GB2341297A

GB2341297A - Mitigating interference in radio communications systems

Info

Publication number: GB2341297A
Application number: GB9906388A
Authority: GB
Inventors: Anthony Peter Hulbert
Original assignee: Roke Manor Research Ltd
Current assignee: Roke Manor Research Ltd
Priority date: 1998-09-02
Filing date: 1999-03-22
Publication date: 2000-03-08
Anticipated expiration: 2019-03-22
Also published as: GB2341297B; EP0984583B1; DE69913273D1; GB9906388D0; EP0984583A2; EP0984583A3

Abstract

A communications system incorporates regular burst format forward error correction (FEC) coded data in at least two transmitters which are synchronised in such a way that transmission bursts are nominally non overlapped in time, but wherein the occurrence of a brief burst overlap period 5, at the location of at least one remote receiver, produced due to propagation delays, is mitigated at least partially by means incorporated in the receiver for estimating information loss associated with the overlap, which estimate is used to assist in the decoding of the FEC code. Application is to Time Division Duplex (TDD) radio communication systems with or without a Time Division Multiple Access (TDMA) component. The technique is called 'Blind Interference Edge Detection', and is applicable in systems with or without a direct sequence spread spectrum component.

Description

1 2341297 IMPROVEMENTS IN OR RELATING RADIO COMMUNICATION SYSTEMS This

invention relates to radio communication svstems and more especially it relates to Time Division Duplex (TDD) radio communication systems with or without a Time Division Multiple Access (TDMA) component.

In such systems, which are especially suitable for use in mobile communication systems comprising a plurality of mobile terminals in communication with a base station, it is desirable to maximise the time of transmission within a time slot and in this connection attention is directed to our co-pending Patent Application F21448 wherein elimination of the guard time associated with the round trip delay is proposed in a system incorporating a TTD element, thereby aligning the absolute timings of the base and mobile stations(s).

Any loss of communication time is compensated by rate matching, for example, declaring the lost part of the signals as erasures and inputting these into the forward error correction decoding algorithm, after deinterleaving. In this way, the loss of information is compensated, with the only effect being the loss of a small amount of signal power.

Whilst being a significant advance on the prior art, the above proposed scheme has some difficulties. The overlapped transmission/ reception periods are always declared as erasures. Although such an overlap may result in lost transmission capability, this is not necessarily the case. For example, a mobile may not interfere with another mobile if they happen to be widely spaced apart. Furthermore, the process cannot cope with unforeseeable interference relationships. These can arise, for example, in a cellular deployment with synchronised base stations. If a terminal in one cell uses a time slot and a terminal in a second cell uses the following time slot, then depending upon the phylsical distances involved, interference can arise between the edges of the time slots.

According to the present invention a communications system is provided which incorporates regular burst format forward error correction (FEC) coded data in at least two transmitters which are s,,-nchronised in such a way, that transmission bursts are nominally' non overlapped in time, but wherein the occurance of a brief burst overlap period, at the location of at least one remote recei-x 7 er, produced due to propagation delays, is mitigated at least partially by. means incorporated in the receiver for estimating information loss associated with the overlap, which estimate is used to assist in the decoding of the FEC code.

The said information loss estimate may take the form of an estimate of the period of overlap, soft received coded data bits over said estimated period of overlap being declared as erasures by setting their numerical value to zero prior to FEC decoding.

Alternatively the said information loss estimate may be derived from of an estimate of the period of overlap and an estimate of the level of interference over the period of overlap the said estimate of the level of interference over the period of overlap being compared with an estimate of the signal power such that if the signal to interference ratio is smaller than a predetermined value then soft received coded data bits over said estimated period of overlap are declared as erasures by setting their numerical value to zero prior to FEC decoding.

3 As a further alternative the said information loss estimate is derived from an estimate of the period of overlap and an estimate of the level of interference over the period of overlap, said estimate of the level of interference over the period of overlap being used in combination with an estimate of the signal power in order to generate a weighting value which is used to scale received soft coded data bits.

