WO2015192909A1

WO2015192909A1 - Controlling a de-sensitisation level to be applied at an uplink radio receiver

Info

Publication number: WO2015192909A1
Application number: PCT/EP2014/062975
Authority: WO
Inventors: Torbjörn WIGREN
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2014-06-19
Filing date: 2014-06-19
Publication date: 2015-12-23
Anticipated expiration: 2016-12-19

Abstract

The invention provides an optimum amount of desensitization to a RDS when it is necessary due to inter-cell interference. In this way, the amount of inter-cell interference created by RDS is reduced, while maintaining the uplink performance in RDS irrespective of any incoming neighbor cell interference. The invention therefore contributes to enhance the general mobile broadband performance for RDS. The advantages of the invention are illustrated in Figures 4 and 3. Note that the difference between Figures 3 and 4 is that, in Figure 4, when the invention detects a too high inter-cell interference, it starts to increase the thermal noise power floor by ΔP_d (t). This is illustrated by the thick line that starts at the arrow marked "invention starts to add desensitization". Since the headroom for scheduling users in the RDS UL is counted with respect to the thick line, this headroom remains until soft handover takes place. Then the interfering UE changes to the RDS cell and the inter-cell interference problem disappears.

Description

Controlling a de-sensitisation level to be applied at an uplink radio receiver

Technical Field

The invention relates to apparatuses and methods to selectively reduce the uplink receiver's sensitivity, primarily, in a Radio Dot System. This helps to balance the difference between the downlink output power and the uplink sensitivity, which results in an inhomogeneous coverage for the Radio Dot Cell. The method and apparatus presented selectively provide a minimal level of desensitisation in the uplink receiver.

Background

Radio Dot System and Power Imbalance

A Radio Dot System (RDS) is an indoor system that comprises a radio unit with dedicated RDS functionalities, 1 -8 radio heads, transceivers and Ethernet cabling that connects the radio heads to the radio unit. The RDS is multi-standard and supports WiFi signal carriers and both 3G's Wideband Code Division Multiple Access (WCDMA) and 4G's Long-Term Evolution (LTE) cellular systems.

A RDS comprises an antenna element in the radio heads, which delivers mobile broadband access to user equipment (UE) in a broad range of indoor locations. The radio heads are connected via standard Ethernet cables to indoor radio units that link to a base station. RDSs aim at providing a users' experience that is consistent regardless of whether they are outdoors or indoors. To achieve this, the indoor and outdoor networks should evolve in lockstep, securing that the link present in a two-way communication is balanced. RDSs have to ensure that the transmitted downlink path to the UE's receiver will match the transmitted uplink path from the UE to the radio unit receiver in the RDS.

The downlink transmitter at the radio unit has normally an output power of 50 mW, while the uplink receiver at the radio head has generally a basic sensitivity that is not much worse than that of a normal cellular radio receiver. Uplink signals received at the radio heads and then transmitted to the radio unit suffer power losses as they travel through the Ethernet cable connecting them. These power losses are of the order of tens of dBs and they depend on the length of the Ethernet cable, which currently may be of up to 200 m. To overcome these power losses in the uplink direction, the signals received in the radio heads can be amplified before sending them to the radio unit, which results in the uplink having coverage comparable to those of standard cellular radios.

On the other hand, the downlink transmitter at the radio unit can never transmit with more power than 50 mW. This signal's power is about 20 times smaller than the standard UE transmission power. Consequently, even if the downlink signal is transmitted from the radio head, this signal will be weaker than the uplink signal, resulting in lower downlink coverage.

The uplink and downlink in the RDS are, therefore, unbalanced, which results in a coverage that is not uniform throughout the RDS cell.

Moreover, to provide receiver diversity and to be able to support Multiple-Input Multiple- Output (MIMO) in LTE cellular systems, the RDS system has two antenna branches. The uplink signals coming from the radio heads are combined in the analogue domain, separately for each antenna branch.

A first consequence of such an imbalance in coverage takes place when a UE switches from one set of radio resources -such as a serving outdoor cell- to another set -such as the RDS-. This switching is known as 'handover' functionality and is controlled by the downlink. A UE enters handover when the downlink powers of the RDS and the serving cell are of comparable size.

In WCDMA the handover is normally soft ("make before break"): the new connection between the UE and the RDS cell is activated before the old connection between UE and serving cell is broken. In LTE, on the other hand, the handover is normally hard ("break before make"): the connection between UE and serving cell is broken before the UE connects to the RDS cell.

