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

net/publication/220250186

Evolution of the Power Control Techniques for DS-CDMA Toward 3G Wireless

Communication Systems.

Article  in  IEEE Communications Surveys & Tutorials · January 2000

DOI: 10.1109/COMST.2000.5340803 · Source: DBLP

CITATIONS READS

45 373

2 authors, including:

Miroslav Dukic
University of Belgrade
17 PUBLICATIONS   182 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

plasma surface treatment View project

All content following this page was uploaded by Miroslav Dukic on 19 May 2014.

The user has requested enhancement of the downloaded file.

 IEEE
COMMUNICATIONS

SURVEYS
The Electronic Magazine of
Original Peer-Reviewed Survey Articles

EVOLUTION OF THE
POWER CONTROL TECHNIQUES FOR
DS-CDMA TOWARD
3G WIRELESS COMMUNICATION SYSTEMS
DEJAN M. NOVAKOVIC AND MIROSLAV L. DUKIC, BELGRADE, YUGOSLAVIA,
FACULTY OF ELECTRICAL ENGINEERING

ABSTRACT
In this article we present a survey of the power control techniques for modern wireless
DS-CDMA communication systems. Power control is the single most important system
requirement for DS-CDMA systems. Well-defined power control is essential for proper
functioning of the DS-CDMA system. In the absence of power control the effect of near/far
phenomena is dominant, and the capacity of the DS-CDMA mobile system is very low, even
lower than that of mobile systems based on FDMA. However, when the power control in
DS-CDMA systems exists, it allows users to share resources of the system equally between
themselves. Also, with a proper power control it is possible to lower total transmitting power
of the mobiles and prolong the battery life.

I n recent years the cellular communications market has

exploded. The main goal of cellular communications sys-
tems that consist of sets of mobiles (MS), base stations
(BS), and mobile switching centers (MSC), as depicted in
Fig. 1, is to enable communication services irrespective of
systems. In order to meet the growing demandss of sub-
scribers for different kinds of services, such as conferencing,
multimedia, data base access, Internet, etc., it is necessary to
have higher data rates up to 2Mb/s and more stringent Quali-
ty of Service (QoS) requirements. A milestone was reached in
time and location. It is already possible to imagine a sub- January 1998 when the basic technology for the Universal
scriber who, in the near future, makes video calls to friends Mobile Telecommunication System (UMTS) Terrestrial
and colleagues from a mobile terminal while simultaneously Radio Access (UTRA) system was selected by the European
accessing a remote data base from the same terminal, or Telecommunications Standards Institute (ETSI) [2]. This deci-
receiving e-mails or phone calls. But how did everything sion contained the following key elements:
begin? • For the paired bands 1920-1980 MHz and 2110-2170
A brief history of cellular communication systems will be MHz, Wideband CDMA (WCDMA) shall be used in
presented here [1]. In 1979 the first analog cellular communi- Frequency-Division Duplex (FDD) operation.
cation system, the Nippon Telephone and Telegraph (NTT) • For the unpaired bands of total 35 MHz, Time-Division
system, became operational. Subsequently, the Nordic Mobile CDMA (TD-CDMA) shall be used in Time-Division
Telephone (NMT) system and the Advanced Mobile Phone Duplex (TDD) operation.
Service (AMPS) were introduced in 1981 and 1983, respec- • Parameters shall be chosen to facilitate easy implementa-
tively. These systems were based on Frequency Division Mul- tion of FDD/TDD dual-mode terminals.
tiple Access (FDMA) and are known as first generation In addition, Japan [3], North America [4], and South
cellular systems. Korea [5, 6] will most likely base their systems on the same
The first digital cellular standard, the Global Standard for technology.
Mobile Communications (GSM), based on Time Division An air interface based on wideband DS-CDMA provides
Multiple Access (TDMA), was deployed in 1992 in Europe. the opportunity to design a system with properties fulfilling
Two digital standards developed in the United States are the 3G requirements. The key properties emphasized in
IS-54, based on TDMA, and IS-95, based on narrowband WCDMA are:
DS-CDMA. The first digital cellular system in Japan was Per- • Improved performance over the second-generation sys-
sonal Digital Cellular (PDC), introduced in 1994. All of these tems, including capacity, coverage, and enabling migra-
digital cellular systems are called second generation cellular tion from second-generation deployment.

2 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

• A high degree of service flexibility, including PSTN
support of a wide range of services with maxi-
mum bit rates above 2 Mb/s and the possibili- MSC
ty for multiple parallel services on a single
connection. Also, a fast and efficient packet-
access scheme is specified.
• A high degree of operator flexibility, including
support of asynchronous inter-base station
operation; efficient support of different BS
deployment scenarios, including hierarchical
cell structure and hot-spot scenarios; support MS
BS
of evolutionary technologies such as adaptive MS
MS BS
antenna arrays; and multi-user detection. MS
Also specified is a TDD mode designed for
efficient operation in uncoordinated environ- MS
ments. MS
Since it is necessary to have higher data rates
and more stringent QoS requirements, new trans- ■ FIGURE 1. The cellular communications system.
mission technologies and improved radio resource
management, including base station assignment,
channel assignment, especially power control, and handoff, compromise between easier implementation and perfor-
are required for cellular communication systems. mances. Because of that different, more complex, techniques
Power control is one of the most important system require- for both reverse and forward link power control are presented
ment, and it is analyzed for cellular networks based on FDMA that improve system performance, such as the integration of
and TDMA [1, 7–12], and for DS-CDMA cellular networks, power control and base station assignment and/or advanced
[13–39]. In most modern systems, both base stations and receiver design. Joint power control, multi-user detection, and
mobiles have the capability of real-time (dynamic) adjustment beamforming in the reverse link and transmit beamforming
of their transmit powers. and power control for the forward link, possibly integrated
The necessity for power control in FDMA/TDMA-based with base station assignment, have the greatest potential for
cellular networks stems from the requirement for co-channel improvement in system performance.
interference management. This type of interference is caused The article is organized as follows. Power control tech-
by frequency reuse due to limited available frequency spec- niques are classified according to different criteria. Different
trum. By a proper power adjustment, the harmful effects of power control techniques for DS-CDMA reverse link are
co-channel interference can be reduced. This allows a more summarized. Power control for DS-CDMA forward link is
“dense” reuse of resources and thus higher capacity. presented, as well as power control in 3G wireless communi-
In DS-CDMA systems, power control allows users to share cation system proposals. A summary and concluding remarks
resources of the system equally between themselves. In the are then given.
event that power control is not used, all mobiles transmit sig-
nals toward the base station with the same power, without
taking into account the fading and the distance from the base CLASSIFICATION OF
station. Mobiles that are closer to the base station will cause
significant interference to the mobiles that are further from
P OWER CONTROL TECHNIQUES
the base station because of non-zero cross-correlation between Several methods and strategies to control the power in cellu-
signature sequences assigned to users. This effect is known as lar radio systems were proposed. When considering the power
near/far effect. control in real systems, the following aspects are interesting:
Because of that, well-defined power control is essential for • Quality measure: Speech quality is a very subjective quan-
proper functioning of the DS-CDMA system. In the absence tity. People have argued that Signal-to-Interference
of power control the capacity of the DS-CDMA mobile sys- Ratio (SIR) is an adequate objective measure, and it has
tem is very low, even lower than that of mobile systems based been used extensively in previous works, even though it is
on FDMA. far from ideal. For example, in the case of data transmis-
One of the reasons for the use of power control both in sion, bit error rate (BER) requirements can be very
FDMA/TDMA and in DS-CDMA networks is to prolong bat- stringent and SIR might not be an adequate measure. If
tery life by using a minimum of transmitter power to achieve the signal and interference powers are constant BER and
the required transmission quality. SIR contain equivalent information regarding quality.
According to the above mentioned facts, for proper opera- But in real systems the SIR is time variant and thus the
tion of a modern high-capacity cellular radio system, power average SIR will not correspond to the average BER. In
control is an essential feature. Among others, there are two this case BER is a better quality measure.
important aspects to be considered: • Available measurements: Usually the measurements are
• Extracting relevant information from the available mea- given in reports comprising a quality indicator (QI)
surements. reflecting the quality and a received signal strength indi-
• Designing an appropriate power control algorithm and cator (RSSI) reflecting the received signal strength at the
tuning the optimized parameters. receiver. These values are coarsely quantized in order to
In this article we present a survey of power control tech- use few bits.
niques for reverse and forward DS-CDMA links. Since the • Constraints: The output power levels are limited to a
IS-95 system is the first operational commercial DS-CDMA given set of values due to hardware constraints. This
system, both reverse and forward link power control for IS-95 includes quantizing and the fact that the output power
systems are presented. Power control for IS-95 systems is a has an upper and a lower limit.

