CONTROL

The present invention relates to a method for determining SIR target for outer loop power control in a wireless communication system. The invention also relates to a device for determining SIR target, use of such a SIR target value in a wireless communication system, a computer program and a computer program product.

Background of the Invention

Wideband Code Division Multiple Access (WCDMA), an air interface standard derived from Code Division Multiple Access (CDMA), is the most widely used air interface for the third generation (3G) telecommunication networks and is also known as the IMT-2000 Direct Spread. Its specification has been created in the 3rd Generation Partnership Project (3GPP). While the earlier systems were developed with voice communications in mind, the WCDMA has been designed for multimedia communication with higher data rates and increasingly demanding Quality of Service (QoS) requirements.

In direct-sequence CDMA communication systems several users share the same radio bandwidth simultaneously through the use of a specific spreading code. These codes are not exactly orthogonal and hence there exists interference from users of the same cell and interference from users in nearby cells. Interference increases as the number of users increase in the system, and hence the system may be interference-limited. Apart from this limitation, the system capacity is further limited by multipath and near/far effects.

Power control has been shown to mitigate both these effects and also result in longer battery lives by controlling the transmit powers of all the users such that the received powers for all users at a base station are equal - assuming uplink transmission and that the users employ the same service, such as speech. Furthermore, power control is used to ensure that the received Signal-to -Interference Ratio (SIR) is good enough to maintain the required QoS.

Power control is performed both in the uplink and the downlink. In the downlink, the power control is used for cases where users experience large path losses such that the received signal is of the order of the noise, but the power control requirement is not as stringent as it is for the uplink. In the uplink, power control is mainly used to mitigate the near/far effects and effects of fading radio channels. As can be seen in Figure 1, the transmit power control on the uplink consists of two parts that are running simultaneously, an inner loop power control (or fast loop power control) and an Outer Loop Power Control (OLPC). The fast loop power control is used to counteract the effects of fast fading by controlling the transmit power of the mobile stations, through power control commands ("power up" or "power down" commands) sent on the downlink, so that received SIR reach a specified SIR target. The fast loop runs between the base station (also known as the Node B) and the mobile station (also known as a User Equipment (UE)).

However, the fast loop power control employed in WCDMA systems has its drawbacks, it is restricted by a limited set of power adjustment steps and an update rate of 1500 Hz, and the received SIR still has some variations due to varying propagation channel conditions. This variation is influenced by factors such as the power delay profile, the number of resolvable propagation paths, the fading maximum Doppler frequency (dependent on the velocity of the UE), etc. Thus, it is difficult to have a fixed SIR target that will result in meeting a fixed QoS target.

Traditionally, channel quality indicators such as the Block Error Rate (BLER) or Frame Error Rate (FER) are used to specify the required QoS targets. Usually, the assessment of the Cyclic Redundancy Check (CRC) is used to determine whether to increase the SIR target in case an error occurs and decrease the SIR target when a good block is received.

One prior art power control algorithm is the sawtooth algorithm that is used in CDMA systems. In the sawtooth algorithm, the SIR target (measured in dB) is adjusted step-wise with a different step size each for step-up and step-down. According to the sawtooth algorithm, whenever a block of data (a transport block in 3GPP terminology) is received in error the SIR target (measured in dB) is increased by a value equal to SIR_step-up and each time a block of data is received with no error, the SIR target is decreased by a value equal to SIR_step-down. The SIR_step-up and SIR_step-down values are related to the BLER target and a common variable SIR step as follows: No error SIR_step_down = SIR step ^■ BLER target,

Error SIR step up = SIR step - SIR_step_down, where SIR step (step size) determines the rate of the algorithm to converge to the ideal SIR target and also the stability of the obtained BLER. An ideal SIR target represents a SIR value, which if maintained, would result in a BLER that will be equal to the BLER target value.

With large step sizes, the system can achieve a high convergence speed, but this is usually associated with significant instability or oscillations of the obtained BLER around the BLER target. In contrast, with small step sizes the system can achieve a quite stable BLER, but it takes a long time to converge to the stable BLER. So, this is a trade off scenario and often a SIR step size of 0.5 dB is employed as the optimum trade off value between overhead and convergence speed. Figure 2 shows the working of a typical sawtooth implementation with BLER target of 1 % and a step size (SIR_step) of 0.5 dB; and figure 3 shows simulation results for the sawtooth algorithm where: A = PedA 30 km/h, B = PedB 3 km/h, C = PedA 3 km/h, D = VehA 30 km/h, E = AWGN 3 km/h, and F = VehA 120 km/h.