Due to the downlink power imbalance between the RDS and the serving outdoor cells, the handover conditions are met when the distance of the UE to the RDS cell is significantly smaller than the distance of the UE to the serving cell. At this point, to maintain the connection with the serving cell, the UE is transmitting at high power, which results in a substantial amount of inter-cell interference (also referred to in this document as "neighbor cell interference") to the RDS cell. This inter-cell interference can easily drain the load headroom for mobile broadband in WCDMA, disturbing the operation of the system and causing dropped calls.

Desensitisation

One way to rebalance the downlink-uplink coverage is by reducing the uplink receiver's sensitivity using a functionality known as 'desensitization'. This can be implemented by adding noise to the uplink signal, at a suitable point in the receiver chain. For instance, this could be after digitalization in the radio unit. An extra noise source adds to the electronics noise already present at the radio heads' antenna, creating an artificial increase of the thermal noise power floor. This results in a degradation of the signal-to- noise ratio in the uplink receiver, which reduces its sensitivity.

Desensitization then tries to balance the RDS cell coverage by reducing the uplink coverage to try to match the downlink coverage. Unfortunately, this is not a perfect solution. As mentioned above, the UE needs to increase its transmitting power before handover due to the downlink power imbalance between the RDS and serving cells. Now that desensitization has also reduced the uplink sensitivity, the UEs need to increase their transmit power further when approaching the RDS cell. Moreover, as the uplink receivers are desensitized, the UE will have to maintain a relatively high transmission power even when it is within the RDS cell's coverage. The overall result is a further increase in the inter-cell interference generated by the RDS cell with desensitization, which consumes the load headroom in the system's scheduler as well as in neighbor cells, diminishing the overall system capacity.

The aim of present invention is to apply desensitisation only to a group of selected users, which helps reduce the increase in inter-cell interference.

Dynamic interference thresholds and Raise Over Thermal (RoT) estimation algorithms

The users within the RDS cell and terminals in the neighbour cells contribute to the interference level in the uplink of WCDMA systems. The load of a RDS cell is directly related to its interference level. Therefore, load should be kept below a certain level to keep the stability of a RDS cell. In WCDMA systems, load management is done via the scheduling of uplink channels to time intervals where the interference conditions are favourable.

Since the scheduling in LTE is not primarily interference controlled and since uplink orthogonalityy properties are different in WCDMA than in LTE, this section applies mainly to WCDMA systems. However, inter-cell interference will become more important in LTE HetNets, with increasing traffic and also due to RDS deployments. Therefore the ideas of this section may bear on load control in LTE.

Inter-cell interference is affected by desensitization but also by the amount of UEs using the system : the larger the number of UEs the more congested a cellular network is likely to be. Therefore, when a new UE tries to join the RDS, it is useful to know how congested the RDS is at that time so that the scheduling of users can be controlled.

In WCDMA scheduling of users, and in WCDMA power control stability monitoring with fast congestion control, a key variable known as Raise over Thermal (RoT) indicates the ratio between the total interference received on an antenna connector and the thermal noise level floor at that same point. Values of RoT can be used to determine when to apply desensitisation to a RDS. The raise over thermal is defined as:

RoT(t) = ^^ (1 ) where P_N {t) is the thermal noise power floor level and P_RTWP (t) is the total wideband power, both as measured at the antenna connector. P_RTWP (t) is defined as:

PRTWP (t) + P_N (t) (2)

=1 where is the radio link power received from neighboring cells and sources outside the RDS cell's coverage, P_t {t) is the power received from UEs within the RDS cell's coverage and P_N (t) is the thermal noise power floor level. RoT is estimated using algorithms, and one of the main problems of this calculation is how to separate the power contributions of thermal noise from those due to neighbor cell interference. In order to understand this fundamental problem, note that

^neighbor ( + W^— neighbor (t )]+ E[P_N (t)] + AP_naghbor (t) + AP_N (t) , (3)

where E[ ] denotes mathematical expectation and where Δ denotes the variation around the mathematical expectation.

As mentioned above, desensitisation can be implemented by adding noise to the uplink signal after digitalization in the radio unit. But it is not clear how much noise can be added. There are no measurements available in the radio unit related to the neighbor cell interference. Wigren et. al. , in their work presented in the I EEE 2007 Vehicular Technology Conference on the 'Estimation of uplink WCDMA load in a single RBS', have shown that, at most, a linear filtering operation can be used to estimate the sum

> ^{Dut tnis} estimate cannot be used to deduce the value of either £|Λ_τ( ] ^or E[P_neighbor (t)] . This issue has been analysed rigorously for the RoT estimation problem by Wigren on his 2009 article on "Soft uplink load estimation in WCDMA" for the IEEE Transactions on Vehicular Technology, and it has been proved that the noise power floor is not mathematically observable. Hence, RoT cannot be calculated, it can only be approximately estimated. Two algorithms for estimating RoT are briefly described below.