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 3

 Reverse link
Forward link
in time delays in the system.
• The possible output powers of
Centralized Power control Continuous power the transmitters are constrained
Decentralized algorithms Discrete power due to physical limits and quan-
tization. Different external con-
straints such as the use of
Closed loop maximum power on specific
Strength-based
Open loop Fixed step size channels affect the output
SIR-based
Combined closed Adaptive step size power.
BER-based
and open loop
• The signals needed for control
may not be available and have
■ FIGURE 2. Classification of power control techniques.
to be estimated.
• Quality is a subjective measure,
and relevant objective measures
• Time delays: Measuring and control signaling take time, have to be employed.
which results in time delays in the network. The surrounding environment as seen by a decentralized
Power control techniques can be classified in many differ- controller, for example, is shown in Fig. 3.
ent ways, as shown in Fig. 2. One of the possible classifica- According to what is measured to determine power control
tions is: command, power control techniques can be classified into
• Power control for reverse link (from mobiles to base sta- three categories:
tions). • Strength-based
• Power control for forward link (from base stations to • SIR-based
mobiles). • BER-based
Power control for DS-CDMA reverse link is the single In strength-based schemes the strength of a signal arriving
most important system requirement because of the near/far at the base station from a mobile is measured to determine
effect. In this case, it is necessary to have a dynamic range for whether it is higher or lower than the desired strength. The
control on the order of 80dB [13]. For the forward link, no command to lower or raise the transmit power is made accord-
power control is required in a single cell system, since all sig- ingly.
nals are transmitted together and hence vary together. How- In SIR-based schemes the measured quantity is the SIR
ever in multiple cell systems, interference from neighboring where interference consists of channel noise and multi-user
cell sites fades independently from the given cell site and interference. Strength-based power control is easier to imple-
thereby degrades performance. Thus it is necessary to apply ment but SIR-based power control reflects better system per-
power control in this case also, to reduce intercell interfer- formance such as QoS and capacity. A serious problem
ence. associated with SIR-based power control is the potential to
Also, there is centralized and decentralized (distributed) get positive feedback to endanger the stability of the system.
power control. A centralized controller has all information Positive feedback arises in a situation when one mobile under
about the established connections and channel gains, and con- instructions from the base station has to raise its transmit
trols all the power levels in the network or part of the net- power in order to deliver a desirable SIR to the base station,
work. Centralized power control requires extensive control but the increase in its power also results in an increase in
signaling in the network and cannot be applied in practice. It interference to other mobiles so that these other mobiles are
can be used to give bounds on the performance of the dis- then forced to also increase their power, etc. In the case of N
tributed algorithms. mobiles in the system, this becomes a typical non-cooperative
A decentralized controller controls only the power of one N-person game problem [14].
single transmitter, and the algorithm depends only on local In BER-based power control, BER is defined as an aver-
information, such as measured the SIR or channel gain of the age number of erroneous bits compared to the original
specific user. These algorithms perform well in rather ideal sequence of bits. If the signal and interference powers are
cases, but in real systems there are a number of undesired constant, the BER will be a function of the SIR, and in this
effects, such as: case the QoS is equivalent. However, in reality the SIR is
• Measuring and control signaling take time, which results time-variant and thus the average SIR will not correspond to
the average BER. In this case the BER
is a better quality measure. Since the
channel coding is implemented in
Quality Environment every practical system, power control
requirements
can be based on the average number
of erroneous frames as well.
Power Also, power control techniques can
Constraints q-n be classified as follows:
regulator
Cellular • Closed-loop power control.
network • Open-loop power control.
Measurement data q-m • A combined technique consist-
ing of closed-loop and open-loop
power control.
Closed-loop power control is feasi-
ble in a terrestrial cellular environ-
■ FIGURE 3. The network block incorporates the effects caused by the radio channel, such ment. However, in mobile
as power gain, noise, and interfering transmitters, while q-n and q-m represent appropriate communications systems using multi-
delays. ple low earth orbital satellites, the

4 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

fades occur too rapidly for the closed-loop power control to the transmission of consecutive packets, it is very difficult to
track, due to the large round trip propagation delay. In this employ closed-loop power control. The validity of feedback
case, the solution is open-loop power control where the information for closed-loop power control decreases as the
mobile user estimates the channel state on the forward link, interval between packets increases.
and this estimate is used as a measure of the channel state on
the reverse link. These techniques can compensate for path
loss and large-scale variations such as shadowing, but it is not POWER CONTROL TECHNIQUES FOR
possible to compensate multipath fading because reverse and DS-CDMA REVERSE LINK
forward links are not correlated. It has been shown that
capacity degrades by 5 percent for a 1dB open-loop power POWER CONTROL FOR THE IS-95 DS-CDMA SYSTEM
control error, by 25 percent for a 2dB power control error,
and by 44 percent for a 3dB power control error [15]. The IS-95 system is the first operational commercial
The combined technique will be presented in the next sec- DS-CDMA system, and problems encountered during its
tion. operation provide valuable experience for future research
According to update strategies, power control algorithms work and development.
can be classified as follows: Power control for DS-CDMA according to the IS-95 stan-
• Those where the transmit power step size is fixed (fixed dard [17] consists of reverse link open-loop power control,
step size algorithm). reverse link closed-loop power control, and forward link
• Those where the transmit power step size is made adap- power control, as shown in Fig. 4. Reverse link open-loop
tive to the channel variation. power control is primarily a function of the mobile stations.
A specific example of the adaptive step size approach is The base stations take an active role in the reverse link
the inverse update algorithm, which increases or decreases the closed-loop power control and the forward link power control.
mobile users’ transmit power by the actual difference between Power control for the reverse link is a combined technique
the received signal power and the desired received signal consisting of closed-loop and open-loop power controls. Also,
power. it is a fixed step size algorithm and strength-based distributed
Power control command in fixed step size algorithms is a algorithm. The goal of open-loop power control is the estima-
simple 1-bit command. It has been shown that the inverse tion of a path loss and a loss due to shadowing between the
algorithm is superior to the fixed step size algorithm [16]. base and the mobile station. According to this process, the
However, the fixed step size algorithm is easier to implement mobiles transmit the initial power control signal.
because the inverse algorithm needs additional bandwidth on However, multipath fading in a reverse and a forward
the return channel to carry the power control step size instead DS-CDMA link is an independent process since the frequency
of the 1-bit control command as in fixed step size algorithm. separation of these links is 45MHz and it greatly exceeds the
A compromise would be to use an adaptive delta-modulation coherent bandwidth of the channel. Thus, closed-loop power
algorithm. control is used. Every cell site demodulator measures the
Power control can be also classified as follows: received signal-to-noise ratio (SNR) from each mobile station.
• Power control where the transmit power level is con- The measured SNR is compared to the desired SNR for that
trolled in continuous power domain. mobile station and a power adjustment command is sent to
• Power control where the transmit power level is con- the mobile station. This power adjustment command is com-
trolled in discrete power domain. bined with the mobile station open-loop estimate to obtain
In practice transmission power in digital systems can be the final value of the mobile station transmit power. This
updated only at discrete levels. It has been shown in [9] that command has the fixed step size of 0.5dB and it is transmitted
by simple discretization of continuous power control algo- at a rate of once every 1.25ms. The base station measures the
rithm using a function ceiling (rounds toward plus infinity) signal quality (BER) and based on that determines the desired
and function floor (rounds toward minus infinity), the conver- SNR for specific mobile station. The result of the open loop
gence and uniqueness of the continuous power control algo- and the closed loop power control is presented in Fig. 5.
rithm are lost. The discrete ceiling power control algorithm In previously described power control technique, the sub-
converges in a weaker sense, i.e., into an envelope of power scribers are power controlled by the base station of their own
vectors rather than to a unique vector. The transmitter powers cell. However, the interference level from subscribers in other
under this algorithm may oscillate, which results in a poorer cells varies not only according to the attenuation in the path
link quality and in a higher outage probability. In order to to the subscriber’s cell site, but also inversely to the attenua-
alleviate these oscillations, it has been proposed in [9] to start tion from the interfering user to his own cell site, which
updating the powers according to the floor algorithm from through power control by that cell site may increase or
any initial power vector until the first entry into the envelope decrease the interference to the desired cell site. It has been
of power vectors, and then to continue with the ceiling algo- shown [13] that the maximal number of subscribers in the cell
rithm. is the highest when there are no subscribers in the neighbor-
Power control can be applied in circuit-switched and pack- ing cells. As the number of subscribers in the neighboring
et-switched modes of operation. Among various multimedia cells increases the maximal number of subscribers in the cell
services circuit switching is suitable for delay-sensitive and decreases. For example, when there is a maximal number of
long-holding-time media, such as voice, video and large-size users in the neighboring cells the reverse link can support 108
file transfer. Packet switching is suitable for delay-less-sensi- users/cell, with 10 – 3 bit error rates better than 99 percent of
tive and short-holding-time media. Packet communications the time. This number becomes 132 users/cell if the neighbor-
are provided to bursty information sources, which are charac- ing cells are kept to half of this loading.
terized as on-off processes, and transmission is discontinued In order to lower processing delay and to save bandwidth
at the end of the data burst and no information is transmitted in the forward link, command bits for power control from the
during the unpredictable off interval. In some cases, synchro- base to the mobile station are not coded and they are suscep-
nization information may be transmitted during these off tible to errors. It has been shown in [18] that every 1dB power
intervals. Because it is possible to have large intervals between control error standard deviation increase roughly translates

