GB2412045A - Iterative interference reduction in multicarrier CDMA systems - Google Patents

Abstract

A multicarrier Code Division Multiple Access (MC-CDMA) receiver forms an estimated interference signal <EMI ID=2.1 HE=6 WI=3 LX=583 LY=732 TI=UI> <PC>by estimating the transmitted signal <B>Z,</B> processing this (100-130) to give a hard or soft output value, reprocessing (140-170) this to give a new estimate <EMI ID=2.2 HE=6 WI=5 LX=1599 LY=847 TI=UI> <PC>and subtracting this from <B>Z.</B> The interference signal is then itself subtracted, after multiplication with channel estimate <EMI ID=2.4 HE=6 WI=5 LX=643 LY=1046 TI=UI> <PC>from the received signal <B>Y</B> to give <EMI ID=2.3 HE=6 WI=5 LX=1280 LY=1043 TI=UI> <PC>This is then combined at 60 according to a scheme such as Minimum Mean Square Error Combining (MMSEC). When enough iterations have occurred that little more improvement is possible, a different combination technique (eg. Maximal Ratio Combining, MRC) may be switched to via option selector 70. Such an iterative process can remove known and unknown interference, and support multimedia, multiuser service with multicode transmission as in Fig. 4.

Description

241 2045 ) 1 Signal Processing in Multi-Carrier CDMA The present invention

relates to a method of, and apparatus for, processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system.

Preferred embodiments relate to the improving of the quality of such a received signal, and can be used for removing known/unknown interference. The invention finds particular application in a mobile multi-carrier system. It is most suitable for downlink transmissions (from a base station to a mobile station), but can also be used for uplink transmissions.

Various diversity reception techniques exist, such as Orthogonal Restore Combining (ORC), Equal Gain Combining (EGC), Maximal Ratio Combining (MRC) and Minimum Mean Square Error Combining (MMSEC). These can be used for a multi- carrier system to resolve a frequency selective fading channel.

Fig. la demonstrates that, in a single user case, MRC outperforms other diversity reception techniques due to the weights of each sub-carrier with respect to its Signal-to- Noise Ratio (SNR). However, in a multiuser system, frequency selective fading could compromise the orthogonality between each user and result in performance degradation, as shown in Fig. lb. Loss of orthogonality may occur in the frequency domain (for MC- CDMA) or time domain (for MC/DS-CDMA) or both (in VSF-OFCDM) and will result in poor performance in Multiple Access Interference (MAI). Loss of orthogonality will occur if the width of the spreading is greater than the coherence bandwidth (in frequency) or coherence time (in time).

Simulation results have indicated that the amplification of certain (lowpower) subcarriers in ORC results in noise-enhancement.

MMSEC has a superior performance in Multiple Access Interference (MAN) as it accounts for the noise in the system. However, it still cannot reach the single user performance due to Inter-Carrier Interference (ICI) and Multiple Access Interference (MAI).

The following publications are hereby referred to: [ref l] N. Yee, J. Linnartz and G. Fettweis, "Multi-carrier CDMA in indoor wireless radio networks," Conference Processing PIMRC '93, Yokohama, pplO9-113, Sept. 1993.

[ref 2] N. Maeda, "Pilot channel assisted MMSE combining in forward link for broadband OFCDM packet wireless access," IEICE Trans. Fund, Vol. ES5AA, No.7, ppl635-1646, July2002.

The present invention has been made in the light of the problems described above.

It is an object of at least the preferred embodiments of the present invention to address these problems.

One or more aspects of the invention is / are set out in the independent claim(s).

Apparatus aspects corresponding to method aspects disclosed herein are also provided, and vice versa.

In one aspect the present invention provides a method of processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: a) forming an estimate of the transmitted signal from the received signal; b) de-spreading, demodulating, de-interleaving and decoding the estimate of the transmitted signal so as to provide an output signal; c) re-encoding, re-interleaving, re-modulating and re-spreading a signal which is based on the estimate of the transmitted signal after de-spreading, de- modulating, deinterleaving and decoding, so as to provide a new estimate of the transmitted signal; and d) forming an estimated interference signal based on the estimate of the transmitted signal and the new estimate of the transmitted signal.

In some applications, forming an estimated interference signal may be of useperse, e.g. the estimated interference signal could be used to determine how much interference a particular system has, perhaps so as to test/optimize transmission conditions.

However, in the most useful application the estimated interference signal is used to cancel or remove (at least some of) the interference contained in the received signal.