The sliding window noise floor estimation algorithm is depicted in Figure 1 and described in detail in Wigren's work presented in the IEEE 2007 Vehicular Technology Conference. This algorithm accurately estimates the thermal noise floor P_N (t) by computing a soft minimum over a relative long window in time. This estimation relies on the fact that the noise floor is constant over very long periods of time (disregarding the small temperature drift). The sliding window algorithm has the disadvantage of requiring a large amount of storage memory, which becomes particularly troublesome in case a large number of instances of the algorithm are needed. As T. Wigren showed in his 2010 article on "Recursive noise floor estimation in WCDMA" for the I EEE Transactions on Vehicular Technology, to reduce the memory consumption, a recursive noise floor estimation algorithm can be used, which reduces the memory requirements of the sliding window scheme discussed above at least by a factor of 100-1000. Moreover, there is a particularly relevant methodology, called the dynamic RoT methodology, which is based on the fact that some network cells can operate with larger load headroom than others. The basis of the method lays in the fact that in case inter-cell interference is high, a higher RoT may be used since the inter-cell interference is only weakly coupled to the inner power control loop stability of the serving cell. Interference generated by users in the own cell are much harder coupled to serving cell inner power control stability and in case the majority of the interference is caused by users in the own cell the RoT needs to be reduced. Dynamic RoT algorithms estimate maximum levels of RoT and then use these values to affect the levels of scheduling and the amount of UEs in the system. They do however not usually account for the origin of the interference, leading to a performance that is not optimal.

However, RoT has a non-linear functional relationship with the UE throughput. In particular, in regions with high RoT, the functional curve is very steep. This means that even modifications of several dBs in the maximum values of the RoTs often have negligible effects on the UEs throughput and inter-cell interference. Consequently, dynamic RoT cannot be effectively used to apply selective desensitization.

Inter-cell interference estimation

T. Wigren in WO2013/184063 refers to new high bandwidth inter-cell interference power estimation algorithms that can be used in WCDMA and LTE systems.

The WCDMA algorithm estimates the uplink neighbor cell interference, experienced in a cell as a result of all the surrounding cells. This estimation algorithm has a high bandwidth, almost at the sampling rate for enhanced uplink scheduling in WCDMA. It is also accurate, with inaccuracies of only 10-15% in the interesting region with high inter- cell interference power. See Figure 2.

The WCDMA algorithm is applicable to LTE systems with minor modifications. First, since scheduled grants must be used by UEs in LTE, while this is a UE choice in WCDMA, there is no need to estimate the load utilization in LTE. This means that the corresponding state of the WCDMA estimator needs to be removed. Secondly, in the WCDMA algorithm the nonlinear relation between the throughput and the RoT is exploited in a nonlinear measurement equation, which relates the measured P_RTWP {t) to the modeled value. In LTE this relation is replaced by a linear or close to linear relation. The consequence is that the estimation problem is in fact easier and less complex to solve online in LTE than in WCDMA.

So far, prior art static desensitization algorithms apply a fix desensitisation value to all the RDS's uplink receivers. On the other hand, prior art dynamic algorithms could, in principle, apply different desensitisation values to the uplink receivers. However, as dynamic algorithms operate in the region of high RoT values, their selective desensitisation would only have a minimum effect on the inter-cell interference. Therefore, it would be useful to improve the selective desensitization of the uplink receivers without unnecessarily increasing the inter-cell interference.

Summary

It is an object of the present invention to obviate some of the disadvantages described above and provide an improved selective desensitization of the uplink receivers without unnecessarily increasing the inter-cell interference.

According to a first aspect of the present invention, there is provided a method of controlling a de-sensitisation level, to be applied at an uplink radio receiver of a cell of a cellular radio communication network. The method comprises defining a high power inter-cell interference level (threshold_high), determining a first mean high inter-cell interference value ( ^ high, neighbour(M )) and sampling received signals and using them to obtain estimated inter-cell interference samples. The method further comprises deriving a time constant 3(t) based upon a fraction (fhigh(t)) of the estimated inter-cell interference samples that exceed the threshold_high. The method further comprises applying a linear filter, which employs the time constant (3(t)), to the first mean high inter-cell interference ( ^ high, neighbour(t-1 )) to determine a second mean high inter-cell interference ( ^ high, neighbour(t)). Finally, the method comprises using said the value of (

^ high, neighbour(t)) to control the de-sensitisation level. Using the defined value for high power inter-cell interference level (threshold _high), the method may comprise iteratively repeating all the other steps in sequence. Such that, in this sequence, the second mean high inter-cell interference of each completed iteration becomes the first mean high inter-cell interference for each new iteration.

