WO1999001945A1

WO1999001945A1 - Signal detection method in a digital cellular receiver

Info

Publication number: WO1999001945A1
Application number: PCT/FI1998/000528
Authority: WO
Inventors: Marko Escartin; Pekka Ranta; Esa Malkamäki; Riku Pirhonen
Original assignee: Nokia Networks Oy
Priority date: 1997-06-19
Filing date: 1998-06-17
Publication date: 1999-01-14
Also published as: FI972688A0; AU7771398A; JP2002506594A; EP0990309A1; AU741617B2; FI972688L; FI104019B; CN1260914A; NO996262L; FI104019B1; NO996262D0

Abstract

The invention relates to a signal detection method in a digital cellular network receiver. A combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to a desired channel is received on the desired channel. In accordance with the invention channel estimates of the useful signal and at least one adjacent channel signal are jointly determined. The estimation is performed either using reference parts, for example a training sequence, known by the receiver or using symbol or bit decisions produced by decision feedback. Then the useful signal is detected utilizing the channel estimates of both the useful signal and the adjacent channel signal. The estimation can also be performed separately, in parallel or sequentially, in which case a reconstructed interfering signal is reduced from the useful signal before detection.

Description

SIGNAL DETECTION METHOD IN A DIGITAL CELLULAR RECEIVER

FIELD OF THE INVENTION

The present invention relates to a signal detection method in a digital cellular network receiver, the method receiving on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel.

BACKGROUND OF THE INVENTION

A cellular network receiver receives a signal transmitted on a par- ticular channel. Signals transmitted on adjacent channels cause interference to the signal, which is referred to as adjacent channel interference. Different frequency division channels are located close to each other. For example, in the GSM system the medium frequencies of the channels are located at a distance of 200 kHz from one another, and the modulation spectrum is slightly wider than 200 kHz. The guard band between channels is thus non-existent. Due to the properties of a GMSK modulation method used adjacent channels leak into the frequency range of each other. Furthermore, signals to be transmitted on the same channel in a cell somewhere further cause interference to the signal, which is known as co-channel interference. In current cellular networks interfering signals are approximated in receivers as random additive white Gaussian noise. The approximation is sufficiently accurate only if interfering signals in relation to a desired signal are sufficiently weak. Adjacent channels thus interfere each other too much, if the difference of powers used thereon is too high. In the GSM system the largest difference of transmission or reception powers allowed of adjacent channels is 9 dB.

Since it is necessary to allow larger power differences with regard tothe functionality of the system, the use of adjacent channels in the same cell is not possible. Cellular networks use a reuse pattern determining how the channels are reused in the cells. When the reuse pattern is for instance 7, then the cell and its six neighbouring cells each use a different channel (or channels). Reuse pattern planning requires thorough planning of radio frequency use. The planning results are tested upon commissioning the network by extensively measuring radio fields. Planning and measuring are important steps also when the network is enlarged.

The signal generally propagates along various paths due to reflec- tions caused by obstacles and arrives at a receiver as signal components delayed in different ways. The phenomenon is referred to as multipath propagation. To signal symbols this causes inter symbol interference, in which symbols are partly superimposed. A predetermined reference part, by which a multipath channel through which the signal has propagated can be estimated at the receiver, is generally included in the signal in cellular radio systems. Using a channel estimate thus formed the received signal can be corrected to correspond to the original.

CDMA systems use a particular pilot signal and/or a broadband spreading code as the reference part.

TDMA systems use a training sequence as the reference part. The GSM system, for example, employs eight training sequences, which are selected so as not to resemble the coded data transferred in the signal, but are a unique part of each burst. Transmitters in a particular cell operating on a par- ticular channel all use the same training sequence. The transmitters on an adjacent channel and a co-channel use a different training sequence. A receiver can thus distinguish the correct signal from an interfering signal arriving at the same time. Since interfering signal powers are kept sufficiently low by means of frequency planning, the interfering signals are processed at a re- ceiver merely as additive white Gaussian noise. Actual channel estimation, correction and detection are performed without information on interfering signals. This does not cause any problems in cellular networks where frequency planning is correctly performed.

However, as mobile communication is becoming increasingly popular network operators tend to utilize the radio frequency areas obtained to best possible effect. This requires reducing the reuse pattern, even the value one of the reuse pattern is possible. Then it is necessary to use the adjacent channels in the same cell.

The cellular network capacity is also increased by introducing hier- archie cell structures, in which cell sizes vary in diameter from dozens of kilometres to some dozens or hundreds of metres. Such cells are referred to as macro, micro and pico cells. In these cells the power level differences of adjacent channels can be very extensive and cause severe interference.