Hence, preferably, the method further comprises forming a refined estimate of the transmitted signal based on the received signal and the estimate of the interference, for example by subtracting the estimate of the interference from the received signal.

Even more efficient interference cancellation can be achieved by performing at least one iteration of: cl) re-encoding, re-interleaving, re-modulating and re-spreading a signal which is based on the refined estimate of the transmitted signal after de- spreading, de- modulating, de-interleaving and decoding, so as to provide a new refined estimate of the transmitted signal; dl) forming a refined estimate of the interference based on the refined estimate of the transmitted signal and the new refined estimate of the transmitted signal; and a2) forming a (further) refined estimate of the transmitted signal based on the received signal and the refined estimate of the interference.

Such an iterative process may be used to suppress known and unknown interference in a MC-CDMA receiver. It may improve the frequency diversity gain and support more multiusers for system level than would be possible without the iterative interference cancellation process. Further, it may be possible to support multimedia service with multicode transmission.

A

In another aspect the present invention provides a method of processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: forming an estimate of the transmitted signal from the received signal; and at least two iterations of: - forming an estimated interference signal based on the estimate of the transmitted signal; - removing at least some of the interference from the received signal, using the estimated interference signal, and - based thereon, refining the estimate of the transmitted signal, wherein an estimate of the transmitted signal as a result of a later iteration has less interference than an estimate of the transmitted signal as a result of a previous iteration.

According to this iterative technique, an amount of interference may be removed with every iteration (until a substantially interference-free signal is obtained). By way of contrast, with conventional techniques such an iterative process would merely reinforce the decisions previously made, so that no improvement is achieved by performing several iterations.

Some preferred embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which: Fig. 1A is a graph showing the performance of conventional detectors (maximal Doppler frequency Fit = 0 Hz) for the single user case; Fig. 1B is a graph showing the performance of conventional detectors (maximal Doppler frequency Fit = 0 Hz) for the multiple users case; Fig. 2 schematically shows a mathematical model of the transmission of signals through a multipath fading channel; Fig. 3 schematically shows a first embodiment of the present invention; Fig. 4 schematically shows a second embodiment of the present invention; r) Fig. 5 is a graph showing the performance of iterative interference cancellation for unknown interference; Fig. 6 shows features of Orthogonal Frequency Code Multiplex Access (OFCDMA); Fig. 7 is a graph showing the BER performance for the iterative multicode interference cancellation using a (SISO) Turbo decoder; and Fig. 8 is a graph showing a BER performance comparison versus system load.

A mathematical model of a multiuser multicarrier system is described first so as to facilitate the understanding of the present invention. In mathematical formulations used herein, matrices are represented by bold lettering and vectors are represented by bold lettering with a line on top.

Assume a multicarrier CDMA system has K active users, and these transmit simultaneously and use a total of Nf orthogonal subcarriers. Each user employs a unique specific spreading code with a length of SFfreq and SF''me in the frequency domain and time domain respectively. For ease of understanding we assume for the purposes of the following description that BPSK modulation is used and assume that SFime =1. The input sequence for the k-th user is first converted into P parallel data sequences ( ak,O (m), ak,l (m), ak, p (m) ), where - oo < m < oo Then each serial-to-parallel (SIP) converter output is multiplied with the spreading code with length SFfreq All the data in total Nf =PxSFfreq (corresponding to the total number of subcarriers) are modulated in baseband by inverse Fast Fourier Transform (IFFT), which performs a Nf" point inverse discrete Fourier transform. In the downlink (DL) case, all users' modulated chips are summed together and are then converted back into serial data. A cyclic prefix () is then added to form a cyclically extended OFDM symbol to avoid the inter-symbol interference. Thus, the transmitted signal is written as o P-l SFIn,'-l K ak,P(m)Ck(q) pS(t-mT)ej2(pxsFl/rq+q)Af (t-mT5) (Eq 1) m=-= p=0 q=0 k=1 (I 6 Ps (t) = l,0 otherwise (Eq.2) where Ts is the data symbol duration, Af is the subcarrier space, and { ck (0), ck (1), , ck (SFfreq -1) } is the spreading code of the lath user.