The method may comprise deriving the fraction (fhigh(t)), of the estimated inter-cell interference samples that exceed the threshold _high, by applying a second linear filter to a previously determined fraction (f_high(t-1 )). This linear filter employs a further time constant a.

The linear filtering used in the method is performed with an auto- regressive linear filter.

The method may comprise evaluating the fraction (f_high(t)) by applying a second auto- regressive linear filter with the following mathematical expression:

ίπ (t ) = (1 - a)f_Ugh [t - l)+ aS(P_neighbor (i)) (4) where

_S(p i_t))= l^{1, NR}i W ^{> threshold}hi_gh

{ neighboA )) - ^_{Q NRi} (t) < threshold_high ^U and where NR_j (t) denotes the noise rise of the inter-cell interference:

and P_N (t) and P_ndghbor{t) are thermal noise power and high power inter-cell interference respectively, and P_d denotes a constant level of desensitization.

The method may comprise evaluating the time constant (3(t)) as follows:

I ° fhi_gh (^t)^≤threshoidfi

where f_hi_gh,min =threshold_p is a predefined minimum fraction.

The time constant (3(t)) then may be used in the first linear auto-regressive filter in the following mathematical expression:

igh,neighbor ( = (l ^~ (t )) igh,neighbor (' ^~ l) + β{ί ^neighbor (') NR, (t) > threshold _Ugh hi_Mhbor if) = (l -

it - 1) , (0 <= threshold _high (8)

The method may comprise using the value of ( ^ high, neighbour(t)) for determining the following noise rise measure:

P neighbor , high (t)

NR_{I gh} (t) = . Τ' (9)

N (t)+ P_d (t)

wherein the total desensitization is defined as P_d {t) = P_D + AP_d (t) and the additional desensitization power AP_d (t) is added according to a control algorithm.

The method may further comprise a control algorithm to calculate the additional desensitization AP_d (t) that is a static function of the estimated NR_{I high} (t) :

AP_d (t) = Y(NR_{i Mgh}(t))■ An example of such static function is a pre-specified function such as:

The method may also further comprise a control algorithm, to calculate said additional desensitization AP_d (t), that is a dynamic function such as x{t) = Mt -l), NR_IMgh {t - l))

(H i

AP_d {t) = h(x{t), NR_IMgh {t))

According to a second aspect of the present invention, there is provided an apparatus for controlling a de-sensitisation level to be applied at an uplink radio receiver of a cell of a cellular radio communication network. The apparatus comprises a memory configured to define a high power inter-cell interference level {threshold _high); a first processor configured to determine a first mean high inter-cell interference ( P high, neighbour(M )) and s. detector configured to sample received signals and use these to obtain estimated inter-cell interference samples. The first processor is further configured to derive a time constant (3(t)), which is based upon a fraction (f_high(t)) of the estimated inter-cell interference samples that exceed the high power inter-cell interference level and apply a linear filter, which employs (3(t)), to the first mean high inter-cell interference ( P high, neighbour(M )) to determine a second mean high inter-cell interference ( ^ high, neighbour(t)). The apparatus further comprises a second processor for using the second mean high inter-cell interference ( ^ high, neighbour(t)) to control the de-sensitisation level.

Using the defined value for high power inter-cell interference level (threshold_high), the first processor is configured to iteratively repeat all the procedures it is configured to perform in a sequence. Such that, in this sequence, the second mean high inter-cell interference of each completed iteration become the first mean high inter-cell interference for each new iteration.

The apparatus may comprise the first processor being configured to derive the fraction (fhigh(t)), of the estimated inter-cell interference samples that exceed the threshold_high, by applying a second linear filter to a previously determined fraction (f_high(t-1 )). This linear filter employs a further time constant a.

The linear filter used by the first processor is an auto-regressive linear filter.

The apparatus may comprise the first processor being configured to derive said fraction by evaluating the equation:

fw it) = (1 - )f {t - 1) + aS(P_neighbor {tj) (12) where

(13)

and where NR, (t) denotes the noise rise of the inter-cell interference:

and P_N {t) and P_neighbor {t) are thermal noise power and high power inter-cell interference respectively, and P_d denotes a constant level of desensitization.

The apparatus may comprise the first processor being configured to evaluate the time constant (3(t))as follows: f_high (^t ) > threshold _β

where

is a predefined minimum fraction.

The time constant (3(t)) may be used by the first processor to apply the first mentioned linear auto-regressive filter to evaluate the following equations:

Kig neighbor (' ) = (l NR, (t) > threshold _Μ≠

PhMhbor it) = (l -

NR, ( <= threshold (16)

The apparatus may comprise the second processor being configured to determine the following noise rise measure:

The apparatus may comprise the second processor being configured to estimate the thermal noise floor estimation before estimating the desensitization, and the second processor being further configured to add additional desensitization power AP_d (t) according to a control algorithm.