For example, in CDMA systems the power control problem of the adjacent channels is solved in such a manner that an attempt is made to attain all signals in the uplink direction to the same power level at a base station re- ceiver. All signals in the downlink direction are transmitted at constant power. A drawback when applying this idea to the GSM system is that the dynamic area of power control should be increased from the current level, 30 dB, up to 60-100 dB. However, this is not possible between different cell types since too much interference would occur.

A theoretical solution for decreasing interference caused by an adjacent channel is to increase the guard band between channels. However, this would waste the frequency area in use, and its implementation to current systems is not possible without changing the specifications. In connection with CDMA systems the use of different interference cancellation techniques is studied. Applying said techniques to TDMA systems is, however, substantially more difficult, since the separation of signals is not optimized at transmitted waveforms as in CDMA systems. In addition co- channel interference cancellation is generally studied in CDMA systems, as the present invention, in turn, focuses on interference cancellation caused by an adjacent channel.

The Finnish patent publication 944 736, for example, describes how co-channel interference cancellation can be implemented at a receiver. However, said publication does not describe how to remove the adjacent channel interference.

In brief, it can be noted that the current operation complicates the radio frequency planning of the network, prevents the use of small reuse patterns and complicates the further development of current systems.

BRIEF DESCRIPTION OF THE INVENTION

An object of the present invention is to provide an improvement on signal detection in a cellular receiver by removing an interfering signal caused by an adjacent channel.

This is achieved by the method of the invention of the type shown in the preamble, characterized by determining jointly channel estimates of a useful signal and at least one adjacent channel signal, detecting the useful signal utilizing the channel estimates of the useful signal and the adjacent channel signal.

The invention also relates to a signal detection method in a digital cellular network receiver, the method receiving on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel. The method is characterized according to the invention by determining channel estimates of the useful signal and at least one adjacent channel signal, reconstructing an adjacent channel signal from the useful signal utilizing the channel estimate of the adjacent channel signal, reducing the reconstructed adjacent channel signal from the useful signal, detecting the useful signal utilizing the channel estimate of the useful signal.

The invention further relates to a digital cellular network receiver arranged to receive on a desired channel a combination of a desired useful sig- nal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel. The receiver is characterized according to the invention by comprising a channel estimator arranged to jointly determine channel estimates of the useful signal and at least one adjacent channel signal, a detection part arranged to detect the useful signal utilizing the channel estimates of both the useful signal and the adjacent channel signal.

The invention further also relates to a digital cellular network receiver arranged to receive on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel. The receiver is characterized according to the invention by comprising at least one channel estimator arranged to determine channel estimates of the useful signal and at least one adjacent channel signal, reconstruction means to reconstruct an adjacent channel signal from the useful signal utilizing the channel estimate of the adja- cent channel signal, means to reduce the reconstructed adjacent channel signal from the useful signal, a detection part arranged to detect the useful signal utilizing the channel estimate of the useful signal.

Considerable advantages are achieved with the invention. By comparing interfering signals one or more of the most interfering signals can be selected as interference cancellation subjects. In accordance with the invention interference caused by an adjacent channel signal can also be removed. This enables the use of smaller reuse patterns and facilitates radio frequency planning. Alternatively if there is no need to increase system capacity the invention can improve the radio connection quality or increase the radio connec- tion range.

In accordance with the invention it is possible to jointly determine channel estimates of a useful signal and at least one adjacent channel signal. This is particularly suitable for implementing a subscriber terminal receiver, since then the invention will not cause additional costs to the implementation of the receiver. Then, when a channel estimate is determined for the adjacent channel signal a generated altered training sequence can be used where a phase distortion caused by the frequency difference between a desired channel and an adjacent channel is taken into account. This improves the accuracy of the channel estimates. Correspondingly the phase distortion can also be taken into account when a desired signal is detected.

In accordance with the invention it is also possible to determine the channel estimates of the useful signal and at least one adjacent channel signal in parallel, and to reconstruct an adjacent channel signal from the useful signal utilizing the channel estimate of the adjacent channel signal. This procedure is particularly suitable for a base station, since it requires a receiver for each channel to be received. The advantage with this embodiment is quality, as it is easier to remove interference more accurately if the structure of the interference is known.

In accordance with the invention it is also possible to receive in ad- dition to a desired channel at least one channel adjacent to the desired channel, and then filter the desired channel and the adjacent channel apart. Signal estimation and detection is then performed sequentially. This embodiment is well suited for a subscriber terminal due to the relative simplicity thereof compared to the previous embodiment. The invention also enables to determine the channel estimates of the useful signal and at least one adjacent channel signal in such a manner that instead of reference parts symbol or bit decisions made from a received signal are used regarding the useful signal and at least one adjacent channel signal. This known method is referred to as decision feedback. The estimation method of a channel estimate is particularly suitable for

- updating the channel estimate, for example, in such cellular radio systems, which include channel estimation based on a reference part, but in which the channel has time to change considerably during the time between reference parts (for example, because a transmitter is in a fast moving vehicle or because there is a long time slot between reference parts).