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 5

 Forward
power
control XMIT power
control

Deinterliver
Selector AGC AMP and decoder
Transmit Data
data
Variable
XMIT HPA Digital
Error rate demod
determination Open Error rate
Eb /N0 Eb /N0 loop determination
set sensor control
Closed
loop
control
Digital
demod Data
source
Data
Deinterliver XMIT data
and decoder processing
Cell site processing Subscriber unit processing

■ FIGURE 4. Block diagram of power control for IS-95 DS-CDMA [17].

into a loss in capacity of 10 users. LINEAR PREDICTION-BASED CLOSED-LOOP POWER CONTROL

The rate of power control adjustment command transmis-
sion must be high enough to permit tracking of Rayleigh fad- One of the ways to enhance a closed-loop power control is to
ing in the reverse link. It is important that the latency in use a linear prediction of power control bits at the mobile
determining the power control signal and the transmission [19]. The linear prediction should mitigate the impact of the
process be kept small so that the channel conditions will not errors in the transmission of the power control bits and the
change significantly before the control bit can be received and effect of the delay from when the power control bit is generat-
acted upon. It has been shown in [16] that in a multi-cell sys- ed until the transmitted power is adjusted accordingly. The
tem under flat fading conditions, increasing the update rate cost of that is a more complex receiver structure.
from 800Hz to 2KHz, results in a capacity improvement on The receiver consists of a linear prediction filter that has
the order of 50 percent. In the case of multipath fading, the accumulated received power control bits at its input. Its out-
capacity improvement is only 10 percent, and it can be con- put is the prediction to the accumulation of the power control
cluded that increasing the update rate results in diminishing bits and power control is executed according to the predictor
capacity improvements, as the channel becomes more and output. Since the performance of the predictor depends on
more frequency-selective. the autocorrelation of its input signal, it is important that the
power control bit stream is a correlated process. After a sim-
plified analysis it has been shown in [19] that predictor input
has the same autocorrelation properties as the signal corre-
sponding to fading channel variations,
whose bandwidth is defined by the
Doppler spread. Simulation results in
[19] have shown that the predictor of the
Rayleigh fading order 60 and tap gain adaptation step
Average 0.002 compared to the conventional IS-
path 95 power control yields improvement in
loss
Power

overall system performance that is

approximately equivalent to increasing
R-1 the SNR by 1dB (from 6dB to 7dB).
Distance
STRENGTH AND
Power received SIR-COMBINED POWER CONTROL
by subscriber
Power

One of the ways to alleviate the possibil-

ity of positive feedback in power control
Power

Received power based on SIR measurement is to use

at cell site
Distance strength-based and SIR-based power
Desired average
Power transmitted received power control schemes in a combined manner
by mobile [20]. Both the received power and the
Distance
received SIR are measured at base sta-
Power received
tions. If both quantities are lower or
by cell higher than the threshold, it is clear that
the associated mobile should be instruct-
■ FIGURE 5. Effects of power control for reverse link [17]. ed to power up in the first case or to

6 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

 650
power down in the second case. In the case that SIR is in sat- 600
isfactory condition and that the power is below the threshold,

Total transmitted powers

the mobile does not need to take any action because SIR 550
reflects better system performance such as QoS and capacity, 500
as noted earlier. In the case that received power is in satisfac-
tory condition and SIR is below the threshold, in principle the 450
mobile needs to raise its transmitting power, but in a more 400
careful manner because of potential positive feedback since
the power of the signal already exceeds its threshold. One way 350
to solve this problem is to take a number of most recent states 300 Uplink: fixed assignment
Downlink: fixed assignment
when the power is satisfactory and SIR is not. The higher that Uplink: uplink assignment
number, the less amount of power the mobile should be asked 250 Downlink: uplink assignment
Downlink: downlink assignment
to increase. It has been shown in [20] that this combined 200
power control technique yields better performances than 0 10 20 30 40 50 60 70
power control based on SIR measurements only. Number of additional local users

UTILITY-BASED POWER CONTROL ■ FIGURE 6. The total transmitted powers vs. the number of users
in a FDD network [24].
Another way to alleviate positive feedback or to solve the
problem of the non-cooperative N-person game is to define
the utility function u j ( p j , SIR j ) for every user as a function where
of its transmitter power reflecting users’ preference for power
saving and SIR reflecting users’ preference for QoS. It is I (kn−1) ( p) = ∑ pu( n−1)Γ( u, k ) + σ2 . (4)
assumed that for the j th user, given a SIR j , the utility function
u≠ i
is monotonically decreasing concave upward as a function of (n-1)
its transmitter power, and for given its transmitter power p j , Ik
is the total received power of all other users at cell
the utility function is a monotonically increasing concave site k at step (n – 1) of the algorithm; user i is in soft han-
upward as a function of the SIR j . The concavity in the first dover with a set of base stations that are denoted by Di; Γ(u,
assumption comes from the fact that the marginal of a user’s k) is the fading from user u to base station k; pu is the trans-
preference will be steeper as larger power amount is needed mitter power of user u; there are M users and K base stations
to maintain a fixed SIR, and that the second assumption is in the system. It is assumed that user i has SIR requirement
due to the fact that users will be indifferent to further QoS SIRi ≥ bi. L is the length of signature sequences assigned to
improvement once it has reached the target SIR. It is a goal the user, and σ 2 is the variance of the Gaussian noise.
to find for every user j the optimal transmitter power such As can be seen from Equations (2) and (3), this algorithm
that it will be most satisfied. This is a typical individual opti- makes it possible to solve the problem of power control and
mization problem, formulated as follows: base station assignment in a decentralized way. At each step
of the algorithm, user i needs only to know its own desired βi,
max u j ( p j , SIR j ). (1) its own path gains to the base stations to which it is in soft
pj
handover, and the total received interference at each of these
The solution of this problem is presented in [14], where it base stations. The user does not need to know any informa-
has been shown that there exists a unique Nash equilibrium. tion about other base stations in the network, and it is unnec-
Also, an economic model of the power control for data trans- essary to know anything about the transmit power levels or
mission is presented in [21], where it has been shown that path gains of any other user.
unique equilibrium is not Pareto-efficient. In order to find a Base station selection can be explained in the following
more Pareto-efficient solution, a pricing function is intro- way. At each step of the algorithm, the user listens to broad-
duced in [21] that yields favorable utility solutions at reduced casted interference levels from every base station in its soft
transmitter powers. handover set. The user then computes the transmit power lev-
els that it would need if it were to send them to every one of
INTEGRATED POWER CONTROL AND these base stations, and then transmits them with the mini-
BASE STATION ASSIGNMENT mum computed power. Apart from this power control mecha-
nism, base station selection plays no other role. Every base
An entirely different approach to power control from those station in the soft handover set can still demodulates the
presented earlier in the text is the integration of power con- user’s signal and send it to the switching center. The switching
trol and base station assignment [22–24]. The algorithm is center makes the base station selection, but the decision can
defined as follows. Updates of transmission powers are be based on frame error rates (FER), rather than on explicit
defined by knowledge of the selection that the mobile itself made when it
1  βi transmitted the signal.
p(i n) = min  Ik ( n−1) , (2)
The performance improvements of the integrated power
k ∈D  L
i  Γ( i , k ) control and base station assignment related to fixed assign-
i = 1, 2,... M ment in the case of reverse link are presented in Fig. 6, where
it can be seen that this algorithm has doubled the number of
and updates of base station assignments is additional local users in the congested cell. The simulation
scenario consisted of 36 hexagonal cells and an initial number
1  βi of 300 users, while additional users are added only to one cell.
b(i n) = arg min I k( n−1) , (3)
The link gain model included only path loss. This algorithm
k ∈Di  L  Γ( i , k )
yields very good performances, but the following problems
i = 1, 2,..., M exist in its practical implementation:

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 7

 Beamformer
t=nT weights
A Matched in [27]. As a performance measure of the algorithm, the
r w*0 failure rate was used. This rate corresponds to the proba-
r filter
a bility that, for a randomly given instance, not all of the
y total number of mobiles are supported with their target
Matched w*1
e filter SIR. It has been shown that compared to distributed con-
l Σ strained power control (DCPC) there is significant capaci-
e
m ty gain from TMPC and CMPA. Also, TMPC has less
e than 10 percent higher computational load than DCPC
n and it is very productive because of significant capacity
t Matched w*k
s filter gain.
Adaptive
beamformer COMBINED POWER CONTROL AND
ADVANCED RECEIVER DESIGN
Array response to the direction of interest
Since WCDMA will employ advanced receiver designs,
■ FIGURE 7. Antenna array and beamformer [28]. such as beamforming with adaptive antenna arrays and
multi-user detection, power control could be integrated
with these receivers.

• High communications overhead in determining the mini- Joint Power Control and Beamforming using Antenna
mum transmit power and selecting base stations. Arrays — Comprehensive analysis of joint power control and
• All base stations must also have large storage capacity to beamforming using antenna arrays is given in [28].
store all the information broadcast from all base stations An adaptive antenna array is depicted in Fig. 7. It consists
in the system. of a set of antennas designed to receive signals from some
The computational requirement for a mobile device can be specific directions and to attenuate signals from other direc-
high since it has to periodically calculate transmit power for tions of no interest. The outputs of array elements are weight-
all neighboring base stations in order to select the minimum. ed and added by a beamformer to produce a directed main
beam.
Pilot Power Controlled Load Sharing — In order to solve The algorithm for joint power control and beamforming
these problems in practical implementation, a power control using antenna arrays is a two-stage algorithm. In the first
algorithm based on the power control of the pilot signal is stage, beamforming weight vectors w i are computed at each
proposed to effectively adjust the size of a cellular area to receiver i such that the co-channel interference is minimized
maximize overall system capacity [26]. According to the basic under the constraint of maintaining constant gain for the
idea of this algorithm, when the SIR decreases below the direction of interest, and for fixed powers from the previous
threshold, the base station starts directing some subscribers to iteration step. Afterward, in the second stage of the algorithm,
less loaded surrounding cells by lowering the pilot signal the updated power vector is calculated. These operations are
power, thus shrinking the cell size and increasing the total done iteratively until the vector of transmitted powers and
number of subscribers that the network system can handle. beamforming weight vectors converge to the joint optimal vec-
Simulation results have shown [26] that under localized tor. The proposed algorithm is decentralized and amenable to
heavy load or asymmetric load this algorithm performs very a distributed implementation.
well. However, under uniformly heavy load or symmetric load Also, joint power control, base station assignment, and
this algorithm performs less ideally. In extreme cases, the QoS beamforming are considered in [28], which converges to the
could be even worse than without this algorithm, since uni- optimal power allocation starting from any initial power vec-
formly heavy load can cause all base stations to reduce their tor. The algorithm is given as follows:
pilot signal power. This can cause a mobile station in the mid- • Each mobile can be assigned to a set of base stations,
dle of two base stations to experience service difficulties or denoted by D i for the i-th mobile.
even be dropped in fading conditions. • At each iteration all base stations in the set will perform
beamforming and the mobile transmitted power for the
Two-Mode Algorithm for Power Control and Base Sta- next iteration is calculated.
tion Assignment — Another way to alleviate the heavy sig- • The base station assignment is performed by comparing
naling and computational burdens of the algorithms proposed the power requirements for different base station assign-
in [22–24] is to use a two-mode power control (TMPC) algo- ments.
rithm for a combined power control and base station assign- • The base station with the least required power will be
ment [27]. The algorithm has two modes: normal and chosen for the mobile.
congestion. In normal mode the power control procedure is This can be expressed mathematically in the following way:
executed without base station assignment. In congestion mode
TMPC updates both power levels and base station assign-  2 2 2 2
ments as a technique considered in [22] with constrained max-  ∑ Γ( u, j ) wijH auj siH su pu( n) + σ 2 wij si 
 
imum power level, and it is called a constraint minimum p(i n+1) = min β i u≠1 ;
wij , j ∈Di  2 2

power assignment (CMPA) problem. Therefore, TMPC oscil- Γ ( i , j ) s H
s w H
a
lates between the normal and the congestion modes in each of  i i ij ij 
 
the power update iterations. A necessary condition for mode i = 1, 2,..., M
change is as follows: if the received SIR of a mobile station at (5)
subject to
the corresponding base station becomes less than or equal to
its target SIR, than the power control mode of the mobile sta- wijH aij = 1
tion can be changed to the congestion mode from the normal
mode. The formal description of the TMPC algorithm is given where wij is the receive beamforming weight vector for user

8 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

 16 Power control algorithm
Power control and base
i at base station j; auj is the array response of the j-th anten- 14 assignment
na array to the u-th source; and su is the spreading vector of Power control with antenna

Sum of mobile powers

array
user u and the other parameters as defined in an earlier 12
section.
It is worth noting that the beamformings at the base sta- 10
tions are done independently, without the knowledge of other
channel responses. The performance of the joint power con- 8
trol, base station assignment, and beamforming is presented in
6
Fig. 8, as the total mobile power vs. the number of users. It
can be seen that using omnidirectional antennas and the 4
power control with fixed base station assignment, the maxi-
mum capacity is 660 users. If the joint base station assignment 2
and power control algorithms are used, the capacity is 800
users. In the case of joint power control, base station assign- 0
0 500 1000 1500 2000 2500 3000
ment, and beamforming, the capacity is 2800 users. The simu-
lation scenario consisted of a network with 36 base stations. Number of users
The link gain model included only path loss.
■ FIGURE 8. Total mobile power vs. the number of users [28].
Integrated Power Control and MMSE Interference Sup-
pression — The integration of power control and Minimum In the second stage of the algorithm, the power vector P is
Mean Square Error (MMSE) interference suppression is pro- updated using c i and w i from the first stage of the algorithm
posed in [29]. This algorithm is a two-stage algorithm: in the following way
• First, the coefficients of the filter c i are adjusted to mini-
mize the required transmitter power p(i n+1)
2 2 2 2
pi ( n) ∑ Γ( u, i) wiH ( n + 1)au i ciH ( n + 1)su pu( n) + σ 2 wi ( n + 1) ci ( n + 1)
ci = A−i 1( p( n))si ; i = 1, 2,... M = βi u≠ i ;
1 + pi ( n) sTi A−i 1( p( n))si (6) 2 2
Γ( i, i ) ciH ( n + 1) si wiH ( n + 1)aii
(9)
Ai ( p( n)) = ∑ pu( n)Γ( u, i)su suT + σ2 I i = 1, 2,..., M
u≠ i