The apparatus may comprise the second processor being configured to apply a control algorithm, to calculate the additional desensitization AP_d (t), that is a static function of the estimated NR, _high(t) : AP_d (t) = Y{NR, _high{t)). An example of such static function is a pre-specified function such as:

The apparatus may comprise the second processor being configured to apply a control algorithm, to calculate the additional desensitization that is a dynamic function of the estimated NR_{I Mgh} (t) such as

x{t) = Mt - \), NR_IMgh {t - \))

AP_d (t) = h(x(t), NR_{I gh} (t))

The apparatus may comprise a plurality of radio heads connected to the uplink radio receiver, such that each radio head is configured to send and receive radio signals to user equipment.

Brief Description of Drawings

Figure 1 is a block diagram illustrating a baseline RoT estimation algorithm;

Figure 2 shows the rms inaccuracy of the neighbour cell interference estimate as a function of the average neighbour cell interference power level;

Figure 3 shows the effect of inter-cell interference from a macro radio base station (RBS) on an radio dot system (RDS) cell without the embodiment of the invention; Figure 3 shows the effect of inter-cell interference from a macro RBS on an RDS cell with the invention;

Figure 5 is a flow diagram illustrating a process carried out to calculate a desensitisation level;

Figure 6 shows the apparatus for controlling the de-sensitisation level in an uplink radio receiver.

Detailed Description of the Invention

The invention comprises means for estimation of a mean high-power inter-cell interference value, P_{hig neighbor}{t) . The quantity is related to the fraction of all time high- power samples that create "high" inter-cell interference, f_high (t) . These samples may be related to things such as UEs. The invention further comprises means that use the value of P_hi„ _nei„_hbor {t) to increase the desensitization of uplink receivers in a RDS cell. When a mean power value of inter-cell interference is 'high' in relation to the estimated thermal noise power floor value in the RDS's radio unit, the noise floor is momentarily increased to achieve desensitisation. To achieve a high high-power inter-cell interference value P_{hig neighbor} {t) , at least one UEs has to be transmitting at high power.

The momentary desensitisation of uplink receivers removes parts of the inter-cell interference from the scheduler's load headroom, thereby increasing the possible throughput of UEs signals significantly, as compared to a dynamic RoT algorithm.

A novel noise-raise variable, NR^t) , is introduced to estimate the amount of mean high-power inter-cell interference P_{hig neighbo} ) :

P (t)

NR_r (t) = r? ^boA ' . (20)

p_N {t) + p_d

NR^t) denotes the specific noise-raise of the inter-cell interference and P_N (t) and

P_neighbor {t) denote the estimates of thermal noise power floor and inter-cell interference respectively. Equation (20) is used self-consistently.

The quantity P_d denotes a constant level of desensitization, possibly applied at a point after thermal noise floor estimation in the receiver signal chain. The thermal noise floor estimator is hence not sensitive to P_d in such a case. In case the desensitization is applied before thermal noise power floor estimation, P_d = 0 .

Equation (20) needs estimates of:

(a) The fraction of all inter-cell interference power samples, f_high (t) , that are classified as "high power" and contributing to the inter-cell interference.

(b) f_{hi h} (t) in turn is then used to estimate mean high-power inter-cell interference value,

These estimations are done using standard linear filtering. In these models, the so- called 'time constant' of the filtering controls the filter bandwidth when measuring the change of a variable against a time-shifted version of itself. For instance, P_{hig neighbor}{t) against P_{hig neighbor}{i - l)■ If a small time constant used for the filtering, it means that P_{hig neighbor}{t) is allowed to change quicker than if a large time constant is used.

A problem now occurs since the mean of the high-power inter-cell interference is obviously only updated when the inter-cell interference is high. Since standard linear autoregressive filters are preferably used, the convergence time of the mean high power inter-cell interference estimate could become very high during times with low intensity traffic, simply because there would be few estimated high power inter-cell interference samples to process. This is where the estimate of the fraction of estimated inter-cell interference power samples that are classified as "high-power" comes into play.

The idea is to re-compute the autoregressive time constant for P_{hig neighbor}{t) , by normalizing it according to the current estimate of the fraction f_high(t) .

The linear autoregressive expression for f_high(t) is: fw (') = (!- a)f_Ugh [t - 1) + aS(P_neighbor {t)) (21 ) where

P (Λ) - ί ^NR, ^) > ^thresholdk_i&h

"* ^{j; "} [0 NR_I(t)≤threshold_hlgh - ^{ '

Here a is the time constant of the fractional update, and threshold_high is the limit above which an inter-cell interference power sample is classified as a high power one.