- estimating and updating the channel estimate in such cellular ra- dio systems, which do not necessarily include an actual reference part or symbols, for example possibly on a transmission path directed towards a base station from a CDMA system subscriber terminal. Then it may be preferable to make a plurality of recursions for estimating a channel estimate and symbol decisions so that the estimated symbols specify the estimate of the channel estimates and the estimated channel estimates specify the estimate of the symbol sequence. This recursive repeatedly specified method is also known and in this application it is intended to be used for blind channel estimation.

In accordance with the invention in a GSM-based system in the downlink direction a specific modulation method can be used for the signal enabling substantially better protection against interference caused by an adjacent channel than a conventional modulation method. Then in the uplink direction a conventional modulation method, such as a GMSK modulation method, can be used for the signal. A considerable advantage is achieved with this embodiment; the necessary changes to the receiver and transmitter of the base station can be carried out in accordance with the invention. The receiver of the subscriber terminal does not, however, require any changes, in which case already existing subscriber terminals can be used in the improved network, where for instance the reuse pattern is reduced. According to studies performed by the applicant an OQAM method (Offset Quadrature Amplitude Modulation), for example, is in this case particularly well suited as the specific modulation method, since its demodulation can be made to match the already existing demodulation used by the subscriber terminal receivers.

The invention allows the use of a neighbouring frequency in the system, for example, in adjacent cells of a hierarchic cellular radio network, in which the power levels used can be of very different sizes. An example of such a situation is a macro cell covering a wide area and an internal pico cell of an office building in said area. This advantage relates to CDMA, TDMA and various hybrid systems.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following the invention will be described in greater detail with reference to the examples of the accompanying drawings, in which Figure 1 illustrates a cellular network of the invention, Figure 2 illustrates interference caused by an adjacent channel signal to a desired signal, Figure 3 shows a discrete time model of signal structure,

Figure 4A shows a receiver of the invention jointly determining channel estimates,

Figure 4B shows a receiver of the invention using a set of antennas, Figure 5 shows a receiver of the invention separately determining the channel estimates in parallel,

Figure 6 shows a receiver of the invention receiving a substantially broader frequency band than the desired signal.

DETAILED DESCRIPTION OF THE INVENTION

The present invention is applicable to be used in all digital cellular radio networks that allow interference caused by an adjacent channel. Here the GSM system is used as an example without restricting the invention thereto. Thus, TDMA systems, CDMA systems, SDMA systems and various hybrid systems concurrently using different multiple methods are examples of the cellular radio networks the invention refers to.

Figure 1 describes a part of the cellular network of the invention. A base station 100 receives a signal 104 transmitted by a moving station 102 on channel 890.2 MHz. A moving station 106 in another cell transmits a signal 108 to its base station on the same channel 890.2 MHz. A moving station 110 in a neighbouring cell transmits a signal 112 to its base station on channel 890 MHz. Correspondingly a moving station 114 in another neighbouring cell transmits a signal 116 to its base station on channel 890.4 MHz. Signal transmissions substantially occur simultaneously, so as to the receiver 100 the de- sired signal 104 (at 890.2 MHz frequency) is now interfered by

- the co-channel signal 108 (also at 890.2 MHz frequency),

- the adjacent channel signal 112 of a lower frequency (at 890 MHz frequency), and

- the adjacent channel signal 116 of a higher frequency (at 890.4 MHz frequency).

Since interference cancellation caused by an adjacent channel signal in particular is discussed in the invention, a closer look is taken at how interference occurs with regard to a desired signal. Figure 2 shows a frequency on x-axis and a power spectrum on y-axis. Curve 200 is the power spectrum of a signal on a desired channel. Curve 202 is the power spectrum of a signal on a channel below the desired channel, for example at frequency f1 - 200 kHz in the GSM system. Curve 204 is the power spectrum of a signal on a channel above the desired channel, for example at frequency f1 + 200 kHz. The spectrum produced by a GSMK modulation method used in the GSM system is in a sense infinite by nature, and therefore adjacent channels inevi- tably overlap each other; these areas 206, 208 are hatched in the Figure. In the hatched area 206 the adjacent channel signal 202 interferes the desired signal 200. In the area 208 the adjacent channel signal 204 interferes the desired signal 200, respectively.