One method of deriving the estimate of the period of overlap may comprise the following steps utilising parameters as hereinafter defined; a) Generate sets of measurement samples of received signal power across the period of the burst from one or more received bursts and average said sets of measurement samples across said bursts to produce a set of average signal power measurement samples, Pi, forl<i:5ns, b) Define the maximum number of measurement samples over which the received burst could be subject to transmitter overlap generated interference, nn,,, c) Compute A? - 1 n, Pi n, -n. - 2 i =n., d) Compute v Pi - n,. A, 1 e) Compute a series of metrics, Mi, where Mo = A and Mi=aili, g K, en that a i = P i - /2 and ai= ai - 1 + Pi fori:s i:5 nmax, f) Select the maximum value of Mi forO s i <nnia.v. The associated index, i, is the MAP estimate of nint, hint. The estimate of the level of interference over the period of overlap required may be obtained by using a current unaveraged set of measurement samples of received signal power 4 1 it- across the period of a burst, e' to for m P = -, P"i which is an fit pit estimate of the total received power over the estimated period of 1 P? N interference and forming PR= jl to give an estimate of ns - hint i -li- -I the received signal plus noise and to facilitate provision of the interference level estimate.

It will be appreciated that a system according to the invention is especially suitable for TDD mobile radio applications for example.

One embodiment of the invention will now be described with reference to the accompanying drawings in which; Figure I is a timing diagram which illustrates the timing relationship between the transmission and reception signal periods of a mobile terminal and, Figure 2 is a diagram which illustrates the relationship wanted signal and interference.

Referring now to the drawings, Figure I shows part of a TDD/TDMA frame, and more specifically it shows the last two downlink slots 1, 2, followed by the first two uplink slots 3, 4, of the frame. It can be seen that the first and last slots 1 and 4 respectively as shown, are communicated completely without anN, interference. However, the mobile terminal receiving the second downlink slot may receive significant interference from the mobile transmitting the first uplink slot, depending on their relative positions and in Figure I this is in effect illustrated by overlap regions 5.

The present invention uses, in effect, a technique which is herein aptly described as 'Blind Interference Edge Detection', in the various receivers to mitigate the above mentioned effects.

Blind Interference Edge Detection may be used in any system where parts of the received signal burst may be lost due to interference and where apriori information exists concerning the approximate position of the parts lost. The approach is applicable in systems with or without a direct sequence spread spectrum component.

In one embodiment of the present invention, as hereinafter described, no direct spread spectrum component is included. A transmission burst may have interference from the beginning of the burst, and continuing into the burst for a period which is unknown but which never exceeds a certain maximum percentage of the burst (e.g. 20%). The situation is illustrated in Figure 2, where Mear is the maximum part of the frame in which it can be guaranteed that interference will not arise, i.e. Tint < Tslot Tclear. The value of Mear can be determined beforehand from system design consideration. The variables, PN, Ps, Pi and Tint are all unknown.

If Tint can be successfully estimated then all of data received over this period can be declared as erasures. For example, the values of the corresponding decision variables are set numerically equal to zero.

In an alternative embodiment of the present invention the optimum procedure for demodulating the signal is to weight all of the decision variables according to their signal amplitude to noise standard deviation ratio. Thus, for the above situation, the optimum receiver would weight the decision variables over the 1 - -PS period Tint by, a factor Vj57- and the remaining decision N +P] variables b-.,,, the factor f_ Alternatively and equi-v-alently, the P-1 1 1 F77, - N 6 decision variables can be weighted over the period Tint by:P,:, iP+1 -P, and leave the others unweighted. Thus, in the ideal case, RV, PY, and Tint would be estimated in order to perform this weighting.

A method for obtaining these estimates is described as follows. The first stage is to obtain an estimate Of PN + PS= PR. This is done by averaging the total received power over the period Tclear. The next stage is to estimate Ps. This can be performed by averaging the modulus of the decision variables. This is equivalent to multiplying by hard decisions, providing a biased estimate of the signal amplitude. The bias arises because of the errors in those hard decisions. It may be that the bias is acceptably small. PN is obtained by subtracting the estimate of PS from the original power estimate.

If the bias is too great, the alternative is to use a signal to noise ratio detector. The principles for this are laid out in the document "A. Signal to Noise Monitoring" by C E Ghilchriest - JPL 7 Space Programs Summary, No 3'1 -27, Vol. IV, pp 169-184 which is included here by reference. In this case, the noise power, PN is given by dividing the total power estimate by one plus the signal to noise ratio.

The maximum a posteriori (MAP) estimate, a preferred method for estimating Tint and Pi is formed by several operations as follows:1) Samples of the signal over the transmit period are collected. 2) The power of the samples is computed. If the samples are generated at complex baseband the power will be obtained by I- I forming 12 + 0 W- - 7 3) Optionally the sample power measurementsvOll be used to create an exponentially weighted average over the corresponding samples from previous frames. This has two benefits. Firstly it improves the accuracy of the measurements. Secondly, because of the Central Limit Theorem it leads to samples which have a distribution approximated by the Gaussian distribution. Working from these averages allows the period of interference to be computed more accurately. It cannot be used to derive an accurate measure of the interfering power in a given frame since this will vary from frame to frame.