The transmitted signals are transmitted through a multipath fading channel (schematically shown in Fig. 2), which consists of L discrete paths, each having different time delays. The time-varying channel impulse response, h(r, t), is represented as a, h(r, t) = a' (t)(r - r') (Eq.3) =o where, a'(t) Complex channel gain of the l-th tap (t) The delta function rim Time delay of l-th propagation path Then the received signal can be written as r(t) = 1 x(t - r) h(r,t)dr + n(t) (Eq.4) where, denotes convolution and n(t) is the noise. We assume that the channel is not changing during one OFDM symbol interval. At the receiver, the cyclic prefix is removed from the received signals, r(t), and SIP converted. Fast Fourier Transform (FFT) is then applied to the received discrete-time signal to obtain Em =HmXm +N'm (Eq.5) (I) 7 where the Y,m, Am and H. m respectively represent the received signal after FFT, the transmitted signal and the channel frequency response, at the i-th subcarrier during the m-th OFDM symbol; N is the additive noise, which is independent of the values of m, i.

Y can be represented as follows: Ye a Ye YO M-] Y = . . (Eq.6) YNI-l O YN, -I M-l where M is the total number of transmitted OFDM symbols in each frame.

Figs. 3 and 4 show the structure and functionality of receivers according to two embodiments of the present invention. Generally speaking, embodiments of the invention aim to provide a technique which efficiently cancels the interference from the received signal (Y) after FFT. However, as mentioned above, in some applications it may be of interest to calculate or estimate the interference without actually cancelling it from the received signal.

Fig. 3 shows the structure and functionality of a single code iterative receiver for MC- CDMA according to a first embodiment of the present invention. The detector can for example form part of a mobile station. We consider a single-code detector, with no additional multicode or broadcast data.

A signal r(t) is received via antenna 10. Fast Fourier Transform is carried out on this signal r(t) at Fast Fourier Transformer 20 to produce the received signal (after FFT) Y. which is distributed over Nf subcarriers. An initial channel estimate (H) is formed at channel estimator 30 from the received signal after FFT Y. H has the following form: Ha) 8 HO I Ho m HO M-I H= Hl Him HiM I (Eq.7) HNr-l,l,-- N/-l,m Nrl,M I Initially, no interference has been estimated, and therefore no interference is subtracted by interference subtracter 50 from the received signal Y. Hence, during this initial part (in the following description referred to as "first iteration") of the iteration process described herein the output Y from interference subtracter 50 is the same as the received signal Y. Next, a suitable weight matrix W is calculated at weight calculator 60, to enable an initial estimation of the received signal to be given. This matrix has the form W = [W r, W. T. .. WM-I] and a matrix size of Nf x (M -1) . The vectors making up this matrix can be written as wm = [we O we, . . . wm N/ ] . We assume that there is only one time-multiplexed pilot symbol at the beginning of each frame.

The combining weights depend on which combining scheme is used. The choice of combining scheme depends on interference quantity. For example, MRC could be selected for a radio environment with good condition, which has little interference, and MMSEC could be used in case the system suffers more interference, so as to restore orthogonality.

As shown in Fig. 3, the choice of combining option is performed by combining option selector 70. Depending on the choice of combining option intermediate steps may be required, for example the calculation of noise by noise calculator 80 (e.g. if MMSEC is chosen).

A signal representative of the choice of combining option, the channel estimate H and any other necessary information/signal (such as the noise calculated by noise calculator 80) is passed to combining weight calculator 60.

For the purposes of describing the first embodiment we assume that there is more than a minimum amount of interference, and therefore MMSEC is chosen as combining scheme during the first iteration.

In the case of MMSEC weights, the combining weight for the '-th subcarrier and the m- th OFDM symbol is given by Wi,m = 1 12 i' / 2. In order to simplify the calculations we consider explicitly only cases where SF = 1 (the extension to the more general case, i.e. where SF does not necessarily take the value 1, will be clear to one skilled in the art). Based on this assumption we have: W m = H m /(|H m| + N2O) (Eq.8) where No represents the noise power density, Cr2 represents the average energy of the symbol (e.g., for QPSK modulation, 2 = 2), and K represents the number of users.

The noise estimate can for example be calculated from the training symbols (assuming that time multiplexed with the desired signal, there will be no contribution from other users within the cell).

These weights W are applied to the received signal Y. and after parallel to serial conversion at parallel to serial converter 90 a first estimate of the desired signal Z is obtained, i.e. of the transmitted signal. This signal Z is passed to de-spreader 100, where correlation with the user's code is carried out (summing over the spreading length), to provide symbol estimates. These are passed to de-modulator 110, to produce coded and interleaved bit information by de-mapping. This is passed to deinterleaver 120, to produce de-interleaved coded bit informational, also referred to as soft decision.

Finally, this is passed to decoder 130 for channel decoding.