The invention can be used in cellular networks that include a pro- tective frequency area between channels. The invention can also be used in cellular networks where the frequencies are as close as possible to each other, even so close that the channels partly overlap each other. The invention can further be used in cellular networks that use an expanded frequency channel i.e. the channel is considerably broader (e.g. 200 kHz) than a normal channel. The cellular network of the invention can also be semi-adaptive regarding the protective frequency and/or the channel width, i.e. the network operator can place them in any way he/she wants according to situation.

The signals to be processed can be described by a discrete time model shown in Figure 3. The system comprises N_c transmitters transmitting on the same channel f1 and in the same time slot. In addition the system comprises Λ/_a transmitters transmitting on channel fO and/or f2 adjacent to channel f1 at the same time as the transmitters of channel f1. Each transmitter transmits the signal within the time slot as a radio frequency burst including a symbol sequence to be transmitted, K by length, on the same channel a_κ,n_c = (a₁,n_c, a₂,n_c, ■■■ a_K,_nJ , and on the adjacent channel a_κ,n_a = (aι,n_a, a₂,n_a, ■■■ a_κ,_n ■ The impulse response of each channel is on the same channel _L,n_c ^~ (h₀,n_c, hι,_nc, ... h_L,_nJ , and on the adjacent channel hL.n_a = (h₀,n_a, _1ιna, ... h_{L a}) ■ L is the length of a channel memory in symbols. Vector r_κ = (r₁, r₂, ■■■ "xj is a received signal sequence. Vector n_κ consists of independent white Gaussian noise. The model is simplified in the sense that it assumes the channel memory length to be finite and equal for all channels. As for the invention, the model is sufficiently accurate, since in practice the receivers process only finite impulse responses.

The received signal r_κ sampled once per symbol can be written as follows N_c L N_a L r_κ = ∑ ∑hι,_nca_k-ι,n_c + ∑ ∑hι,_na ^ak-l,n_a n_k n_c=1l=0 n_a=1l=0 (1 )

The receiver has to be able to detect the N_c+N_a transmitted data sequence a_{K n} from the received signal r_κ. This is possible with high probability if the channel states do not overlap. In cellular networks the channels are summed in random phases and amplitudes, the probability of overlapping thus being very small.

Figure 4A shows a simplified block diagram of the receiver of the invention. Figure 4A includes only the blocks essential for describing the in- vention, but for one skilled in the art it is obvious that a conventional cellular network receiver also comprises many more functions and structures, the detailed description of which is not necessary in this context. In practice the receiver may be for example a receiver normally used in the GSM system comprising the changes required by the invention. A sum signal received by an antenna 400 and including a desired signal, interference caused by an adjacent channel signal and interference caused by a co-channel signal is applied to radio frequency parts 402 from where the sum signal is applied to a filter 404, where selection filtering of a desired channel is performed for the sum signal by a band-pass filter. The sum signal is converted in further processing means 406 to an intermediate frequency or directly to a base-band. The signal is demodulated into I and Q parts. An A/D transformation is performed for the I and Q signals. The signal can thereafter be more accurately filtered. After the described measures the signal is applied from the further processing means 406 to a detection part 414 and to a channel estimator 408. The channel estimator 408 estimates channel estimates i.e. for each channel a vector h describing the amplitude and phase of the multipath propagated components of the signal at different delays. A joint channel estimation is performed for the signals. Let us assume that N_c synchronic co-channels are received i.e. a desired channel and N_c-1 interfer- ing co-channels and in addition N_a synchronic interfering adjacent channels. The channel responses of the co-channels are denoted

^c _n = (ho,n, ₁,n,... _L.n), n = 1,2,...N_c (2)

the length of each of which is L+1 with complex channel tap weights. Correspondingly the channel responses of the adjacent channels are denoted h^an = (ho.„,h_1ιn,...hL.n), n = 1,2,...N_a (3)

Both co-channel and adjacent channel impulse responses are collected into the vector h as follows

h = (h h^C2,...h^c _Nc,h h^a ₂,...h^a _NJ (4)

The training sequence divided into preamble and midamble codes is denoted on the n^th co-channel by

m^c _n = (mo,n,m₁,n,...m_P+L-₁,n)^T , n = 1,2,...N_c (5)

and on the n^th adjacent channel by

m^a _π= (m₀,n, m^ -β^, m₂,_π -e^J2m, m₃,_n -e^J3(2πfAT>,

(6) m_(PWe«^p+L-¹»^2nΑT>)^τ, n = 1,2,...N_a

When a channel estimate is determined for an adjacent channel signal an altered training sequence is used where a phase distortion caused by the frequency difference between the desired channel and the adjacent channel is taken into account. In the formula a complex value mixer wave signal is generated by raising term e at a power, and fΔ is a frequency difference, or the difference of medium frequencies between the desired channel and the adjacent channel, and j is an imaginary number by which the wave becomes sinusoidal. The formulas include L+P element m_{p n}, where L is the length of the preamble code which equals the length of the channel memory, and P is the length of the midamble code. The first L bits are preamble code bits and the following B bits are midamble code bits.