In practice it has been observed that though the sawtooth algorithm more or less maintains the required BLER it has some major drawbacks, such as:

• Slow in adjusting to a sudden improvement in channel conditions - a convergence rate of around 0.12 dB/sec;

• Slow in reacting to a sudden deterioration in channel conditions - a convergence rate of around 0.5 dB/sec;

· No SIR target upper limit - thus when a mobile station hits its maximum power limit or channel conditions remain very poor, the SIR target can increase to very high values and will take a long time to come back to reasonable values when the channel conditions become better leading to wastage of power. Hence, enhancements of the sawtooth algorithm that include a SIR target upper limit are widely employed. In order to overcome these issues, channel quality estimates other than CRC errors were investigated and one such was derived based on the Power Delay Profile (PDP) of the multipath fading channel. According to another prior art solution changes in received power in the strongest path and the ratio of the received power between the strongest path and the second strongest path is used to determine the required change in SIR target. A Channel Discriminating Filter (CDF) is used to track the changes of the channel and whenever a change in channel is detected, the new channel is mapped to a predefined channel type and the SIR target is immediately adjusted to correspond to the new channel type. This solution is also able to detect the difference between a static channel and a dynamic channel - with the former the first channel tap does not fade, i.e. fluctuate in magnitude. A certain set of radio channels are preprogrammed and the method is able to adjust to the dynamically changing channel conditions quite quickly, but is limited by the time to assess the current channel against the pre- programmed set of channels.

Specifically, this method reacts quite well to a sudden improvement in channel conditions when compared to the sawtooth algorithm. However, this method works best if the channel conditions can be mapped to one of the predefined channels as defined in the 3 GPP standards. Indeed, the method only works for channels which have been pre-programmed into the algorithm. Another possible issue with this prior art solution is the fact that the method does not take into account the difference for two UEs with the same channel taps but different speeds. Faster UEs suffer more rapid fading, and require larger SIR targets but the rate of fading is not apparent from the channel response measured by this CDF method.

According to yet another prior art solution, knowledge of stochastic distribution of the fading channel models is used to find the SIR fade margin that satisfies a given outage probability in that link. This SIR fade margin is a function of the current channel propagation conditions and the SIR target value is determined as follows:

SIR target = SIRBLER target + SIRoutage target, where SIR _BLER target is the error driven component (sawtooth based) and SIRoutage target is the component based on the SIR fade margin calculated. The fade margin = SIRmean - SIRoutage _point measured over a set of SIR samples. According to this prior art solution, the measured SIR fade margin (or many fade margins calculated at different outage probabilities) is used as an input to a neural network, which is trained for different channel conditions, to determine the SIR target value. Figure 4 shows how to calculate the SIR fade margin from a Cumulative Distribution Function (CDF) of SIR samples. The results for this prior art solution show a good response to sudden improvement in channel conditions as compared to the sawtooth algorithm. However, the algorithm is complex and requires training for various channel conditions. The algorithm also demonstrates some latency (delay) in responding to a new channel for which the neural network has not yet been trained.

Adaptive step size algorithms have also been proposed that make use of memory based adjustments to the transmit power step size of the inner loop, this is the change in transmit power on receiving a power up or down command, typically 1 or 2 dB. The transmit power step size is decreased when the UE receives consecutive alternate up and down power control commands, thus reducing the power control error when the transmission power is close to the required value. Update step size is increased when several consecutive up (or down) power control commands are received to increase the convergence rate.

It has also been proposed to modify the step size based on the difference between the SIR target and the received SIR. Yet another approach decreases the step size when a number of down commands are received consecutively so that the power never goes below the required value and the condition of oscillating up and down commands never occurs.

Results show an improvement in the convergence rate of the SIR target, with these adaptive step size algorithms, the convergence rate increases six fold compare to that of the sawtooth algorithm. However, it was also noticed that there is a trade-off between the convergence speed and the stability of the system. Algorithms which converge faster were found to be rather instable causing some abrupt increases in outage probabilities, thereby resulting in instable BLERs. Fuzzy control based algorithms have also been proposed where the transmit power of an UE is set based on the output of a membership function which takes as an input an error and an error change quantity. The error described here is the difference between the SIR target and the received SIR, or the BLER target and the BLER. This leads to faster convergence rates, and better stability while also reducing the time delay that is inherent between the Node B and the UE. Implementation of these fuzzy control schemes is however indeed complex when compared to the other schemes and it is used in the inner loop power control rather than outer- loop power control. Summary of the Invention

An object of the present invention is to provide a method for determining SIR target for outer loop power control in a wireless communication system which solves or mitigates the drawbacks of prior art. Another object of the invention is to provide an alternative solution for determining SIR target for outer loop power control in a wireless communication system.

