-
these applications will need to evolve as well. The 3GPP’s evolution in lor’s degree in El ectrical Engineering. He
high speed data transmission will definitely be a good candidate in pro- attends Ryerson University in Toronto, ON. His
viding users with the increased data rates and minimal delay necessary
to support these applications. Only time will show the actual benefit of research interests include communication sys-
HSDPA, but at the present time, it appears to be a viable protocol for tems, signal processing and comp uter
the future of cellular communications. networking. He is also the fourth year represen-
tative of Ryerson’s IEEE student branch.
8.0 Read more about it:
[1]. 3GPP RAN TS 25.848 v4.0.0 (2001 - 03). “Physical Layer Aspects Alagan Anpalagan is an Asst. Professor at
of UTRA High Speed Downlink Packet Access (Release 4)”, April Ryerson University where he co-founded WIN-
5, 2001. CORE laboratory and dir ects the Wireless
Access and Netwo rking R&D group. His
[2]. 3GPP RAN TS 25.950 V4.0.0 (2001-03), “High Speed Downlink research interests are in QoS-aware radio
Packet Access (Rel. 4)”, Mar. 27, 2001. resource management; joint study of wireless
[3]. T.E. Kolding, F. Frederiksen, and P.E. Mogensen. “Performance physical and link layer characteristics; and cel-
Aspects of WCDMA Systems with High Speed Downlink Packet lular/multihop radio communication for 4G
Access (HSDPA)”, Proc. of 56th IEEE VTC, September 2002. pp systems. He has published more than 30 arti-
477 - 481 cles in his research area. He currently serves as Technical Program
Co-Chair, IEEE CCECE’04; Chair, Communications Chapter -
[4]. M. Chatterjee, G.D. Mandyam, S.K. Das. “Fast ARQ in High IEEE Toronto Section; and Counselor - IEEE Ryerson. He holds a
Speed Downlink Packet Access for WCDMA Systems”, Proc. of Ph.D. degree in Electrical Engineering from University of Toronto.
24 IEEE Canadian Review - Spring / Printemps 2004
View publication stats