The received signal corresponding to the midamble code bits can thus be presented in the form

y=Mh+n (7)

where n represents Gaussian noise samples with a covariance matrix R, and the matrix M = (Mι ,M₂,...M^c _Nc,M^a,M^a ₂,M^a _Na) includes the transmitted training sequences arranged as matrixes to the co-channels WΛ , n = 1,2,...N_c as follows

m L,n ™ι,n m_on m L+1,n m₂,_n m

M_n ^c = (8) m P+L-1,n m_P. m P-1,n

Correspondingly a matrix can be formed for the adjacent channels taking the rotated training sequence into account

jLΣUfAT m _n - e^jL(mfAT> ^{■ ■} m ™₀,n j(L₊i) (₂πfAV , _ej2(₂πf&T)

"'L+1,n ^c ^{■ ■} m₂,n m_{1 n} - e^j2ΩΑT

M„' = (9) _m „j(P₊L-1) (2τifAT) _ejP( πfA _m „J(^p-¹) (∑UfAT) "p+L-1,n ^c ^■ mp,n "' P-1,n ^G

The maximum likelihood channel estimate is obtained by the equation

(h)_ML = (M^HR-¹Mr M^HR-¹y (10)

Assuming that the noise is white the equation is reduced to the form

(h)_ML = (M^HM M^Hy (1 1 )

This results in the multipath channel estimates of the desired signal, the multi- path channel estimates of the interfering co-channel signals and the multipath channel estimates of the interfering adjacent channel signals.

When a channel estimator 408 has estimated a desired channel, channel estimates for co-channels and adjacent channels either sequentially or in parallel for each signal, the channel estimates are applied to means 410 where a receiving power is formed for each signal. The channel estimates are further applied to means 412. From the means 410 the receiving power formed for each signal is applied to the means 412. Interference caused to a useful signal by an adjacent channel signal and a co-channel signal is com- pared in the means 412, and a decision is made concerning which interference or interferences should be removed. From means 414 the channel estimates of the desired signal and at least one of the most interfering signal are applied to a detection part 414. In the detection part 414 the desired signal is detected by removing from the useful signal the effect of the most interfering signal.

It is known that an optimal joint detection algorithm in a situation where interference between symbols and white Gaussian noise occur is a joint maximum likelihood sequence estimation (JMLSE) that can recursively be implemented by a Viterbi algorithm. By using the standard trellis search technique it is possible to find the symbol sequences most likely transmitted by the JMLSE algorithm among all possible sequences using the maximum likelihood criterion

max\ P(^rκ

'^aK,₂ ' - - -^aK,N ) ^aK,n (12) n [l,N]

where

is a joint probability density function for random variables r_κ depending on the transmitted sequences a_{K n}. The most likely transmitted symbol sequences maximize the above value. If an independent noise is assumed, then the equation can be denoted

'K,n (13) n ≡[l,N] where vector a_{L+1 n} includes L+1 of the symbol previously transmitted on the n^th channel i.e. a_{L+1 n} = (a_{k n}, a_k._{1 n}, ...a_k._{L n}). Assuming that the noise is Gaussian whose average and deviation σ² are 0, then the conditional probability density function can be written

P(r_κ ^aL+1,1 > ^aL+₂,₂ > L+1,N )

Equation 12 can be formed using the previous equation 14 into the form

^JK,n (15) n e ]

This equation returns the minimum sum of the Euclidean distances of all possible sequences.

Since the use of the Viterbi algorithm requires a recursive formulation of the probability density function, the final form of the JMLSE path metrics is

(^ak,n) =

where term J (a _n), n=1 ,2,..N_c+N_a represents the survivor path metrics in the previous phase of the trellis.

The number of states in the JMLSE trellis diagram is 2^N . In practice the calculation capacity of the receivers limits the number of signals that can be jointly detected or the number of multipaths of the signals. In the cellular network one interferer is more dominating than the others, so it is preferable to remove the effect of at least one of the strongest interferers.

A joint estimation and detection provide the best results in a synchronic network. If the network is not synchronic, the interference situation may change in the middle of the burst. Then interference is estimated from the part of the burst it affects. Channel estimation is performed independently for each training sequence, which is assumed to cause interference. Channel es- timation is then a sliding measure by nature as the timing of the interference is not known. Sliding means that according to the timing of the desired signal training sequence a search for the interfering signal training sequence is started.