According to one aspect of the invention, the objects are achieved with a method for determining SIR target for outer loop power control in a wireless communication system, said method comprising the steps of:

- obtaining a plurality of SIR values for a communication node;

- determining a SIR spread value SIR _Spread based on a mean SIR value SIR _Mean of said plurality of SIR values and a ^-percentile value SIR _nVa , said ^-percentile value being a SIR value below which n % of said plurality of SIR values falls; and

- determining a SIR target SIR _{j a} for outer loop power control based on said SIR spread value SIR _Spread and a SIR target value SIR _TargetAWGN in an AWGN channel for a given QoS target.

Different embodiments of the method above are disclosed in the dependent claims 2-13.

The invention also relates to use of a SIR target determined according to any of the above methods for downlink or uplink outer loop power control in a wireless communication system. Further, the invention also relates to a computer program and a computer program product. According to another aspect of the invention, the objects are also achieved with a device for determining SIR target for outer loop power control in a wireless communication system, said device being adapted to:

- obtain a plurality of SIR values for a communication node;

- determine a SIR spread value SIR _Spread based on a mean SIR value SIR _Mean of said plurality of SIR values and a ^-percentile value SIR _nVa , said ^-percentile value being a SIR value below which n % of said plurality of SIR values falls; and

- determine a SIR target SIR _j , ^ for outer loop power control based on said SIR spread value SIR _Spread and a SIR target value SIR _TaigetAWGN in an AWGN channel for a given QoS target.

The device above can be modified, mutatis mutandis, according to the different embodiments of the method above.

The invention provides a method and a device which realizes an outer-loop power control algorithm in a wireless communication system.

According to an embodiment of the present invention, the power control algorithm may be applied to either downlink or uplink; on the downlink the algorithm may be performed in the mobile station and in the uplink the algorithm may be located at the base station and the RNC.

A method and device according to the invention provides a high performance power control algorithm for wireless communication system that is able to quickly tune a SIR target close to an ideal value as the radio channel experienced by the mobile station changes - this could be a change in the multipath profile of the channel or a change in the speed of the mobile station, or both of these.

As a result the received SIR from the mobile station fluctuates less - so interference to other mobile stations is smaller and more stable. Moreover, the BLER is more tightly bound to the target value, so helping the perceived performance by the end user.

Further, there is a reduction in power consumption which results in prolonged battery e.g. in mobile stations. By employing the present algorithm, it is possible to realize an effective outer loop power control with low complexity that can work with different BLER targets and in different radio conditions for different code rates, modulations, etc.

Other applications and advantages of the present invention will be apparent from the following detailed description.

Brief Description of the Drawings

The appended drawings are intended to clarify and explain different embodiments and aspects of the present invention in which:

- Figure 1 shows uplink power control in a wireless communication system;

Figure 2 shows an example of the sawtooth power control algorithm;

Figure 3 shows simulation results of the sawtooth algorithm;

Figure 4 shows a prior art method to determine the "Fade Margin";

Figure 5 shows AWGN BLER curves for evaluated services;

- Figure 6 shows SIR margin versus SIR spread for different radio channels and mobile station speeds, and different BLER targets. Service simulated is normal speech, and the fitted lines assume a non-linear mapping;

Figure 7 SIR margin versus SIR spread for different radio channels and UE speeds, and different BLER targets. Service simulated with a shallower BLER curve, and the fitted lines assume a non-linear mapping;

Figure 8 shows soft handover uplink power control according to the invention in a wireless communication system;

Figure 9 shows a typical BLER vs. SIR curve with linear interpolation to determine ideal SIR target values; and

- Figure 10 shows a flow chart of an embodiment of the present invention.