In the preferred embodiment decoder 130 is a Turbo decoder. It firstly produces an output (o/p), which is passed to the user interface (not shown), usually subject to additional processing, amplification etc. as is known in the art. Decoder 130 secondly generates a "hard output" or a "soft output" (or hard / soft value) and passes this to re- encoder 140.

We distinguish two cases, soft-in soft-out (SISO) and soft-in hard-out (SIHO). In principle, either the SISO or the SIHO technique can be used. Both techniques can be used to achieve the preferred aim of the present invention of removing interference from the received signal. The choice of which technique is to be used can be based on trial runs. The inventors have found that the SISO technique provides better results at low SINR (signal to interference and noise ratio).

In the SISO case the decoder 130 generates a soft output. This is given by (P(b - +1 a) )) (Eq. 9) In(x) is IOge(X)' P(xLy) is the probability of x given y, bk are the transmitted code bits (bk e {+ 1,-1}), and ark is the coded bit information.

In the SIHO case, the output o/p, which is passed to the user interface, is also passed to the re-encoder 140 as the "hard output". This is the quantized value of the soft output.

This output can, for example, be expressed as {1 if soft output L(bk |a)k) > 0 (Eq. 10) - 1 otherwise The re-encoder 140 receives the hard or soft output and re- encodes it. It also receives information as to which modulation is used (e.g. QPSK, 16QAM etc.). In the SISO case < 11 the soft re-encoded bits are given by E{bk 1 6 k3 = tanh(L(bk I a)k)/2). In the SIHO case the output of re-encoder 140 is the same as the input to re-encoder 140. The output from re-encoder 140 is re-interleaved at re-interleaver 150, re- modulated at re-modulator 160 and re-spread at re-spreader l 70, to produce a new estimate Z of the transmitted signal.

Subtracter 180 forms the difference 1Sea between the original estimate Z of the transmitted signal and the new estimate Z of the transmitted signal. Iser,a,can be expressed by Isern(n) = Z(n) - Z(n) where n = tO, 1,

., Nf.' (M -1)l (Eq.11) This difference ISena, represents an estimated interference serial sequence and may be of use per se, e.g. if it is desired to determine how much interference a particular system has, perhaps so as to test/optimize transmission conditions. However, in the most useful application the estimated interference serial sequence is used to cancel or remove (at least some of, the interference contained in the received signal. To this end the estimated interference serial sequence is subjected to serial-to-parallel conversion at serial-to- parallel converter 190 to generate interference estimate I, which can be represented as: fo,l fO,M-I I = . .. . (Eq. 12) _ Nrll fNf-l,M-I This interference estimate is then weighted with interference weighting (^i/m = [Ym.o Ym tm,Nf -']' matrix size = Nf X (M -1) ), a set of suitable weights Y'm at weighting device 210. These weights are generated by interference weighting generator 200. Usually, y, m should be between O and 1, e.g. = 0 if H estimate confidence = 0 = l if H estimate confidence = 1 O < < l otherwise O 12 Suitable values for Ym (in case the H estimate confidence 0 and 1) can be determined empirically, but preferably, the higher the H estimate confidence the higher should be...DTD: The H estimate confidence can be determined based on the SNR at each subcarrier. The SNR of the i-th subcarrier can be obtained for the m-th OFDM symbol as follows: SNR(i) = |x(i,0)h(i,0)l (Eq. 13) |Y(i,O) - X(i,O)h(i,o)l where X and Y respectively are the transmitted pilot signal and the received pilot signal after FFT. Smaller interference weights should be applied to poor subcarriers (or can be set to zero). If the SNR is below zero, then the channel estimation will be unreliable, i.e. ^t =o.

The weighted interference estimate is then multiplied at multiplier 40 with the initial channel estimate H (formed at channel estimator 30), so as to produce the subcarrier interference estimates. These are then subtracted (by interference subtracter 50) at sub- carrier level from the original received signal Yi.m after FFT, so as to produce a first received signal Y with (at least some) interference removed. The interference subtraction procedure is given below:

A A A

Yi,m = Yi,m-HimYi,mI,m i = {O. . , Nf -1}, m = {1, . . ., M -1} (Eq.14) For Y'm = 0, this corresponds to a conventional detector with no interference subtraction. This is required, as some of the error in subcarrier signal will be due to noise as well as interference. < 13

The system then performs another iteration of forming a (refined) estimate of the transmitted signal, forming a (refined) estimate of the interference, and removing the interference from the received signal.

At least in the ideal case, with each iteration more and more interference is removed, up to an optimum level, after which no more interference is removed.