where Γ(u, i) is defined as gain from user u to the assigned

base station of user i. These calculation s are done iteratively, and if the SIR tar-
• In the second stage the transmitter powers are adjusted gets are feasible then starting from any initial power vector
to meet the SIR constraints for the chosen filter coeffi- and filter coefficients, the algorithm converges to the mini-
cients mum power with the best possible temporal-spatial filter in
the constrained space (CTSF), where the constraint is that the
2 2 matrix filters x i = c i w Ti are defined. Physically, it is combining
∑ Γ( u, i) cTi su p(un) + σ 2 ci the chip matched filter outputs using a linear filter at the out-
p(i n+1) = β i u≠ i ; i = 1, 2,... M . (7) put of each of the antenna followed by linear combination of
2
the resulting statistics. The performance of the proposed algo-
Γ( i, i ) cTi si
rithm is given through the comparison of the total transmis-
sion power usage for a different algorithm, depicted in Fig. 9.
Simulation results [29] have shown that this algorithm is The simulation scenario consisted of nine base stations and 60
superior in terms of the total transmitter power and conver- mobile users. Base stations were equipped with four element
gence rate when compared with the conventional power con- antenna arrays. The link gain model included only the path
trol algorithm. With this algorithm the same system loss. The following algorithms are compared:
performance is achieved with less total transmitter power, • Conventional power control (C-PC): One antenna and
increasing the capacity of the CDMA system when compared matched filter receiver in time domain.
with conventional power control. • Power control and MMSE receiver filter (MMSE-PC):
An antenna and SIR maximizing (MMSE) receiver in the
Joint Power Control, Multiuser Detection and Beamform- time domain.
ing — In order to further increase the capacity of a CDMA • Power control and beamforming (BF-PC): MMSE com-
system, joint power control, multi-user detection, and beam- bining in the spatial domain and matched filter receiver
forming are proposed [30]. This algorithm is also a two-stage in the time domain.
algorithm. In the first stage of the algorithm, coefficients c i of • Power control with constrained temporal-spatial filtering
the linear (temporal) filter and beamforming weight vectors ((c-w) L -PC): L = 5 iterations of CTSF before each
w i (spatial filter coefficients) are calculated in order to sup- power update.
press co-channel interference. First, the coefficients c i are cal- • Power control with single CTSF (c-w-PC): L = 1.
culated for fixed w i according to the MMSE principle, and • Joint unconstrained filtering in temporal and spatial
then w i are calculated for previously computed c i , i.e., at iter- domain (OTSF) where there are no constraints for
ation (n + 1) for user i matrix filter x (X-PC).
It can be seen from Fig. 9 that only power control algo-
c i (n + 1) = MMSE (w i (n))
(8) rithms with joint processing in temporal and spatial domains
w i (n + 1) = MMSE (c i (n + 1)). are feasible and that the total transmit power is less compared
to algorithms not utilizing both temporal and spatial filterings.
All these calculations are done for the fixed power vector.
All of these algorithms are highly complex for implementa-

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 9

 1015
C-PC
MMSE-PC POWER CONTROL TECHNIQUES FOR
BF-PC
c-w-PC DS-CDMA FORWARD LINK
(c-w)L-PC
Total transmit power

1010 X-PC The purpose of the forward link power control is to reduce
power for users that are either stationary, relatively close to
the base station, impacted little by multipath fading and shad-
owing effects, or experiencing minimal other cell interference.
Therefore, extra power can be given to users that are either in
105 a more difficult environment or far away from the base station
and experiencing high error rates.
A simple power control technique without taking into
account fading is considered in [33], where transmitted power
from the base to the mobile station is proportional to the n th
100
0 5 10 15 20 power of the distance between the base and the mobile station.
Iteration index It has been shown that without power control the perfor-
mances are poor. A simple analytical model is developed for
■ FIGURE 9. Performance comparison of power control tech- the neighboring cell interference experienced by users as a
niques [30]. function of distance from their home base station, without tak-
ing fading into consideration in [34]. Using this model and the
previously mentioned nth power of distance power control, the
tion and they could be termed as inverse power control algo- optimum radial distance dependent power control is designed
rithms according to classification in an earlier section. providing a uniform service to all users and approximately
The performances of these algorithms for the fixed step doubles the capacity compared to the no power control case.
size power control command are presented in Fig. 10. Joint This is a very optimistic result since fading and shadowing is
power control, multi-user detection, and beamforming has the not considered. Optimal forward link power control law in the
highest performance again, but the number of iterations is presence of fading is presented in [35], where the expression
higher. Also, the influence of power control command error is for the average BER of a mobile located at some distance
presented in Fig. 11. It can be seen that probability of outage from the base station is given analytically. This expression for
increases as the power control command error variance σ p the desired BER is the optimality criterion that is equivalent to
increases. maximization in capacity. It has been shown that the perfor-
mances with optimal power control law are improved and the
POWER CONTROL FOR number of admitted users is increased by a factor of two. How-
PACKET SWITCHED MODE DS-CDMA SYSTEMS ever, this optimal power control law is given implicitly as the
solution of an equation and is not available in closed form.
As noted earlier, because of the random access nature of Because of that, an extension of the power control law based
packet switched systems, it is very difficult to employ closed on the nth power of the distance termed as floor power con-
loop power control. Therefore, only open loop power control trol is proposed in [36-37]. It has been shown that floor power
is applicable to the packet switched DS-CDMA system. The control always achieves nearly optimal performance [37].
conventional approach to power control in this case would be
to increase the transmission power for the random access POWER CONTROL FOR IS-95 DS-CDMA SYSTEM
burst after an unsuccessful access attempt. Since open loop
power control compensates only path loss and shadow fading, Besides power control for reverse link, IS-95 also includes for-
the received power of packet switched users is fluctuated by ward link power control [17], as shown in Fig. 4. This power
the multipath fading. control is called forward link slow power control [38]. In order
As the variation of a received power degrades the perfor- to execute forward link power control, the base station peri-
mance of a packet success rate, open loop power control with odically reduces the power transmitted to the mobile station.
power margin for packet switched users is introduced in [32]. This process continues until the mobile station requests addi-
The power margin raises the transmission power for the pack-
et switched users by a certain amount and they use higher
transmit power than they would use in the case of ordinary
104
open loop power control. For circuit switched users in the MMSE-PC
Total transmitted power [W]

same system both open and closed loop power control are BF-PC
used. Since DS-CDMA systems are interference limited, when c-w-PC
103 MMSE-PC, ∆p=1.5dB
an unnecessary higher power margin is employed the average
BF-PC, ∆p=1.5dB
message delay is reduced due to higher packet access rate, but c-w-PC, ∆p=1.5dB
interference to circuit switched users is increased and thereby 102
the capacity of the system for the circuit switched users would
be reduced. Because of that, the power margin is optimized as
a trade off between the reduction of the average message 101
delay of packet switched users and increase of interference to
the circuit switched users.
100
It has been show that open loop power control with a 5dB
margin significantly improves message delay performance for 0 10 20 30 40 50 60 70
the packet switched users without decreasing the capacity of Number of iteration
the system for the circuit switched users.
■ FIGURE 10. Performance comparison of inverse and fixed step
size power control techniques [31].

10 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

 0,4
c-w-PC, σp=0
tional power due to higher measured FER. The base station c-w-PC, σp=0.25
c-w-PC, σp=0.5
receives the power adjustment requests from each mobile sta- 0,3 c-w-PC, σp=0.75
tion and responds by adjusting the power allocated to the
transmitted signal by a predetermined amount. The adjust-
ment is small, approximately 0.5dB. The rate of change of

Poutage
0,2
power is slower than that used for the reverse link, either
once per vocoder frame, or nominally once per 15 – 20ms. The
dynamic range of the adjustment is also limited approximately
to a ±6dB range around the nominal power. It has been 0,1
shown in [13] that for a multi-cell layout this power control
algorithm can support more than 114 users/cell with 10 – 3 bit
error rates better than 99 percent of the time. This result cor- 0,0
responds to the total spread bandwidth W = 1.25MHz, the bit
rate R = 8kb/s, voice activity factor 3/8, and log-normal shad- 7 8 9 10 11 12 13
owing. Also, it is assumed that 20 percent of every base sta- Target SIR [dB]
tion’s power is devoted to the pilot signal.
■ FIGURE 11. The influence of power control command error on
INTEGRATED POWER CONTROL AND the performance of fixed step size joint power control, multi-user
BASE STATION ASSIGNMENT detection, and beamforming [31].