Detailed Description of Embodiments of the Invention

To achieve the aforementioned objectives, the present invention relates to a method for determining SIR target value for outer loop power control in a wireless communication system. The method comprises the steps of: obtaining a plurality of SIR values for a communication node; determining a SIR spread value SIR _Spread based on a mean SIR value SIR _Mean of the plurality of SIR values and a ^-percentile value SIR _n% , the ^-percentile value being a SIR value below which n % of the plurality of SIR values falls; and determining a SIR target SIR _j , _α for outer loop power control based on the SIR spread value SIR _Spread and a SIR target value SIR _TaigetAWGN in an AWGN channel for a given QoS target.

Preferably, the SIR mean value SIR _Mean and ^-percentile value SIR _n% are determined from a

CDF calculated from the plurality of SIR values obtained e.g. by sampling received radio signals.

According to an embodiment of the invention, the QoS target is a BLER target for the AWGN channel. Different services will require different BLER targets, and this may be used to determine the SIR _TargetAWGN from the characteristic curve of BLER against SIR target SIR _Twget for the AWGN channel. The characteristic curve is itself dependent upon the radio bearer employed to carry the data blocks, principally according to the code rate and modulation employed.

In another preferred embodiment, the SIR spread value SIR _Spread is defined as a difference between the mean SIR value SIR _Mean and the ^-percentile value, and determined by the relation:

The SIR _Spread may be used to determine a SIR margin - i.e. the difference from the ideal SIR target for the channel and the AWGN SIR target - as a function of the SIR spread.

Furthermore, n is a real number preferably larger than 0 and smaller than 10, and preferably smaller than 1.5. Simulations have shown the SIR spread value SIR _Spread for n = 1.0 (approximately) is able to clearly distinguish one channel type from the other. Other n percentiles values, such as 10-percentile or 5-percentile, did not vary much from one channel type to another, whilst even smaller percentiles, such as 0.1 -percentile or 0.5-percentile required many more SIR samples to calculate them accurately thereby increasing the time over which the SIR target SIR _Twget is determined, thus reducing the responsivity of the present algorithm.

According to yet another embodiment of the invention, the SIR target SIR _Tsiget is determined by a first function f _x of the SIR spread value SIR _Spread and given by the relation:

SIRT zig et ⁼ fl (^^Sprea ) + zig et _ AWGN ·

Preferably, the first (parametric) function f _x is also dependent on further parameters derived from BLER as a function ("curve") of SIR, or vice versa, in the AWGN channel. The inventors have realised that the following parameters describing the relation between BLER and SIR in the AWGN channel can be used: BLER target, a first derivative/gradient of BLER as a function of SIR, a second derivative of BLER as a function of SIR, and a difference in a SIR target for two different BLER values works well with the present power control algorithm. The present algorithm is, among other things, based upon the observation that the ability of the closed-loop power control to track the fading of the channel and give a measured SIR equal to the SIR target value SIR _Twget deteriorates as the radio channel fades become deeper

(channel with less diversity) and more closely packed in time (UE travels faster). The SIR target SIR _Tmget in these difficult conditions is found to be higher to compensate for these greater received SIR variations. Furthermore, the characteristics of the BLER versus SIR curve (in the radio channel) influence the impact of the SIR fluctuations about the intended SIR target value SIR _Tsigei (because of the imperfections discussed above). If the curve is very steep, a small SIR fluctuation can lead to a large fluctuation in BLER. Thus, the gradient of the curve impacts the BLER, and to achieve the target BLER the SIR must be adjusted accordingly. There is an assumption that the shape of the AWGN curve is related to that of the curve in the (non-AWGN) radio channel. Furthermore, it is not only the first derivative/gradient of the BLER versus SIR curve which is of interest. When the BLER target is high (e.g. when BLER equals to 0.1) the BLER curve is more curved than for lower BLER target values. Consequently, when the inner loop power control is unable to set the SIR at the target value, and the error/difference is large, the SIR spread and gradient terms in the mapping first function f _x does not fully capture the impact on the BLER, and therefore the impact on the SIR target SIR _Twget increase over the AWGN value to hit the desired BLER target. Since the BLER curves are concave (negative second derivative) the linear approximation is somewhat pessimistic and to correct for this the SIR target SIR _Taget may be adjusted by a term proportional to the second derivative of the curve measured at the BLER target point which reduce the SIR target SIR _Tsigei . Using further information derived from the BLER/SIR curve for the AWGN channel is possible.

An example of the first mapping function f _x linearly dependent on SIR spread SIR _Spread is given by: f _x = 0.1 - x - (0.45 + 0.15 - . - (4 - z)) , where

x - SIR _Spread ,

j = -lo _g 10(i?JER _raig J , and

z = AWGN _{SIR Taiget} (BLER = 0.001) - AWGN _{S1R T} ^ _get (BLER = 0.1) .