If MMSEC is used again during any iteration after the first, it is necessary to recalculate the noise estimate, by means of noise calculator 80. If the channel estimate is recalculated it may be possible to determine the noise power at the same time.

However, the present inventors have devised a method of calculating the noise explicitly, by examining the desired (or transmitted) signal Z and considering the interference which has been subtracted. In general, the noise signal can be calculated from the difference between the input of the decoder and the corresponding output from the re-encoded data (assuming no errors in the decoding process). However, this value has been used to calculate the interference contribution, a portion of which has been subtracted, thus reducing the overall noise (and interference) power.

No =- |H, (Z(i) - Z(i))- H. ,y, 11, ,| (Eq. 15) This expression only considers spreading in the frequency domain (by a factor SFfreq) If spreading occurs in the time domain simultaneously (e.g. as in VSF-OFCDM), it is possible to average in both domains. In this case the spreading factor in both domains should be determined by channel coherence bandwidth and mobility. The averaging operation in the time domain can e. g. be multi symbol moving averaging (MSA) or linear interpolation. In the frequency domain, MSA can be used, but the averaging span needs to take into account the channel coherence bandwidth.

If the channel coherence bandwidth is narrow, then the spreading factor in the frequency domain should not be particularly high. If the mobile stations have faster movement, then the channel coherence time is short and the spreading factor in the time domain (mu) 14 should not be particularly large. Otherwise, the correlation at the receiver can be adversely affected.

The channel estimation in the frequency dimension can be obtained by the moving average: 1 m+Nvg | m + N Ym(t)cmd;m/sFl(t) m < NaVg I 1 m+Nat Am = 2NaVg + 1 J nNYm (t) Cmd;m / SF1 (t) NaVg < m < Nf-Nary (Nf-m) + N. + 1 m (t)Cmd;m/sFl(t) m > Nf-Nnvg (Eq. 1 6) where, d;m/SF1 is the common pilot symbol pattern before spreading on the m-th subcarrier at the t-th time.

Linear interpolation is applied in the time dimension, i.e. hm (t) = {m + (m-1) 9m'Shik-{m,l (Eq. 17) block where Bock is the number of OFDM symbols in each block.

However, as mentioned above, for the sake of simplicity we consider only the case of SF = SFfreq and SF'me = 1 explicitly.

Continuing now with the description of the technique shown in Fig. 3, a new set of single-tap equalization weights (combining weights) W is calculated by weight calculator 60. MMSEC is performed again by a modified version (K = 1): W m = H m /(IH m I + 2) (Eq. 18) De-spreading, de-modulation, de-interleaving and decoding is carried out as described in connection with the first iteration. A refined output o/p is thus produced. Likewise, re-encoding (at 140), re-interleaving (at 150), remodulating (at 160) and re-spreading (at 170) takes place, so as to produce a refined estimate of the transmitted signal.

Interference is calculated and then removed from the received signal.

This iterative process is repeated, using MMSEC as weighting technique, until a sufficient level of unknown interference, i.e. intracell interference / multi access interference (MAI) has been removed.

Once the interference level has been sufficiently reduced, it is possible to use a different cost function, for example EGC or MRC. Thus, when interference levels are high, MMSEC is used. MRC can be used as the system approaches the single user bound. For MRC, the weights W to be calculated by weight calculator 60 are given by w, m = H' m.

Generally, as long as the packet error rate (PER) is high, MMSEC should be used. The PER will normally be high if the system has many multiusers and has larger interference. The moment when switching to MMSEC should occur can be based on several parameters, such as the number of users, SNR, channel Doppler fading and rms delay spread. The bit error could reflect the channel environment. In practice, the switching criteria can be based on empirical values.

One possible way of determining whether a sufficient degree of interference has been removed so as to "permit" switching from MMSEC to e. g. MRC would be to compare the output signals of two successive iterations. For example, the absolute value of the difference between the output signals of two successive iterations could be compared, and once it is determined that the absolute value of this difference is sufficiently small then the system could switch to e.g. MRC. As a refinement, the absolute value of this difference could be put in relation to the absolute value of the output signal of one of the iterations, e.g. divided by the absolute value of the output signal of one of the iterations, and the result compared with a predetermined threshold. As a further refinement, three successive iterations (n, n-1, n-2) could be taken into account, and if it is determined (A 16 that the absolute value of the difference between the output signals of iteration n-2 and n-1 is not significantly different from the absolute value of the difference between the output signals of iteration n-1 and n (i.e. that further iterations of MMSEC would not result in much improvement), then the system could switch to e.g. MRC.