Another solution to the interference synchronization problem is that a receiver is used which can receive, estimate and detect signals in parallel. Such an operation is particularly well applicable to a base station typically re- ceiving several signals concurrently and detecting them.

Figure 5 shows such a receiver. The receiver is in a sense the receiver described in Figure 4A, but provided with several receiver branches and deviating properties described below. The receiver shown in Figure 5 com- prises three receiver branches. Each receiver branch filters the desired channels fO, f1 , f2 apart with filters 404A, 404B, 404C.

A closer look is taken below at the function in the receiver branch of channel f1. In the receiver branches a sum signal of each channel including a desired signal, interference caused by an adjacent channel signal and inter- ference caused by a co-channel signal is processed as described in Figure 4A.

What deviates is a channel estimator 408A, 408B, 408C estimating channel estimates for a desired channel and a co-channel. Channel estimation is thus not performed for adjacent channels. The channel estimates and power values are applied to a channel estimate bus and to a power bus 504. The channel estimate bus 506 transmits the channel estimates formed in each receiver branch to reconstruction means 500B of an own receiver branch and to reconstruction means 500A, 500C of other receiver branches, and in addition to a detection part 414B of the own receiver branch. The power bus 504 cor- respondingly transmits the power estimates formed in each receiver branch to the reconstruction means of the own 500B and the other 500A, 500C receiver branches. In the reconstruction means 500B an adjacent channel signal is generated from a desired signal utilizing the adjacent channel signal channel estimate obtained from the channel estimate bus 506 and the signal sequence received in the second receiver branch obtained from a symbol bus 508. The generated adjacent channel signal is then applied to means 502B. A reconstructed adjacent channel signal, or the adjacent channel signal of channel fO and/or channel f2, is reduced from the received useful signal in the means 502B. Then a useful signal is detected in the detection part 414B utilizing the channel estimate of the useful signal. The symbols identified from the detection part 414B are applied to be further processed and to the symbol bus 508 transmitting the identified symbols to those adjacent channels (in this case channels fO and f2) to which said channel possibly causes adjacent channel interference. Said receiver in Figure 5 thus utilizes the property that the interfering adjacent channels are desired regarding a receiver branch. One of the channels is likely to be stronger than the others, and therefore easier to detect. The detection of other channels can utilize the fact that based on the content of said strongest channel the interference caused by the strongest channel to adjacent channels can be reconstructed. If a sufficient amount of capacity is available, some receiver branches can detect the signal causing interference also merely because interference cancellation of some channels is thus facilitated in the described manner.

The function of the reconstruction means 500B is based on the fact that when the impulse response h and pulse mode p of the adjacent channel are known the interference received at a side frequency of the desired channel in accordance with the equation p * h, where ^* is a convolution operator.

The pulse mode p(t) is formed for an adjacent channel at a lower frequency using a Fourier synthesis or a corresponding convolution in time by equation

p(t) = mfvx (t) e'^ωt * mf_RX (t) (17)

where mf-rχ (t) represents the impulse response of the transmission filter pulse mode, and e^iωt represents channel difference, and mf_RX (t) represents the im- pulse response of the receiving filter pulse mode, and * represents a convolution operator.

Correspondingly the pulse mode p(t) is formed for an adjacent channel at a higher frequency by equation

p(t) = mf™ (t) e+* * mf_RX (t) (18)

where mf_τx (t) represents the impulse response of the transmission filter pulse mode, and e^jωt represents channel difference, and mf_RX (t) represents the impulse response of the receiving filter pulse mode, and * represents a convolu- tion operator.

An antenna arrangement employing several antennas, for example 2-20 separate antennas, can be connected to the receiver. The receiver is then like the one described in Figure 4B. The sum signal received by an antenna 400A-400N is applied to its own radio frequency parts 402-402N, from where the sum signal is applied to a filter 404A-404N and further to further processing means 406A-406N. For processing the signal received by the an- tennas 400A-400N there are N means, each antenna branch having its own. Then the signals of all antenna branches are in turn multiplexed in a multiplexer 420, for example, in bursts from each antenna branch to a queue. After said multiplexing the channel estimator 408, the means 410, the means 412 and the detection part 414 are the same as in Figure 4A and there is only one common set of them for all antenna branches. Manufacturing costs of the receiver are thus saved.