Here the term z is related to the gradient of the BLER/SIR curve in the region between BLER values 0.1 and 0.001. Another example of the first mapping function f _x , which is a non-linear mapping of SIR spread SIR _Spread , is given by:

x = SIR _Spread , y = -loglO(BLER _Taiget )

z = AWGN _{slR T∞get} (BLER = 0.001) - AWGN _S1R ^BLER - 0.1)

This non-linear mapping function was derived from a set of simulations which determined the SIR target SIR _{j a} for a number of different radio channels - i.e. different tap profiles and different speeds. Simulations were performed for two different radio bearers (representing two different speech-like services carried with different code rates), and therefore two different AWGN BLER versus SIR characteristics (see Figure 5). The first function f _x is applicable for both services at any desired BLER target.

Other non-linear mappings can be derived, and the described above are only exemplary. The first function f _x is in general a non-linear function of SIR spread SIR _Spread and one or more of the parameters described above. Figure 6 shows SIR margin versus SIR spread for different radio channels and mobile station speeds, and different BLER targets. Service simulated is normal speech, and the fitted lines assume a non-linear mapping. Further, figure 7 shows SIR margin versus SIR spread for different radio channels and UE speeds, and different BLER targets. Service simulated with a shallower BLER curve, and the fitted lines assume a non-linear mapping (using the same first function f _x but with a different parameter value z).

The inventors have also realised that the first function f _x may be derived using measurement data, and hence the present method may further comprise the steps of: fixing one or more different values for SIR target SIR _Ta[get ; and measuring BLER and a SIR spread SIR _Sprmd value for each SIR target SIR _Twget value so as to determine the first function f _x .

The mapping between SIR spread SIR _Spread and the first function f _x may be derived using measurements taken with mobile stations in the "field". Different fixed SIR targets are set, and the BLER is measured for each SIR target value. It is important that the channel does not change (i.e. the mean tap powers should be constant) and the mobile station speed is constant which necessitates that the measurements are taken over a short period as possible. However, the measurement period must be sufficient such that the BLER can be measured reasonably accurately. For example, to measure a mobile station which is using a speech service delivering 50 blocks per second, a measurement of 10 s or more is needed when the SIR target SIR _Twget is close to the value that gives a 1 % BLER. The relationship between SIR(dB) and log(BLER) is approximately linear over a large span of different BLER values so linear interpolation can be used to determine the ideal SIR target for the target BLER value as shown in Figure 9. In this figure it is shown how measurements of BLER at 3 different SIR values (the circles) may be used to determine the SIR value for the target BLER value by linear interpolation. Since measurements at low BLER take a long time it may be better to take measurements for high BLER and extrapolate to the BLER target point. Using the method described above, one point can be found on the SIR target margin vs. SIR spread curve. More points require further measurements of other UEs, or the same UE at different times. Gradually, more and more points are added and the curve can be sketched.

It is also common in wireless communication systems to use RAKE receivers with a number of RAKE fingers. Therefore, for determining the first mapping function f _x : instead of looking for mobile stations with different channels, the channel perceived by the mobile station is changed by adjusting the combination of RAKE fingers employed by the Node B in the uplink case. Fewer RAKE fingers will reduce the diversity of the channel and result in a larger SIR spread, thus points on the right hand side of the first mapping function f _x can be obtained. The mapping function can evolve over time - for example, to begin with a default mapping is employed (pre-programmed) but this can be refined as field measurements are taken as described above. Hence, according to an embodiment the step of measuring further involves: using different combinations of RAKE fingers for determining the first function f _x so as to take advantage of different diversity levels. According to yet another embodiment of the invention, the SIR target SIR _Target is further determined by a second function f ₂ and hence given by the general relation: where the second function f ₂ is dependent on BLER and BLER target BLER _{Twg et} .

Preferably, the second function f ₂ is given by the sawtooth algorithm described above so that the value of the second function f ₂ is:

· increased with SIR _{Step Down} = SIR _Step ^■ BLER _{Taig et} if a block of data is received correctly; and

• decreased with SIR _{Step Up} = SIR _Step - SIR _{Step Down} if a block of data is received in error.