If the system has a good channel environment, a MRC single user detector without any interference cancellation (IC) can be selected to reduce receiver complexity.

Turning now to Fig. 4, a multicode iterative detector in accordance with a second embodiment of the present invention is shown. Large parts of the detectors shown in Figs. 3 and 4 are substantially similar, and will therefore not be described again in detail. Like elements carry like reference symbols in Figs. 3 and 4.

In the multicode transmission case of Fig. 4, multicode or broadcast channels (including a code-multiplexed pilot channel) are detected separately and decoded if required.

Outputs of these detectors can then be used to calculate (and subtract) known interference, as compared to the unknown interference which is calculated above.

The functionality of the detector is similar to the single-code detector described above, except that in the system of Fig. 4 the interference estimate is a composite of known and unknown interference.

As in the case of Fig. 3, initial channel estimation, followed by combining using MMSEC and generation of an output signal o/p takes place as previously described. The soft or hard output from the decoder 130 is also re-encoded, re- interleaved, re- modulated and re-spread as described in connection with Fig. 3.

The known interference (common channel interference) Zmc is then constructed by multi-code receiver 207. If more than one common channel is transmitted along with the user by code-multiplexed transmission, then a separate multicode receiver is required for each common channel. The known interference (multi-code interference) Zmc is then passed to subtracter 185, together with the estimate of the transmitted signal O 17 Z and the new estimate Z of the transmitted signal. Subtracter 185 then reconstructs the unknown interference ( Iu) signal, by forming Iu(n)=Z(n)- Z(n)- Zmc (n) n=0, 1, ..., Nf x(M-1)l (Eq.19) Interference generator 200 generates two sets of interference weights, Ymc and Yu. The Yu are used to weight the unknown interference fa, at multiplier 187. The Ymc are used to weight the known interference Zmc' at multiplier 205.

Both the weighted unknown interference Yu fU and the weighted known interference ymc Zmc undergo serial-to-parallel conversion respectively at serial-to- parallel converters 197 and 195, and the respective parallel outputs are added by adder 215. The combined interference is then multiplied with the channel estimate H at multiplier 40 (as in Fig. 3). This is followed by interference subtraction at each sub- carrier (at interference subtracter 50) so as to produce a revised subcarrier signal y m (as in Fig. 3). In other words, the revised subcarrier signal Ym is obtained by subtracting the weighted estimated interference from theoriginal sub-carrier signal, Yjm The interference subtraction procedure is given below: (m = Y,m-Hm(Yu ((m-1)x Nf +i)IU((m-1)xNf +i) + Ymc ((m-1) x Nf + i)ZmC ((m-1) x Nf + i)) i= {O. ..., Nf -1}, m = {1, .. ., M-1} (Eq.20) where, Iu is the unknown interference. The values of Ymc and Yu again depend on the confidence level with which this interference is calculated, and should be between zero and one. In practice, there will be more confidence in the known interference than in the unknown interference, i e Yu < Ymc id>; 18 On the reasonable assumption that there is only one pilot symbol at the beginning of each frame, the noise calculation to be used in connection with the system of Fig. 4 is modified as: N; = I I |H '(z(i)-z(i)) H,,Ymc()Zmc(i) H',IYu(i)lu()l (Eq. 21) The noise variance can be obtained by: NO = N Il Y.m- Ymc (i)2nc (i) H',' | (Eq.22) where, Zmc (i) denotes the detected received signal for the 1 it OFDM symbol of the k-th spreading code at the i-th subcarrier. Again, this expression only considers spreading in the frequency domain (by a factor Nf). If spreading occurs in the time domain simultaneously (e.g. as in VSF-OFCDM), it is possible to average in both domains, as explained in connection with the Fig. 3 embodiment.

The iterative process of the Fig. 4 embodiment, including the switching from e.g. MMSEC to MRC, follows the same principles as the Fig. 3 embodiment.

As will be seen, the invention may improve system performance by successively calculating (and subtracting) an interference estimate. In conventional systems this would merely reinforce the decisions that have been previously made in the decoder.

However, by the use of updated combining weights the performance is modified, depending on the current levels of noise and interference.

In addition, known data (e.g. from multicode, broadcast or pilot channels) may be available and can be used to subtract interference. This can be either instead of, or in addition to, the unknown interference subtraction. In either case, it is expected that the modification of weights would occur as normal. O 19

Preferred embodiments of the invention also enable multi-user (rather than multicode detection) to be performed at an access point in a similar way.