The detection part 414 receives N sample bursts. The channel estimates corresponding to each burst are likewise applied to the detection part 414 from the means 412. Detected bits of about one burst come out from the detection part 414. The detection part 414 can be for example a VMLSE type (Vector Maximum Likelihood Sequence Estimation) described in article "Performance of the Vector MLSE Technique for Antenna Arrays in TDMA Mobile Systems" Escartin, Marko and Ranta, Pekka A. The channel estimator 408 and the means 410 can process each burst regardless of the antenna 400A-400N through which it is received.

The means 412 make decisions on the basis of all N antenna data and decide on interferences to be removed. The means 412 deliver to the detection means 414 with regard to the N burst the estimates of said interfer- ences with regard to the N antenna

Figure 5 shows a receiver which can also be implemented so as to receive a wide frequency band comprising a frequency area of three co- channels. The received frequency band is afterwards digitally filtered to the channels. Figure 6 shows such a receiver, filters 600, 602 and 604 each filter- ing one channel. The adjacent channel signal is transferred using a frequency transmission to a base frequency, the frequency transmission means being 606 and 608 in the Figure. Base frequency signals are multiplexed with one another in means 610 and are sequentially applied to the channel estimator 408 where channel estimates of a desired channel and co-channels are esti- mated for each signal. Then the strongest signal is detected in the detector 414 by which adjacent channel signal interference can be removed on other channels as shown in connection with the receiver in Figure 5.

The above described procedures are well suited to be used in future cellular network receivers. In order to apply them in full-scale to existing networks requires changing receivers at both base stations and subscriber terminals. The invention can be employed in current GSM-based systems in such a manner that the receiver to be used in the downlink direction is arranged to receive a signal modulated by a specific modulation method enabling a substantially better protection against interference caused by an adjacent channel than a conventional modulation method. The receiver is arranged to detect the signal modulated according to the specific modulation method in accordance with prior art, i.e. the subscriber terminal receiver does not require any changes. The receiver to be used in the uplink direction is arranged to receive the signal modulated by a conventional modulation method, for example by a GMSK modulation method, and the receiver is arranged to detect the sig- nal modulated according to the conventional modulation method in accordance with the invention. The changes are thus restricted to the base station transmitter and receiver. According to the tests performed by the applicant the OQAM method (Offset Quadrature Amplitude Modulation) is well applicable as the specific modulation method. The minimum C/l ratio allowed improves from -9 dB in a GMSK modulation method to -30 dB in the OQAM method.

Even though the invention has been described above with reference to the example of the accompanying drawings, it is obvious that the invention is not restricted thereto but can be modified in various ways within the scope of the inventive idea disclosed in the attached claims.

Claims

1. A signal detection method in a digital cellular network receiver, the method receiving on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel, characterized by determining jointly channel estimates of the useful signal and at least one adjacent channel signal, detecting the useful signal utilizing the channel estimates of the useful signal and the adjacent channel signal.

2. A method as claimed in claim 1, characterized by using reference parts in the useful signal and the adjacent channel signal known to the receiver when determining the channel estimates.

3. A method as claimed in claim 1, characterized by using symbol or bit decisions produced by decision feedback when determining the channel estimates.

4. A method as claimed in claim 1, characterized by employing a generated altered training sequence, in which a phase distortion caused by a frequency difference between the desired signal and the adjacent channel is taken into account, when determining the channel estimate for the adjacent channel signal.

5. A method as claimed in claim 1, characterized by taking into account the phase distortion caused by the frequency difference between the desired channel and the adjacent channel when detecting the useful signal in the reconstruction of the adjacent channel.

6. A signal detection method in a digital cellular network receiver, the method receiving on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel, characterized by determining channel estimates of the useful signal and at least one adjacent channel signal, reconstructing an adjacent channel signal from the useful signal utilizing the channel estimate of the adjacent channel signal, reducing the reconstructed adjacent channel signal from the useful signal, detecting the useful signal utilizing the channel estimate of the useful signal.

7. A method as claimed in claim 6, characterized by determining the channel estimates of the useful signal and at least one adjacent channel signal in parallel.

8. A method as claimed in claim 6, characterized by receiving in addition at least one channel adjacent to the desired channel, filtering the desired channel and the adjacent channel apart, transferring the adjacent channel using frequency transmission to the desired channel frequency, determining sequentially the channel estimates of the useful signal and at least one adjacent channel signal.

9. A method as claimed in any one of the preceding claims 6-8, characterized by using reference parts in the useful signal and the adjacent channel signal known to the receiver when determining the channel estimates.

10. A method as claimed in any one of the preceding claims 6-8, characterized by employing symbol or bit decisions produced by decision feedback when determining the channel estimates.

11. A method as claimed in any one of the preceding claims, characterized by asymmetrically selecting modulation methods for the downlink and uplink directions, and detecting the signal of at least one of the transmission paths by the method of the invention.