In order to develop the MAP analysis we make the further assumption that the variance of the samples with interference is the same as that of the samples without interference. This greatly simplifies the analysis and algorithms based on this assumption have been found to work well in practice.

For the NIAP estimation process we jointly estimate the period and average level of the interference. As stated earlier, the latter is not of direct value but is an inevitable by-product of estimating the period.

We have previously introduced the burst periods in terms of time. For a practical implementation the signal will be sampled so that we will henceforth refer to the periods in terms of number of sample intervals T.

We thus estimate the total power with and without interference. Suppose there are ns samples, Pi frame by frame averaged power measurements across a frame (1:5i:5ns). Thus Tslot = ns. T.

According to the Gaussian assumption the probability density function of the samples is given by, 8 (pi - P7.

P(Pi) (Y.J-2-,T exp 2 (12 where cj is the unknown standard deviation and p, j = 1 P,...i:5n,,t PR-1>ni.t where nin t is the number of samples with interference, PR= RN-+ PS is the total power when there is no interference present and P.A= PR+ PI is the total power when there is interference present.

We first form 1 P. This provides an estimate of P.A nint +PR(ns- nint). --1 Next we form an estimate of PR, AR by, measuring forming, 1 17S M' - i= n,, - where nrnax is the maximum number of samples for which overlapped interference may, be present. Thus, Tclear = T(ns-nmax) so that nniax=ns -TclearIT 11 Together these provide a new quantity Pi - 11, R 1 Now y provides an estimate of (PA- PR) nin t= PI nint The probability, of a given set of average power measurements is, ns 1 -- exp (Pi -P - T i)2 cy,,r2,-r 2cr' Taking the logarithm and ignoring all terms which are independent of nin t and PI we form a metric 1 (R - {A+ +2 + 1 (M - A)2 where n is a hypothesis for nint and A is an estimate of Pi Now if nin t =n, then the best estimate we have for Pi is n 9 n v 2 n 2 Thus our metric becomes (A - L- + A,) + 2 (pi - k) {n 01 2My Y 2V 2 This simplifies to 1 + + PR + (Pi - R) 1-1 n n n i-n+l The latter summation is independent of n and so can be removed from the metric. We thus have:- 2 Pi 2PR 7j---+ + 1-1 n n n The term y outside the expression is constant and so can be removed. The metric then becomes, M= 2PR _ 2 n pI + Y n i-, n Because k is constant we form the metric M-2 PR. Then, scaling and dividing by two our final metric is, n M-2PR M 2 =---+ PR which we can now maximise.

2 2 n For the case n=0 the first summation disappears so N1=0 Our metric is therefore h.

Thus we have a complete procedure for forming our MAP estimate. The steps are as follows:- 1. Generate negative exponentially weighted samples, Pi.

n, 2. Compute PR lpi n, - n.

n, 3. Compute Pi - ns R 4. Compute a series of metrics, Mi, where A10 = PR and Ali ==aili, g and ai= ai-,+Pi given that a i = P i - /2 5. Select the maximum value of Ali forO:,- i < nclear. The associated index, i, is the MAP estimate of nint.

When computing,/, the estimate Of PR removes most of variance associated with the samples from nrnax to ns. However, none of the variance is removed from the contribution of samples from nint+1 to nmax. For this reason it is desirable that the value used for nma-,. should be as close to nint as possible. In fact each estimate of nmax can be used to generate a close value for nmax. The criteria for updating nmax from the estimates of nint are as follows:1. The estimates of nint will vary about the actual value. The I updated nrnay must be higher than the estimated nint by a margin large enough to encompass the actual nint with high probability. ? As the receiving mobile and/or the mobile interfering with it change location, the,,,alue of nint will alter. Additional margin is required to allow for this change. In practice, this requirement is likely to be small compared with the need to accommodate inaccuracies in the estimation of nin t.

One possibility for updating nrnax is to form a negative exponential average of the estimates of nint, tTint, and make nrnax = f x n int, Also, nmax should have a minimum value to avoid the algorithm becoming blocked when nint =o. Alternatively we could use nma-, =h + nniargain or any suitable mapping function.