Initial simulation results demonstrate a satisfactory ability of iterative interference cancellation. Time-multiplexed pilot transmission was employed for the simulation.

Hence, the noise variance could be initially obtained from pilot parts. Fig. 5 demonstrates the IC ability for cancelling unknown MAI. As compared with the case where there is no iterative IC, about 2.4 dB improvement can be achieved at a BER of le-5 by employing the SIHO based iterative IC, and 3dB gain can be achieved by means of the SISO based iterative IC.

Fig. 6 illustrates features of Orthogonal Frequency Code Multiplex Access (OFCDMA).

In OFCDMA, each user occupies a different frequency band within the total bandwidth, BW, but multicode transmission is used to achieve a high date rate. Figs. 7 and 8 present initial results for the multicode transmission case. The noise variance is estimated from the pilot part at the initial stage (before IC iteration), and the modified noise variance after IC is obtained from Eq. 22 but with Ymc (i' m) = 1. Fig. 7 shows that, after the 2 iteration, the BER curve is close to the single user bound and only has about 0.5 dB degradation at a BER of le-3. At a BER of le-4, the iterative IC has 2.3dB and 3.5 dB performance improvement for one iteration and two iterations respectively, compared with the case where there is no IC. Fig. 8 indicates that, by using iterative multicode IC, the system load (ratio of number of code multiplexing to the spreading code length, C/SFfreq) and BER performance can be dramatically improved. For example, at a BER of 3e-3, CnuxlsFfreq = 100% (full system load) can be supported by the iterative structure, but only C,,,a, /SFfreq = 31.25% can be supported by the non-iterative structure.

Various features of embodiments of the invention have been described with reference to iterative processes. It will be appreciated that some features described in the present specification find application, and may be useful, without there being an iterative process. For example, generating an initial interference estimate (during the "first iteration" may be useful. Those skilled in the art will recognize other features disclosed herein which may be of use after initial processing, i.e. without "iteration".

(id) 20 Although the invention has been described in terms of preferred embodiments as set forth above, it should be understood that these embodiments are illustrative only and that the claims are not limited to those embodiments. Those skilled in the art will be able to make modifications and alternatives in view of the disclosure which are contemplated as falling within the scope of the appended claims. Each feature disclosed or illustrated in the present specification may be incorporated in the invention, whether alone or in any appropriate combination with any other feature disclosed or illustrated herein.

Claims (31)

1. (me) 21 CLAIMS: 1. A method of processing a received signal which is

   spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: a) forming an estimate of the transmitted signal from the received signal; b) de-spreading, de-modulating, de-interleaving and decoding the estimate of the transmitted signal so as to provide an output signal; c) re-encoding, re-interleaving, re-modulating and re- spreading a signal which is based on the estimate of the transmitted signal after de spreading, de- modulating, de- interleaving and decoding, so as to provide a new estimate of the transmitted signal; and d) forming an estimated interference signal based on the estimate of the transmitted signal and the new estimate of the transmitted signal.
2. 2. A method according to claim 1, wherein forming the estimated interference signal comprises forming the difference between the estimate of the transmitted signal and the new estimate of the transmitted signal.
3. 3. A method according to claim 1 or 2, further comprising al) forming a refined estimate of the transmitted signal based on the received signal and the estimate of the interference.
4. 4. A method according to claim 3, wherein forming the refined estimate of the transmitted signal based on the received signal and the estimate of the interference comprises subtracting the estimate of the interference from the received signal.
5. 5. A method according to claim 3 or 4, further comprising bl) despreading, de-modulating, de-interleaving and decoding the refined estimate of the transmitted signal so as to provide a refined output signal.
6. 6. A method according to any of claims 3 to 5, further comprising at least one iteration of: By) 22 cl) re-encoding, re-interleaving, re- modulating and re-spreading a signal which is based on the refined estimate of the transmitted signal after de-spreading, de modulating, de- interleaving and decoding, so as to provide a new refined estimate of the transmitted signal; dl) forming a refined estimate of the interference based on the refined estimate of the transmitted signal and the new refined estimate of the transmitted signal; and a2) forming a (further) refined estimate of the transmitted signal based on the received signal and the refined estimate of the interference.
7. 7. A method according to claim 6, wherein forming the refined estimate of the interference comprises forming the difference between the refined estimate of the transmitted signal and the new refined estimate of the transmitted signal.
8. 8. A method according to claim 6 or 7, further comprising b2) despreading, de-modulating, de-interleaving and decoding the (farther) refined estimate of the transmitted signal so as to provide a (further) refined output signal.
9. 9. A method according to claim 8, comprising providing the (further) refined output signal for each iteration.
10. 10. A method according to any of claims 1 to 9, wherein re-encoding comprises re encoding a soft output from the decoding step.
11. 11. A method according to claim 10, wherein the soft output is given as L(bk I ark) = Inn p(bk _ 1 k) if wherein In(x) is IOge(x), P(xLy) is the probability of x given y, bk are the transmitted code bits, and aJk are the result of the de-spreading, de-modulating and de-interleaving steps.
12. 12. A method according to any of claims 1 to 9, wherein re-encoding comprises re- encoding a hard output from the decoding step.
13. 13. A method according to claim 12, wherein the hard output is said output signal.
14. 14. A method according to any of claims 1 to 13, wherein the estimated interference signal and, where provided, the refined estimate of the interference comprises an estimate of unknown interference.
15. 15. A method according to claim 14, further comprising forming a signal representative of known interference.
16. 16. A method according to any of claims 1 to 14, wherein the received signal is based on a single-code transmission.
17. 17. A method according to any of claims 1 to 15, wherein the received signal is based on a multi-code transmission.
18. 18. A method according to claim 6 or any of claims 7 to 17 as directly or indirectly dependent on claim 6, wherein forming the estimate of the transmitted signal from the received signal comprises a first combining technique, and forming the (further) refined estimate of the transmitted signal either during the first iteration or, where provided, during a later iteration comprises a second combining technique different from the first combining technique.
19. 19. A method according to claim 18, wherein the selection of the combining technique is based on the amount of interference remaining in the refined or the further refined estimate of the transmitted signal.
20. 20. A method according to claim 19 as indirectly dependent on claim 8, wherein an indication of said amount of interference is derived by comparing the output signals of two successive iterations.