The value of f_high(t) is then used to calculate the time constant β( of the linear autoregressive expression used to estimate a mean high-power inter-cell interference value, P_{hi Knei hbor}{t) . The expression for β{ΐ) is:

^₌ Wf_high(t), f_high (t) > threshold _{β (23)}

I ⁰ f_high {t)≤ threshold _β

In case f_high {t) becomes smaller than the preconfigured limit of f_{hi h n} = threshold _β , it is replaced by 0 in (23). Note that 0 < β₀/ f_{high a} < 1. With available, the mean high inter-cell interference high,neighbor (i) can be calculated as:

P high,neighbor (t) = (i-p(t))P_hhigh,neighbor fi{t)P_neighbor (i) , NR, (t) > threshold _high

P high, neighbor (t) = (i-p(t))P_hhigh,neighbor (i - l) , NR, {t) <= threshold _{hi h}

Note that (23) guarantees that P_{hig neighbor}{t) is never updated in case f_high (t) is small, and hence, unreliable.

The converged value for P_{hig neighbor}{t) can then be used to estimate a still more specific noise-over-rise variable:

which in turn can be used to estimate the load at the antenna connector and the corresponding increase in the desensitization level that needs to be applied to balance the RDS coverage.

Desensitization located after the thermal noise floor estimation

In this case the estimation of the thermal noise floor P_N (t) is located before the desensitization functionality. Hence, the value of P_N (t) is not sensitive to any desensitization value, i.e. the nominal desensitization is represented by P_d in (20). The newly calculated level of desensitization, AP_d (t) , will be added to the constant level of desensitization P_d using a control algorithm. The total desensitization is defined by the sum P_d (t) = P_d + AP_d (t) at each time. Hence, there is no integration involved.

The simplest control algorithm is to compute AP_d (t) as a static function of the estimated NR_{I high} {t) :

AP_d (t) = _r{NR_{I h} (t)) (26) where γ(·) is a pre-specified function. An example embodiment is . (27)

Equation (27) makes sure that the additional desensitization power is constrained to a pre-configurable interval. Note also that AP_d (t) is not to be included in (20) and (25), which would immediately counter the effect of the additional desensitization AP_d (t) .

Since desensitization is added after the thermal noise power estimator, it is essential that AP_d (t) is not included in the thermal noise power floor that is used to compute

P_neighbor{t) . If it were, the effect of the increase provided by AP_d (t) would be cancelled in (20) immediately and in (25) after.

The desensitization control can also be made dynamic. A very general way of describing this, that includes linear dynamic control as a special case, is to compute the additional desensitization with a nonlinear dynamic feed forward controller according to

x(t) = f {x(t -l), NR_{! gh} (t -l))

AP_d {t) = h{x{t),NR_{I gh}{t)). (28)

Desensitization located before the thermal noise floor estimation

In this case the estimation of thermal noise floor P_N(t) is done after the desensitization functionality. P_N{t) is then sensitive to the constant desensitization. Even though the additional desensitization power AP_d (t) is added before the thermal noise floor estimation, the long time constant applied together with minimum estimation means that the thermal noise power floor estimation will not be sensitive to AP_d (t) . The ideas used when the desensitization is located after the thermal noise floor estimation will then be applicable with obvious modifications. The estimation of AP_d (t) can also be used to redefine RoT and then use it for fast congestion control and scheduling:

RnT (A— P tOQ\

Here P_d = 0 in case the desensitization is located before the thermal noise power floor estimator. Note again that AP_d (t) shall not be fed into equations (20) and (25).

The approach described here provides an optimal, minimum, level of desensitization to a RDS. In this way, the RDS can reduce its levels of inter-cell interference, while maintaining its uplink performance irrespective of any incoming neighbor cell interference. The approach therefore contributes to enhancing the general mobile broadband performance in a RDS. This enhancement is illustrated in Figures 3 and 4, which show the effect of the approach in the scheduling headroom.

The scheduler is the entity responsible for controlling which system users get what data rate at which time. This is a balance between permitting high UE transmission powers, meaning a high data transmission rates, and interference. If the interference level is very high, some transmissions in the cell, or uplink transmissions, may not be received properly. On the other hand, if the interference level is too low the full system capacity is unlikely to be used. The scheduler gives users the highest possible data rate that does not exceed the maximum tolerable interference level in the cell. Both inter-cell interference and intra-cell interference are taken into account by the scheduler when allowing a UE to transmit at a high data rate. The amount of common uplink resources used depends on the data rate being used. For the higher data rate, it requires larger transmission power and thus the higher headroom consumption.