As in the reverse link, the joint power control and base station
assignment are proposed for the forward link [24]. The algo- fixed assignment in a TDD CDMA network are depicted in
rithm is given as follows: Fig. 12; it can be seen that capacity is almost doubled. In the
• First, the reverse link selects the base station and power case of FDD CDMA networks with a 10 percent difference
vector according to between the reverse and the forward link frequencies, the per-
formance improvement can be seen in Fig. 6. The simulation
  scenario was the same as described earlier.
 ∑ Γ( u, j ) pu( n) + σ2 
bi( n) = arg min  u≠ i ; i = 1, 2,..., M
j ∈Di  i
β (10) TRANSMIT BEAMFORMING AND POWER CONTROL
Γ( i , j ) 
 
Joint transmit beamforming and power control for forward
link are proposed in [39], where it is assumed that only base
∑ Γ( u, bi( n) ) p(un) + σ2 stations use antenna arrays. The block diagram of a transmit
beamforming system is shown in Fig. 13.
p(i n+1) = β i u≠ i ; i = 1, 2,..., M (11)
Γ  i, bi( n) 
A base station may transmit to more than one mobile with
  different beamforming weight vectors w i. In this algorithm,
• Afterward, power control iteration for forward link is run the virtual reverse link is constructed. In the first step of the
using the same assignment, algorithm, the beamforming weight vector w i is computed in
order to minimize the total power used in the network and
∑ Γ( i, bu( n) ) ˜pu( n) + σ2 with the constraint that the desired SIR at each mobile is
˜p(i n+1) = βi u≠ i ; i = 1, 2,..., M (12)
higher or equal to the target value. Afterward, the power
Γ  i, bi( n) 
update vector for the virtual reverse link and the power
  update vector for the forward link are computed using the
It has been shown that this algorithm converges to the beamforming weight vector w i computed for the virtual
optimal base station assignment and power allocation for the reverse link. Mathematically it can be expressed as:
reverse link, which minimizes each mobile power, and con- • Diversity combining for virtual reverse link
verges to a feasible base station assignment and power alloca-
tion for the forward link. In the foregoing it has been assumed  
that the link gain for the reverse and the forward link are the  pi( n) wiH H ii H iiH wi 
same. This assumption is valid in TDD networks or in FDD w(i n) = arg min  ; i = 1, 2,..., M
networks where the reverse and forward frequencies are close. wi  ∑ pu( n) wiH H iu H iuH wi + wiH wi  (13)
 u≠ i 
In a FDD network where the reverse and forward link gains
are different, the algorithm updates the forward link powers
where all link gains are replaced by the forward link gains, where H i u is the channel vector from the antenna arrays of
i.e., where there are virtual reverse links with the same link user i to the u-th mobile user.
gains as in the forward link. This algorithm converges to a • Virtual reverse link power update
feasible base station assignment that is different from that for
the reverse link. Since the virtual reverse link gains are not ∑ p(un)wiH H ui H uiH wi + wiH wi
related to any receiver, base station assignment and virtual p(i n+1) = βi u≠ i ; i = 1, 2,..., M (14)
reverse link power vector updates must be implemented in the pi( n) wiH H ii H iiH wi
central unit. The forward link power update can be imple-
mented locally by measuring the total interference at every • Forward link power update
mobile station. In networks where the gains in the reverse and
the forward link are the same, both reverse and forward link ∑ ˜p(un)wuH H iu H iuH wu + σ˜ 2
power allocation and the base station assignment can be ˜p(i n+1) = βi u≠ i ; i = 1, 2,..., M (15)
decentralized. ˜pi( n) wiH H ii H iiH wi
The performances of the proposed algorithm related to the

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 11

 750

700 power control is considered in the 3G wireless proposals with

other system requirements.
Total transmitted powers

650
In all the system proposals the power control is specified
600 for the FDD mode of operation. Power control is also speci-
550
fied for the TDD mode of operation in all system proposals
except in S. Korea TTA1 [5] and TTA2 [6], proposals that are
500 not designed for the TDD mode of operation. The general
property of power control for FDD mode is that power control
450
consists of a fast reverse and fast forward link power control.
400 The power control for TDD mode is slower than that for FDD
Fixed assignment mode. In FDD mode power control is SIR-based in all of the
350
With base assignment proposals except on TIA TR-45.5 proposal CDMA2000, [4]
300 and S. Korea TTA1, where power control is strength-based.
0 10 20 30 40 50 60 70 We will now present power control techniques in the
Number of additional local users UTRA system proposed by ETSI [2], and in the CDMA2000
system.
■ FIGURE 12. The total transmitted powers vs. the number of
users in a TDD network [24]. POWER CONTROL FOR THE UTRA SYSTEM

Since the UTRA proposal contains both the FDD and TDD
where σ ~
2 is the variance of the noise at the i-th mobile receiv- modes of operation, power control for both cases will be pre-
er. sented.
This algorithm requires the full knowledge of the channel In FDD mode the transmitter uses fast closed loop SIR-
and array responses for the entire network. This requires based power control and slow quality-based power control on
channel measurements at the mobile and a feedback channel both the reverse and the forward link. The parameters of
to send the information to the base stations. Also, base sta- power control are presented in Table 1 for the reverse and
tions should transfer the measured channel responses to forward links.
other base stations, and that can lead to a great communica- The reverse link power control consists of a closed loop, an
tions overhead through the wire links. In this case the algo- outer loop, and an open loop power control. The closed loop
rithm has to be implemented in a centralized way. In the case power control adjusts the mobile station transmit power in
that the forward and the reverse links are reciprocal, the vir- order to keep the received reverse link SIR at a given SIR
tual reverse link is the same as the real reverse link, which target. The base station estimates the received Dedicated
makes this algorithm amenable to a decentralized implemen- Physical Control Channel (DPCCH) power after RAKE com-
tation. bining of the connection to be power controlled. Simultane-
The performances of the joint transmit beamforming and ously, the base station estimates the total uplink received
power control algorithm with four element antennas and sin- interference in the current frequency band and generates a
gle tap (B: P = 4) and multi-tap (B: P = 4, Q = 3) transmit SIR estimate (SIRest). Power control command is generated
diversity compared to algorithms D (algorithm maximizing the according to the following rule:
received power at the desired mobile with a fixed norm trans-
mit beamforming vector), E (algorithm that maximizes the SIRest>SIRtarget → PC command = “down”
gain toward the desired user while the total transmitted power SIRest<SIRtarget → PC command = “up”
to all other users is minimized), and R1 and R2 (these are the
algorithms D and E with fixed power allocation, without After the reception of the power control (PC) command
power control) for FDD networks with 10 percent difference the mobile station adjusts the transmit power of both reverse
between the reverse and the forward link frequencies are pre- link DPCCH and Dedicated Physical Data Channel (DPDCH)
sented in Fig. 14. It can be seen that this algorithm can signifi-
cantly reduce the total transmitted power in the forward link.
Power control integrated with transmit beamforming can
achieve almost 15 dB higher signal-to-noise-interference ratio Beamformer
weights
(SNIR) than conventional power control algorithm. The simu-
lation scenario consisted of 36 base stations and one user in
w0*
each cell. The link gain model included path loss, shadow fad-
ing, and three-path Rayleigh fading. Base stations were
equipped with four-element antenna arrays.
w1*
Array
elements
POWER CONTROL IN
RADIO TRANSMISSION TECHNOLOGY
PROPOSALS FOR wk*

3G WIRELESS COMMUNICATION SYSTEM

Wideband DS-CDMA is the strongest candidate for the 3G Feedback Beamformer
wireless communication system because most of the proposals channel
for 3G wireless communication systems are based on it [2–6].
As noted earlier, power control is the single most important ■ FIGURE 13. Block diagram of a transmit beamforming system
system requirement for DS-CDMA system. Because of that, [39].