    l\ ) 24
21. 21. A method according to claim 20, wherein comparing said output signals comprises forming the absolute value of the difference between said output signals and putting said absolute value of said difference in relation to the absolute value of one of said outputs.
22. 22. A method according to any of claims 18 to 21, wherein the first combining technique is MMSEC and the second combining technique is EGC or MRC, preferably MRC.
23. 23. A method according to any of claims 1 to 22, wherein forming the refined estimate of the transmitted signal based on the received signal and the estimate of the interference comprises weighting the estimate of the interference.
24. 24. A method according to any of claims 1 to 23, further comprising applying a Fast Fourier Transformation to the received signal.
25. 25. A method according to any of claims 1 to 24, wherein the received signal is a signal received in a downlink transmission.
26. 26. A method according to any of claims 1 to 25, further comprising forming a signal representative of the noise, based on the estimated interference signal.
27. 27. A method of processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: forming an estimate of the transmitted signal from the received signal; and at least two iterations of: - forming an estimated interference signal based on the estimate of the transmitted signal; removing at least some of the interference from the received signal, using the estimated interference signal, and - based thereon, refining the estimate of the transmitted signal,

    -

    wherein an estimate of the transmitted signal as a result of a later iteration has less interference than an estimate of the transmitted signal as a result of a previous iteration.
28. 28. Apparatus for processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: a) means for forming an estimate of the transmitted signal from the received signal; b) means for de-spreading, de-modulating, deinterleaving and decoding the estimate of the transmitted signal so as to provide an output signal; c) means for re- encoding, re-interleaving, remo dul ating and re- spreading a signal which is based on the estimate of the transmitted signal after de- spreading, de- modulating, de-interleaving and decoding, so as to provide a new estimate of the transmitted signal; and d) means for forming an estimated interference signal based on the estimate of the transmitted signal and the new estimate of the transmitted signal.
29. 29. Apparatus for processing a received signal which is spread over a plurality of subcarriers in a multicarrier CDMA system and which is based on a transmitted signal and may include interference and noise, comprising: means for forming an estimate of the transmitted signal from the received signal; and means for carrying out at least two iterations of: - forming an estimated interference signal based on the estimate of the transmitted signal; - removing at least some of the interference from the received signal, using the estimated interference signal, and - based thereon, refining the estimate of the transmitted signal, wherein an estimate of the transmitted signal as a result of a later iteration has less interference than an estimate of the transmitted signal as a result of a previous iteration. ) 26
30. 30. A method substantially as herein described with reference to, or as illustrated in, Figs. 3 or 4 of the accompanying drawings.
31. 31. An apparatus substantially as herein described with reference to, or as illustrated in, Figs. 3 or 4 of the accompanying drawings.