The present power control algorithm may also be extended to soft handover cases. In soft handover on the uplink, the RNC performs selection combining of data blocks sent on all the radio links ("legs") of an active set. The outcome of this can be used to drive the sawtooth algorithm. Additionally, since the radio channel may be different on each radio link separate SIR spread values may be sent to the RNC, one for each radio link. Consequently, different SIR targets SIR _Twget may be calculated for different radio links. In the uplink case, it is reasonable for the RNC to issue different SIR target values for each radio link. These would follow the same method described above which is for the hard handover case (i.e. one radio link only in the active set), but would be reduced by an amount dependent upon the number of radio links in the active set to compensate for the combining gain. If all radio links have an equal probability of an erroneous transmission given by p, then the probability that all radio links provide an unsuccessful transmission is p ^r , where r is the number of radio links. Thus, as the number of radio links changes this could be managed by adjusting the BLER target for each individual radio links (new BLER target p is (overall BLER target) ^1/r ). The change in BLER target is easily done if the parameterized first function f _x can be used. Hence, a SIR target SIR _{Targ et} for each radio link is determined in the soft handover case.

Figure 8 shows a UE with two radio links in the active set to the same RNC (if the active set spans more than one RNC then the SIR spread would be passed over the lur interface to the Serving RNC but there would be no other impacts). The two base stations (BS) involved measure the SIR spread (individually) and send the SIR spread value together with the transport blocks to the RNC. The RNC thereafter calculates the SIR target SIR _{Tsig et} to use for each radio link and sends this to each base station, respectively. In the uplink case, the invention can be implemented as follows:

Sample (measure) the received SIR values at the Node B over a time interval (e.g. 1 s) Node B calculates a SIR spread value SIR _Spread using the relation espread ⁼ ^^Mean ^~ *^n%

RNC calculates the SIR target using the relation:

SIR _T arg et = + SIRj _ AIVGN ·

A variant of the algorithm adds the SIR adjustment determined by the sawtooth algorithm to the SIR target, i.e. : S R/ et - fl (SIRSpread ) + fl + etA WGN ' where the second function takes values according to the sawtooth algorithm as aforementioned. Figure 10 shows the operation of an embodiment of the invention applied to uplink power control. The Node B measures the SIR of UE transmissions every timeslot (0.666 ms) and when 1 second (for example) worth of samples has been taken the SIR spread value

SIR _Spread is calculated. This value is passed to the RNC over the Iub interface. The RNC calculates a revised SIR target SIR _{Ts[g et} using the SIR spread value SIR _Spread , and optionally from the errors in blocks passed over the Iub interface - if the second function f ₂ is included in the power control algorithm. The RNC thereafter sends the revised SIR target value SIR _Twget to the Node B over the Iub interface whenever it is recalculated.

It should be obvious to the skilled person that the present power control algorithm is applicable both in the uplink and the downlink. Further, the power control algorithm can be employed in a number of different suitable wireless communication systems, such as GSM, WiMAX and 3 GPP UTRAN systems.

The present invention further relates to the step of using a SIR target SIR _j , _a , determined by a method according to the present invention, for downlink or uplink outer loop power control in a wireless communication system.

Furthermore, as understood by the person skilled in the art, the method for determining SIR target for outer loop power control according to the invention may be implemented in a computer program, having code means, which when run in a computer causes the computer to execute the steps of the method. The computer program is included in a computer readable medium of a computer program product. The computer readable medium may consist of essentially any memory, such as a ROM (Read-Only Memory), a PROM (Programmable Read-Only Memory), an EPROM (Erasable PROM), a Flash memory, an EEPROM (Electrically Erasable PROM), or a hard disk drive.

The invention relates further to a device for determining SIR target for outer loop power control in a wireless communication system. The device is adapted to: obtain a plurality of SIR values for a communication node; determine a SIR spread value SIR _Spread based on a mean SIR value SIR _Mean of said plurality of SIR values and a ^-percentile value SIR _n% , said ^-percentile value being a SIR value below which n % of said plurality of SIR values falls; and determine a SIR target SIR _j , _a for outer loop power control based on said SIR spread value SIR _Spread and a SIR target value SIR _TaigetAWGN in an AWGN channel for a given QoS target. The device may according to an embodiment be a base station, mobile station, or a RNC. It should also be noted that the device above may be modified, mutatis mutandis, according to the different embodiments of the method described above. Finally, it should be understood that the present invention is not limited to the embodiments described above, but also relates to and incorporates all embodiments within the scope of the appended independent claims.