12 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

 Downlink transmit power
1
with a power control step given in Table 1. In the case of the B: P=4,
receiver diversity, the PC command should be generated upon 0.9 Q=3
B: P=4
diversity combining. In the case of soft handover, the quality 0.8 D
measurement of the received signal is performed. E
0.7 R1
The PC command is sent to a mobile station according to R2

Total power
the above rule. If the PC command “up” is sent, the mobile 0.6 PC
compares the commands received from different base stations 0.5
and increases its power only if all commands are “up.” If at
least one “down” command is sent, the mobile station should 0.4
adjust the power with the largest decrease step ordered by the 0.3
PC command received from each base station.
Outer loop power control is used for target SIR adjust- 0.2
ment that is used by closed loop power control. It is based on
0.1
BER and FER measurements and its role is to change the
target SIR when the situation of the mobile is changing or for 0
power control planning. Open loop power control is used for -20 -15 -10 -5 0 5
the initial power setting. Target SNIR [dB]
Forward link power control consists of a closed loop and
an outer loop power control. Closed loop power control ■ FIGURE 14. Total base station power as a function of target
adjusts the base station transmitter power in order to keep the SNIR [39].
received forward link SIR at a given target SIR. The mobile
station estimates received forward link Dedicated Physical
Channel (DPCH) power after RAKE combining and esti- forward link dedicated channels with 800 updates per second.
mates the total forward link received interference in the cur- The closed loop power control compensates medium to fast
rent frequency band. Afterward, the mobile station generates fading and inaccuracies in open loop power control.
a PC command according to the same rule as in the reverse In the case of TDD mode, it is proposed to use the base
link. After the reception of the PC command, the base station station transmission space diversity and an open loop power
adjusts the transmit power in a given direction with a power control. Applying transmission diversity in the forward link,
control step given in Table 1 that may differ in different cells. the instantaneous fluctuation of the received power at the
The outer loop, as in the reverse link, adjusts the SIR tar- mobile station can be reduced. The design of the detailed
get used by a closed loop power control. control for the open loop power control is left as an open
In TDD mode the power control is applied to the limit of issue to each mobile station’s manufacturer’s design.
the interference level within the system in order to reduce the CDMA2000 uses the slotted Aloha principle for packet
inter-cell interference and to reduce power consumption in data transmission. Transmitted power for the random access
the mobile. The proposed power control is a slow C-level- burst is increased after an unsuccessful access attempt.
based power control (similar to GSM) and it is used both for
a reverse and forward link. Power control has the characteris-
tics listed in Table 2. SUMMARY AND DISCUSSION
Also, all codes within one time slot allocated to the same
bearer service use the same transmission power. For real-time In this article, we have presented a survey of the research in
services, closed loop power control is used both for the reverse the field of power control for DS-CDMA reverse and forward
and the forward link. For non-real-time services, both the open link. First, power control techniques are classified according
and the closed loop power control are used according to the to different criteria. Then different power control techniques
mobile’s state and the operator’s needs. The initial power for the DS-CDMA reverse and forward link are presented
value is based on the path loss estimate of the mobile to the with their properties. Also, power control in 3G wireless com-
serving base station path. In the case of a user with simultane- munication system proposals is presented.
ous real-time and non-real-time bearer service, the closed loop Power control is the single most important system require-
power control is used for both bearer services. But depending ment for the DS-CDMA system, because power control miti-
on the current service, different power levels are used. gates harmful degradation of the performance of the users
In the case of packet data transmission, short data packets caused by the near/far effect. Also, with proper power control
can be appended directly to random access bursts. This is called it is possible to lower the total transmit power of the mobiles
common channel packet transmission and is typically used for and prolong battery life. Forward link power control is also an
short, infrequent packets, where the link maintenance needed important requirement because with well defined power con-
for a dedicated channel would lead to unacceptable overhead. trol it is possible to reduce power for the mobile users that
Also, the delay associated with a transfer to a dedicated chan- are either stationary, relatively close to the base station,
nel is avoided. Only open loop power control is applied.
Algorithm Reverse link Forward link
POWER CONTROL FOR THE CDMA2000 SYSTEM
Power control steps Variable Variable
Power control for the CDMA2000 system in FDD mode con-
0.25–1.5db 0.25–1.5db
sists of a fast power control for the reverse link and forward
link. The parameters of power control are given in Table 3. Minimum transmit power –50dBm [ ]dBm
The reverse link power control is the same as in the IS-95
proposal described earlier. The forward link power control is Power control cycles per second 1.6KHz 1.6KHz
different from that defined in IS-95. In the CDMA2000 sys-
Power control dynamic 80dB 30dB
tem, the fast forward link power control is proposed. The
CDMA2000 system uses the fast closed loop power control in ■ Table 1. Parameters of power control in FDD mode [2].

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 13

 Parameter Reverse link Forward link

Dynamic range 80dB 30dB

beamforming and power control for the forward
Power control rate Variable; 2–100 cycles/s Variable; 2–100 cycles/s DS-CDMA link, possibly integrated with base sta-
tion assignment, are very promising techniques in
Step size 2dB 2dB that direction.
Remarks Cycle rate of 100 means Within the one time slot the
that in every frame the powers of all active codes are
REFERENCES
power level is controlled balanced to be within a [1] J. Blom and F. Gunnarsson, “Power Control in Cel-
range of 20dB. lular Radio Systems,” Licentiate Thesis, Department
of Electrical Engineering, Linköpings Universitet,
■ Table 2. Parameters of power control in TDD mode [2]. Sweden, 1998.
[2] ETSI, “The ETSI UMTS Terrestrial Radio Access
(UTRA) ITU-R RTT Candidate Submission,” June
Step size 1dB nominal (0.5 dB & 0.25 dB are available as options) 1998, available at http://www.itu.int/imt/2-radio-
dev/proposals/
Cycles per second 800Hz nominal [3] ARIB, “Japan’s Proposal for Candidate Radio
Transmission Technology on IMT-2000:W-CDMA,”
Dynamic range Open loop: ±40dB June 1998, available at http://www.itu.int/imt/2-
Closed loop: ±24dB around open loop estimate radio-dev/proposals/
[4] TIA, “The CDMA2000 ITU-R RTT Candidate Sub-
Minimum transmit –50dBm mission,” June 1998, available at
power level http://www.itu.int/imt/2-radio-dev/proposals/
[5] TTA, “Global CDMA I: Multiband Direct-Sequence
Accuracy Power control error varies from 1.3dB (low mobility case) CDMA System RTT System Description,” June
to 2.7 dB (high-speed vehicular). 1998, available at http://www.itu.int/imt/2-radio-
dev/proposals/
■ Table 3. Parameters of power control in FDD mode [4]. [6] TTA, “Global CDMA II for IMT-2000 RTT System
Description,” June 1998, available at
http://www.itu.int/imt/2-radio-dev/proposals/
impacted little by multipath fading and shadowing effects, or [7] J. Zender, “Transmitter Power Control for Co-Channel Interfer-
experiencing a minimal amount of other cell interference. ence Management in Cellular Radio Systems,” Proc. 4th Winlab
Therefore, extra power can be given to mobile users who are Wksp., New Brunswick, NF, USA, Oct. 1993.
either in a more difficult environment or far away from the [8] S. A. Grandhi, J. Zander, and R. Yates, “Constrained Power
base station. Control,” Wireless Pers. Commun., vol. 1, No. 4, 1995.
The integration of power control, multi-user detection, and [9] M. Andersin, Z. Rosberg, and J. Zander, “Distributed Discrete
beamforming offers the highest performance of all presented Power Control in Cellular PCS,” Proc. Wksp. Multiaccess,
power control algorithms and can convert some highly loaded Mobility and Teletraffic for PCS, MMT’96, Paris, France, May
1996.
infeasible systems into feasible ones. Also, this power control [10] M. Andersin, Z. Rosberg, “Time Variant Power Control in Cel-
algorithm is of the highest complexity. Similarly, the integra- lular Networks,” Proc. PIMRC’96, Taipei, Taiwan, 1996, pp.
tion of power control and transmit beamforming for forward 193–97.
link offers the highest performance, but the complexity is [11] Z. Rosberg, “Fast Power Control in Cellular Networks based
extremely high as well. In this case additional complexity is a on Short-Term Correlation of Rayleigh Fading,” Proc. 6th Win-
necessity of the centralized implementation of the algorithm, lab Wksp. 3rd Generation Wireless Information Networks, New
while presented algorithms for reverse link could be imple- Brunswick, NJ, Mar. 1997.
mented in a distributed manner. Beamforming and multi-user [12] S. C. Chen, N. Bambos, and G. J. Potie, “On Distributed
detection, combined with the determination of the power con- Power Control for Radio Networks,” IEEE Int’l. Conf. Communi-
cations, ICC’94, 1994, pp. 1281–85.
trol command, must be done for each user communicating via [13] K. S. Gilhousen et al., “On the Capacity of a Cellular CDMA
the base station in question. To perform these in real time Systems,” IEEE Trans. Vehic. Tech., vol. 40, no. 20, May 1991,
could be a very challenging task even for the powerful signal pp. 303–12.
processing units available today. This could cause delays in [14] H. Ji and C-Y Huang, “Non-Cooperative Uplink Power Control
determining the power control command that, in combination in Cellular Radio Systems,” Wireless Networks 4, 1998, pp.
with round-trip propagation delay, could lower the perfor- 233–40.
mance gain of the algorithms. But the delay is inevitable in [15] A. Chockalingam and L. B. Milstein, “Open Loop Power Con-
the conventional power algorithm too. However, develop- trol Performance in DS-CDMA Networks with Frequency Selec-
ments in computer technology in the last few years are such tive Fading and Non-Stationary Base Stations,” Wireless
Networks 4, 1998, pp. 249–61.
that these high-complexity algorithms could be implemented. [16] A. Chockalingam et al., “Performance of CLPC in DS-CDMA
But there is much work to be done to optimize the proposed Cellular Systems,” IEEE Trans. Vehic. Tech., vol. 47, no. 3, Aug.
algorithms for practical implementation and to test them in 1998, pp. 774–89.
real environments. The power control algorithms presented [17] “An Overview of the Application of CDMA to Digital Cellular
here could be the foundation for the development of power Systems and Personal Cellular Networks,” Technical Report,
control algorithms for future wireless systems. Qualcomm Inc., May 1992.
[18] F. Vatalaro et al., “CDMA Cellular Systems Performance with
Imperfect Power Control and Shadowing,” IEEE Vehic. Tech.
CONCLUSION Conference VTC’96, Atlanta, USA, May 1996, pp. 874–78.
[19] Q. Shen and W. A. Krzymien, “Performance Improvement of
CLPC in CDMA Cellular Mobile Communications Systems,”
Power control is a necessary system requirement for wireless Proc. IEEE Vehic. Tech. Conf., Atlanta, USA, 28 Apr. 28–May 1,
systems based on DS-CDMA. One way to further improve 1996, pp. 56–60.
system performance is to implement more complex techniques [20] Y.-J. Yang and J.-F. Chang, “A Strength and SIR Combined
for power control. Joint power control, multi-user detection, Adaptive Power Control for CDMA Mobile Radio Channels,”
and beamforming in the reverse DS-CDMA link, and transmit Proc. IEEE 4th Int’l. Symp. Spread Spectrum Techniques and