12. A method as claimed in claim 11, characterized by using in the downlink direction in the GSM based system a specific modulation method for the signal enabling a substantially better protection against interference caused by an adjacent channel than a conventional modulation method, and detecting the modulated signal according to the specific modulation method in accordance with prior art, and using the conventional modulation method, for example the GSMK modulation method, for the signal in the uplink direction and detecting the uplink signal according to the method of the invention.

13. A method as claimed in claim 12, characterized by the specific modulation method being an OQAM method (Offset Quadrature Amplitude Modulation).

14. A method as claimed in any one of the preceding claims, characterized by the reference part being a training sequence.

15. A method as claimed in any one of the preceding claims 1-13, characterized by the reference part being a pilot signal.

16. A method as claimed in any one of the preceding claims 1-13, characterized by the reference part being a broadband spreading code.

17. A digital cellular network receiver arranged to receive on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel, characterized by comprising a channel estimator (408) arranged to jointly determine channel estimates of the useful signal and at lest one adjacent channel signal, a detection part (414) arranged to detect the useful signal utilizing the channel estimates of both the useful signal and the adjacent channel signal.

18. A receiver as claimed in claim 17, characterized by the channel estimator (408) being arranged to use reference parts in the useful signal and the adjacent channel signal known by the receiver when determining the channel estimates.

19. A receiver as claimed in claim 17, characterized by the channel estimator (408) being arranged to use symbol or bit decisions produced by decision feedback when determining the channel estimates.

20. A receiver as claimed in claim 17, characterized by the channel estimator (408) being arranged to the adjacent channel signal when determining the channel estimate to generate an altered training sequence, in which a phase distortion caused by a frequency difference between the desired channel and the adjacent channel is taken into account.

21. A receiver as claimed in claim 17, characterized by the detection part (414) being arranged to take into account the phase distortion caused by the frequency difference between the desired channel and the ad- jacent channel when reconstructing the adjacent channel signal.

22. A digital cellular network receiver arranged to receive on a desired channel a combination of a desired useful signal and at least one interfering adjacent channel signal originating from a channel adjacent to the desired channel, characterized by comprising at least one channel estimator (408B) arranged to determine channel estimates of the useful signal and at least one adjacent channel signal, reconstruction means (500B) to reconstruct an adjacent channel signal from the useful signal utilizing the channel estimate of the adjacent channel signal, means (502B) to reduce the reconstructed adjacent channel signal from the useful signal, a detection part (414B) arranged to detect the useful signal utilizing the channel estimate of the useful signal.

23. A receiver as claimed in claim 22, c h a r a c t e r i z e d by including at least two channel estimators (408A, 408B) arranged to determine the channel estimates of the useful signal and of at least one adjacent channel signal in parallel.

24. A receiver as claimed in claim 22, c h a r a c t e r i z e d by being arranged to receive in addition to a desired channel at least one channel adjacent to the desired channel, and by comprising means (600, 602) to filter the desired channel and the adjacent channel apart, means (606) to transfer the adjacent channel by frequency transmission to the desired channel frequency, only one channel estimator (408) arranged to determine the chan- nel estimates of the useful signal and of at least one adjacent channel signal sequentially.

25. A receiver as claimed in any one of the preceding claims 22-24, c h a r a c t e r i z e d by the channel estimator (408) being arranged to use reference parts in the useful signal and the adjacent channel signal known by the receiver when determining the channel estimates.

26. A receiver as claimed in any one of the preceding claims 22-24, c h a r a c t e r i z e d by the channel estimator (408) being arranged to use symbol or bit decisions produced by decision feedback when determining the channel estimates.

27. A receiver as claimed in any one of the preceding claims 17-26, c h a r a c t e r i z e d by the receiver used in the downlink direction in the GSM-based system being arranged to receive a signal modulated by a specific modulation method enabling substantially better protection against interference caused by the adjacent channel than a conventional modulation method, and the receiver being arranged to detect the signal modulated according to the specific modulation method in accordance with prior art, and the receiver in the uplink direction being arranged to receive a signal modulated by the conventional modulation method, for example by a GMSK modulation method, and the receiver being arranged to detect the signal modulated according to the conventional modulation method in accordance with the inven- tion.

28. A receiver as claimed in claim 27, c h a r a c t e r i z e d by the specific modulation method being an OQAM method (Offset Quadrature Amplitude Modulation).

29. A receiver as claimed in any one of the preceding claims 17-28, characterized by the reference part being a training sequence.

30. A receiver as claimed in any one of the preceding claims 17-28, characterized by the reference part being a pilot signal.

31. A receiver as claimed in any one of the preceding claims 17-28, characterized by the reference part being a broadband spreading code.