Once h int is obtained, it can be used on the current frame (i.e. not using the exponentially averaged samples) to compute VA 11 In this case we compute R, = 1 An, hint P"i., where all the terms now refer to the samples for the individual frame, P,' rather than the average.

Similarly Pi ni.,,i As before we can now determine the mean variance of the signal component to compute the optimum metric weighting. It may be, of course, that Pi is strong enough with respect to RR that the only, practical solution is to erase the data. As the wanted and interfering signals fade up and down independently, the relative levels of Pi and PR will alter significantly whilst nint will remain substantially constant.

Thus one solution would be, on a per frame basis, to check the relative levels of Pi and PR, erasing only where necessary-.

Once all of the estimates are available the appropriate weightings can be applied and the burst decoded. As will be appreciated by those skilled in the art, the same procedure can be used to handle interference appearing at the end of the frame. As will be also appreciated, different levels of interference can be handled at both end of the burst. For other applications, similar procedures can be applied for interference bursts within a burst.

In the case of spread spectrum, the processing gain assists the measurement of the wanted signal power. However, the fundamental procedure is the same. With automatic gain control (AGC) operating on the received signal envelope, the variations in level will be translated into inverse variations in the wanted signal level. This will inherently achieve the required weighting. However, the AGC loop bandwidth should be optimised for the best trade-off between detecting the change in 1 12 interference level in a timely, m-,ay.,, on the one hand, and averaging the setting of the level on the other.

1 1 13

Claims

1. A communications system which incorporates regular burst format forward error correction (FEC) coded data in at least two transmitters which are synchronised in such a way that transmission bursts are nominally non overlapped in time, but wherein the occurance of a brief burst overlap period, at the location of at least one remote receiver, produced due to propagation delays, is mitigated at least partially by means incorporated in the receiver for estimating information loss associated with the overlap, which estimate is used to assist in decoding of the FEC coded data.

2. A communications system as claimed in Claim 1, wherein the said information loss estimate takes the form of an estimate of the period of overlap and wherein soft received coded data bits over said estimated period of overlap are declared as erasures by setting their numerical value to zero prior to FEC decoding.

3. A communications system as claimed in Claim 1, wherein the said information loss estimate takes the form of an estimate of the period of overlap and an estimate of the level of interference over the period of overlap and wherein said estimate of the level of interference over the period of overlap is compared with an estimate of the signal power and if the signal to interference ratio is smaller than a predetermined value then soft received coded data bits over said estimated period of overlap are declared as erasures by setting their numerical value to zero prior to FEC decoding.

14 4. A communications system as claimed in Claim 1, wherein the said information loss estimate takes the form of an estimate of the period of overlap and an estimate of the level of interference over the period of overlap and wherein said estimate of the level of interference over the period of overlap is used in combination with an estimate of the signal power in order to generate a weighting value which is used to scale received soft coded data bits.

5. A communications system as claimed in Claim 2, Claim 3, or Claim 4, in which the estimate of the period of overlap is generated by means of a method comprising the following steps utilising parameters as hereinbefore defined; a) Generate sets of measurement samples of received signal power across the period of the burst from one or more received bursts and average said sets of measurement samples across said bursts to produce a set of average signal power measurement samples, Pi., forl:5i5ns, b) Define the maximum number of measurement samples over which the received burst could be subject to transmitter overlap generated interference, nm,,, c) Compute PR - 1 ns Pi n, - n.

n' d) Compute Pi - 17, e) Compute a series of metrics, Ali, where MO and Ali=aili, g ive n that a i = P i - /2 and ai= ai - 1 + Pi for 15 i:5 nmax, 1 f) Select the maximum value of Mi forO:5 i < nmax. The associated index, i, is the NIAP estimate of nint, h int.

6. A system as in Claim 3, or Claim 4, or Claim 5 as it depends from Claim 3 or Claim 4, wherein the estimate of the level of interference over the period of overlap required is obtained by using a current unaveraged set of measurement samples of received signal power across the period of the 1 i;i_ burst, t' to form P, = YjY'i. which is an estimate of the total hint 1-1 received power over the estimated period of interference and 1 ns AI = I P,' to give an estimate of the received signal, plus n, - hit i-ttifit+l noise and to allow an interference level estimate, A communication system as claimed in any preceding Claim comprising a plurality of mobile terminals arranged in TDD communication with a base station.

8. A communication system as claimed in Claim 1 and substantially as hereinbefore described with reference to the accompanying drawings