Figure 3 illustrates the problem that is addressed by the approach described here. A graph plots UE position against RDS uplink (UL) power. Shown on the plot are a Macro RBS cell, the uplink coverage of a RDS cell (RDS UL cell), and the downlink coverage area of the RDS cell (RDS DL cell). The plot shows that, at some position between the RDS DL cell extent and the RDS UL cell extent, the presence of the UE will consume the scheduling headroom of the RDS cell before the handover. On the other hand, Figure 4 illustrates what happens when the approach of the invention described above is applied. The scheduling headroom is maintained despite encroachment of the UE into the RDS UL cell. Scheduling headroom is maintained substantially up to the point when the encroaching UE performs a soft handover from the RBS cell to the RDS cell.

Figure 5 is a flow diagram illustrating the method described above. It illustrates the steps of:

a) defining a high power inter-cell interference level (threshold_high);

b) determining a first mean high inter-cell interference ( P high, neighbour(M )); c) sampling received signals and using these to obtain estimated inter-cell interference samples;

d) deriving a time constant ((3(t))) based upon a fraction (f_hi_gh(t)) of the estimated inter-cell interference samples that exceed said high power inter-cell interference level;

e) applying a linear filter to said first mean high inter-cell interference, wherein the filter employs said time constant, in order to determine a second mean high inter-cell interference ( P high. neighbour(t)); and

f) using said second mean high inter-cell interference to control said de- sensitisation level.

Figure 6 is a block diagram illustrating an apparatus for implementing the method described above. The apparatus comprises a memory that defines a high power inter- cell interference level (threshold_high); a first processor, which is used to determine a first mean high inter-cell interference ( ^ high, neighbour(M )) and a detector, which samples received signals and use them to obtain estimated inter-cell interference samples. This first processor is also used to derive a time constant (3(t)). The derivation is based upon a fraction (fhigh(t)) of the estimated inter-cell interference samples that exceed the high power inter-cell interference level. Moreover, the first processor can apply a linear filter to the first mean high inter-cell interference. This filter employs the time constant

(3(t)) to determine a second mean high inter-cell interference ( ^ high, neighbour(t)). Finally, the apparatus also comprises a second processor, which uses the second mean high inter-cell interference to control the de-sensitisation level. It will be appreciated by the person of skill in the art that various modifications may be made to the above described embodiments without departing from the scope of the present invention.

Claims

CLAI MS :

1 . A method of controlling a de-sensitisation level to be applied at an uplink radio receiver of a cell of a cellular radio communication network, the method comprising:

a) defining a high power inter-cell interference level (threshold _high); b) determining a first mean high inter-cell interference ( P high, neighbour(t-

1 )) ;

c) sampling received signals and using these to obtain estimated inter- cell interference samples;

d) deriving a time constant ((3(t))) based upon a fraction (f _hi_gh(t)) of the estimated inter-cell interference samples that exceed said high power inter-cell interference level;

e) applying a linear filter to said first mean high inter-cell interference, wherein the filter employs said time constant, in order to determine a second mean high inter-cell interference ( P high. neighbour(t)); and f) using said second mean high inter-cell interference to control said de-sensitisation level.

2. A method according to claim 1 and comprising iteratively repeating steps c) to f), where the second mean high inter-cell interference of each completed iteration become the first mean high inter-cell interference for each new iteration.

3. A method according to claim 1 or 2 and comprising deriving said fraction by applying a second linear filter to a previously determined fraction (f_hi_gh(t-1 )), wherein the filter employs a further time constant a.

4 A method according to any of the preceding claims where said linear filtering is performed with an auto- regressive linear filter

5. A method according to claim 3, wherein said step of deriving said fraction comprises evaluating the equation:

where N#₇ threshold _hhigh

^V ^' , n neeiigghhbboorr V )) ^~ f ,\ „1„„ Ί _J

0 NR, {t) < threshold high and where N ^{i V}R"^^/ t^J )/ denotes the noise rise of the inter-cell interference:

and ^N ^ ' and neighbor \ / _are thermal noise power and high power inter-cell interference respectively, and ^d denotes a constant level of desensitization.

6. A method according to any one of the preceding claims, where the time constant ((3(t))) is evaluated as follows: β(ή = ^{fhish fhish} ^ ^{> threshold}fi

I ° fhi_gh (^t)^≤threshoidfi

where f_hi_gh,min =threshold_p is a predefined minimum fraction.