14 IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000

 Applications, vol. 3, 1996, pp. 1167–71. Yugoslavia. From 1995 to 2000 he was a telemetry system engi-
[21] D. Famolari et al., “A New Framework for Power Control in neer at NIS-GAS, Novi Sad, Yugoslavia. In 2000 he joined the Cen-
Wireless Data Networks: Games, Utility and Pricing,” Wireless tre for Wireless Communication, University of Oulu, Finland,
Multimedia Network Technologies, Kluwer Academic Publish- where he currently works as a research scientist. His research
ers, 1999, pp. 289–80. interests include multi-user detection, power control, multiple-
[22] R. D. Yates and C.-Y. Huang, “Integrated Power Control and antenna systems, and communications over fading channels.
Base Station Assignment,” IEEE Trans. Vehic. Tech., vol. 44, no.
3, Aug. 1995, pp. 638–44. MIROSLAV L. DUKIC ([email protected]) received the B.Sc. degree
[23] S. V. Hanly, “An Algorithm for Combined Cell-Size Selection in electronics and communications in 1973, the M.Sc. degree in
and Power Control to Maximize Cellular Spread Spectrum 1975, and the Ph.D. degree in 1981, all from the faculty of elec-
Capacity,” IEEE JSAC, vol. 14, no. 7, Sept. 1995, pp. 1332–41. trical engineering, the University of Belgrade, Yugoslavia. In 1973
[24] F. Rashid-Farrokhi, K. J. R. Liu, and L. Tassiulas, “Downlink he joined the faculty of electrical engineering, the University of
and Uplink Capacity Enhancement in Power Controlled Cellular Belgrade, where he currently holds the position of associate pro-
Systems,” Proc. IEEE Vehic. Tech. Conf. VTC’97, Phoenix, USA, fessor in the department of communications. Since 1973 he has
May 1997. taught several courses in communications theory, radar systems
[25] S. V. Hanly and D. Tse, “Power Control and Capacity of design, and modern electronic warfare systems. He also partici-
Spread-Spectrum Wireless Networks,” Automatica, vol. 35, no. pated as a project engineer in several communications projects in
12, Dec. 1999, available at http://degas.eecs.berkeley.edu/ Yugoslavia, and he has published a number of papers in the field
~dtse/pub.html, pp. 1987–2012. of communications. His research interests lie in the area of spread
[26] J. X. Qui and J. W. Mark, “A Dynamic Load Sharing Algo- spectrum communications, radar, and electronic warfare systems.
rithm through Power Control in Cellular CDMA,” Proc. 9th IEEE
Int’l. Symp. Personal Indoor and Mobile Radio Communications
PIMRC’98, Boston, USA, Sept. 1998.
[27] S-L. Kim, “A Two-Mode Algorithm for Combined Power Con-
trol and Base Station Assignment,” Proc. 10th IEEE Int’l. Symp.
Personal, Indoor, and Mobile Radio Communications, Osaka,
Japan, 1999.
[28] F. Rashid-Farrokhi, L. Tassiulas, and K. J. R. Liu, “Joint Opti-
mal Power Control and Beamforming in Wireless Networks
Using Antenna Arrays,” IEEE Trans. Commun., vol. 46, no. 10,
Oct. 1998.
[29] S. Ulukus and R. D. Yates, “Adaptive Power Control and
MMSE Interference Suppression,” Baltzer/ACM Wireless Net-
works, vol. 4, no. 6, June 1998, pp. 489–96.
[30] A. Yener, R. D. Yates, and S. Ulukus, “Joint Power Control,
Multiuser Detection and Beamforming for CDMA Systems,”
IEEE Vehic. Tech. Conf., Houston, USA, May 1999.
[31] D. M. Novakovic and M. L. Dukic, “An Analysis of Integrated
Power Control and Advanced Receivers Algorithms for Reverse
DS-CDMA Link,” IEEE 6th Int’l. Symp. Spread Spectrum Tech-
niques and Applications ISSSTA 2000, New Jersey, USA, Sept.
2000.
[32] T. Ito, S. Sampei, and N. Morinaga, “A Wireless Packet Trans-
mission with Adaptive Processing Gain and Transmitter Power
Control Scheme for Circuit-Switched and Packet-Switched
Modes Integrated DS/CDMA Systems,” IEEE 49th Vehic. Tech.
Conf., vol. 3, 1999, pp. 2039–43.
[33] W. C. Y. Lee, “Overview of Cellular CDMA,” IEEE Trans. Vehic.
Tech., vol. 40, no. 2, May 1991.
[34] R. R. Gejji, “Forward-Link-Power Control in CDMA Cellular
Systems,” IEEE Trans. Vehic. Tech., vol. VT-41, Nov. 1992.
[35] M. Zorzi and L. B. Milstein, “Power Control on the Forward
Link in Cellular CDMA,” Proc. Int’l. Zurich Seminar’94, Mar.
1994, pp. 391–99.
[36] M. Zorzi, “Optimization of a Class of Power Control Laws for
the Forward-Link in a Cellular CDMA System,” IEEE Globe-
com’94, San Francisco, USA, Nov. 1994.
[37] M. Zorzi, “Simplified Forward-Link Power Control Law in Cel-
lular CDMA,” IEEE Trans. Vehic. Tech., vol. 43, no. 4, Nov.
1994.
[38] R. Prasad and T. Ojanpera, “An Overview of CDMA Evolution
Toward Wideband CDMA,” IEEE Commun. Surveys,
http://www.comsoc.org/pubs/surveys, vol. 1, no. 1, 4th Quarter
1998.
[39] F. Rashid-Farrokhi, K. J. R. Liu, and L. Tassiulas, “Transmit
Beamforming and Power Control for Cellular Wireless Sys-
tems,” IEEE JSAC, vol. 16, no. 8, Oct. 1998.

BIOGRAPHIES
DEJAN M. NOVAKOVIC ([email protected]) received the B.Sc. degree in
electronics and communications in 1995 from the faculty of tech-
nical sciences, department of electrical engineering, the University
of Novi Sad, Yugoslavia, and the M.Sc. degree in 1998 from the
faculty of electrical engineering, the University of Belgrade,

IEEE Communications Surveys • http://www.comsoc.org/pubs/surveys • Fourth Quarter 2000 15

View publication stats