7. A method according to claim 6, wherein said first mentioned linear filter comprises an evaluation of the followin e uations:

igh,neighbor (' ) = (l ^~ + - NR, (i) > threshold _u≠

Kig neighbor (0 = (l ^" high, neighbor (' ^" l) . NR, (t) <= threshold ^

8. A method according to any one of the preceding claims, wherein step f) comprises determining the following noise rise measure:

(Λ P neighbor, high (t)

^NKi,high \t I - * , ——

9. A method according to claim 8, wherein the thermal noise floor estimation is located before the desensitization functionality, and step f) comprises adding additional desensitization power AP_d (t) according to a control algorithm.

10. A method according to claim 9, wherein the control algorithm is a static function of the noise raise measure

AP_d (t) = y{NR_{I gh} (t))

11 . A method according to claim 10, wherein the static function of the noise raise measure is:

0, NR_{I gh} {t) < AP_d^

Δ¾( = NR_{I h} - AP_{d n} AP_{d n}≤ NR_{I high} (i) < AP^

12. A method according to claim 9, wherein the control algorithm is a dynamic function of the noise raise measure

x{t) = f{x{t - \), NR_IMgh {t - \))

AP_d {t) = h(x{t), NR_{I gh} {t))

13. A method according to any one of the preceding claims, wherein said cell comprises a plurality of radio heads connected to said uplink radio receiver, and said step of sampling comprises sampling signals received from the plurality of radio heads.

14. Apparatus for controlling a de-sensitisation level to be applied at an uplink radio receiver of a cell of a cellular radio communication network, the apparatus comprising: a memory (2) configured to define a high power inter-cell interference level

{threshold _high);

a first processor (3) configured to determine a first mean high inter-cell interference ( P _{high) n}ei_ghbour(t-1 ));

a detector (4) configured to sample received signals and use these to obtain estimated inter-cell interference samples;

said first processor being further configured to

derive a time constant (3(t)) based upon a fraction (fhigh(t)) of the estimated inter-cell interference samples that exceed said high power inter-cell interference level; and

apply a linear filter to said first mean high inter-cell interference, wherein the filter employs said time constant, in order to determine a second mean high inter-cell interference ( P high, neighbour(t)); and

a second processor (5) for using said second mean high inter-cell interference to control said de-sensitisation level.

15. An apparatus according to claim 14, said first processor being configured to iteratively repeat all the procedures it is configured to perform, where the second mean high inter-cell interference of each completed iteration becomes the first mean high inter-cell interference for each new iteration.

16. An apparatus according to claim 14 or 15, said first processor being configured to derive said fraction by applying a second linear filter to a previously determined fraction (f_hi_gh(t-1 )), wherein the filter employs a further time constant a.

17. An apparatus according to any of the preceding claims, wherein said first processor is configured to apply a linear filter that is an auto-regressive linear filter.

18. An apparatus according to claim 16, wherein the first processor is configured to derive said fraction by evaluating the equation:

f_high (t) = (1 - a)f_high [t - 1) + d{p_ndghbor {t))

where

I , , Λ [l, NR_j (t) > threshold _hhigh

nneeiigghhbboorr \" J / ~ \ 0 n. NR_{ (t)≤ thresh.o„l(dJ

^ _hhigh and where ^{J Vi / i} ; denotes the noise rise of the inter-cell interference:

and

are thermal noise power and high power inter-cell interference respectively, and ^d denotes a constant level of desensitization.

19. An apparatus according to any one of the preceding claims, wherein the first processor is configured to evaluate the time constant (3(t)) as follows: >^thresholdfi

^≤threshoidfi

where

is a predefined minimum fraction.

20. An apparatus according to claim 19, wherein the first processor is configured to apply said first mentioned linear filter by evaluating of the following equations:

(i) > threshold _Ugh igh,neigh_bor (') = (l ^~ fiW igh.neighbor (' ^~ l) . (0 <= threshold _high

21. An apparatus according to any one of the preceding claims, wherein the second processor is configured to determine the following noise rise measure:

22. An apparatus according to claim 21 , wherein the second processor is configured to estimate the thermal noise floor estimation before estimating the desensitization, and the second processor is further comprised to add additional desensitization power AP_d (t) according to a control algorithm.

23. An apparatus according to claim 22, wherein the second processor is configured to apply a control algorithm that is a static function of the noise raise measure

AP_d (t) = 7{NR_IMgh (t))

24. An apparatus according to claim 23, wherein the expression for the static function of the noise raise measure is:

Δ^( = NR_r

Δ NR_r (t) > Δ d,max

25. An apparatus according to claim 22, wherein the control algorithm is a dynamic function of the noise raise measure

x{t) = f{x{t - l), NR_{I h} {t - l))

AP_d (t) = h(x(t), NR_{I gh} (t))

26. An apparatus according to any one of claims 14 to 25, wherein said cell comprises a plurality of radio heads connected to said uplink radio receiver, wherein each radio head is configured to send and receive radio signals to user equipment.