WO2008065462A2

WO2008065462A2 - Method, transceiver and mimo communication system to obtain channel reciprocity

Info

Publication number: WO2008065462A2
Application number: PCT/IB2006/003343
Authority: WO
Inventors: Mattias Wennstrom
Original assignee: Huawei Technologies Co., Ltd.; Huawei Technologies Sweden Ab
Priority date: 2006-11-27
Filing date: 2006-11-27
Publication date: 2008-06-05
Also published as: WO2008065462A3; CN101444054B; CN101444054A

Abstract

The invention concerns a method for determining communication channel characteristics in a MIMO communication system, the system comprising at least one transceiver A having N antennas and at least one transceiver B having N antennas, the system using at least a first communication channel HAB from A to B and a second communication channel HBA from B to A. The method is distinguished in that it includes steps for rendering the communication channels HAB and HBA identical by the use of precoding. The invention also concerns a transceiver and a MIMO communication system implementing the method of the invention.

Description

Method, Transceiver and MIMO Communication System to

Obtain Channel Reciprocity

Field of the Invention

The present invention relates to wireless communication, further concerns a method for determining communication channel, a transceiver to perform the method and a MIMO communication system.

Background of the Invention

In wireless communication systems utilizing multiple antennas at both the transmitter and receiver sides, commonly known as multiple input, multiple output (MIMO) systems, the performance is greatly enhanced if the channel over which the communication is going to take place, is known beforehand to the transmitter. If this is the case, then many types of precoding can be applied, as have been suggested in the past, for instance pre-equalization [1] and singular value decomposition with water filling [2] in the single user MIMO case and dirty paper precoding [3] which also applies to the multi-user MIMO case. (Prior art references in brackets [] are listed at the end of the description.)

Initially, MIMO communication studies were focused on improving the performance of a single direction in a wireless link since it was believed that the throughput demands were different in the two link directions. However, there has recently been an increased interest in performing MIMO communication in both directions of the wireless link, here denoted as the uplink and the downlink, respectively. This two-way MIMO communication, which gives high throughput and/or highly robust performance in both directions, has value in wireless links where the demands are high for both directions.

In a flat fading, non-precoded MIMO system with N₇, transmitter antennas and

N_R receiver antennas, the input-output relation can be described as

y = Hx + n (1) In equation (1), x is the iV_r x l vector of transmitted symbols, y₅n are the N_R x l vectors of received symbols and noise, respectively, and H is the N_R x N_τ matrix of channel coefficients. The transmitter symbols are thus N_τ -fold spatially multiplexed over the MIMO channel H , or in other words, N_τ streams are transmitted in parallel, leading to a theoretically N₁. -fold increase in spectral efficiency.

Linear precoding implies that a N_τ x N_s precoding matrix W is introduced in

(1) to precode the symbols in the vector x . The column dimension N_s can' be selected smaller than N_τ in which case x is modified to dimension N_s x \ . Hence, N_s streams are transmitted in parallel. The input-output relation for a linear precoded

MIMO system is described as

y = HWx + n (2)

Also non-linear precoding can be implemented using for instance Tomlinson- Harashima precoding [7] where feedforward and feedback filter matrices are defined. See [8] for an overview of linear/nonlinear precoding.

To select the best precoder matrix W in (2), or the feedforward and feedback precoding matrices in non-linear precoding, knowledge about the channel H and the receiver noise statistics is necessary. Some knowledge of the forward channel quality is also required to set the correct modulation and code rate for systems using adaptive modulation and coding (AMC).

In frequency division duplex (FDD) wireless communication systems, the frequency distance between the forward and the reverse link is usually much larger than the coherence bandwidth of the wireless channel. This implies that the channel at the frequency where the transmission takes place is unknown and to use AMC and/or more advanced MIMO precoding techniques, feedback of information about the forward channel must be transmitted from the receiver to the transmitter on the reverse channel. This feedback increases the signalling overhead on the reverse channel and thus reduces the spectral efficiency of the system. Furthermore, when feedback is used, there is a risk of feedback errors which will severely degrade the performance of the MIMO precoding technique, Likewise, the MIMO transmission in the opposite link (reverse channel is now forward channel and vice versa) face the same overhead and feedback error problems.

The prior art can be divided into two categories. In the first category, the transmitter is informed about the MIMO channel coefficients and can thus apply any type of desired precoding. In the second category, the receiver, which has full knowledge of the MIMO channel coefficients, selects a precoder from a known set of allowed precoders and feeds back the index to this selected precoder.

In the prior art reference [4], belonging to the first category, the system consists of a base station with M transmit antennas and K single antenna users and the single direction link transmission from the base station to the users is considered. Each user estimates the M times 1 forward channel and it is then conveyed to the transmitter by modulating the reverse carrier with the estimated forward channel coefficients during a training period. The transmitter will obtain the knowledge of all the K users forward channels.

In [5] [6], also belonging to category one, the Direct Channel Feed Back (DCFB) method is introduced in a MIMO-OFDM system to feed back the forward channel coefficients by modulating reverse pilot sequences with the forward channel coefficients. Other reverse pilot sequences are used to estimate the reverse channel coefficients. The obtained forward channel coefficients can then be used in the transmitter to optimize the forward transmission.

In another prior art example, PCT/CN 2006001403, which belongs to category two, the approach is to use a codebook of allowed linear precoding matrices W. The receiver estimates the MIMO channels, selects a desired precoding matrix that in some sense is close to the ideal precoding matrix obtained from the eigenvalue decomposition of the MIMO channel and feeds back the index to the matrix in the codebook to the transmitter.

In the case of a system transmitting with multiple antennas both in uplink and downlink, the problems mentioned above are even more pronounced. If uplink and downlink channels are separate, optimal transmit and receive filters have to be calculated for both uplink and downlink channels. Thus, a twofold increase in calculation capacity is needed. This may entail a need for more expensive hardware and/or higher energy consumption in transceivers.

Summary

It is an object of the present invention to propose a solution for or a reduction of the problems of prior art. A main object is consequently to devise a method for determining communication channel characteristics in a MIMO communication system in both uplink and downlink such that calculation complexity can be held at a minimum and to minimise the need for communication channel properties feedback.

A corresponding transceiver and a corresponding MIMO communication system are also provided.

The main idea of the present invention is to utilize linear MIMO precoding to create an equivalent MIMO system where the uplink and downlink MIMO channels are identical (after a Hermitian transpose and up to a scalar constant) see Figure I₅ even though the original uplink and downlink MIMO channels are completely different, due to the frequency division duplex distance. This equivalent system can then, in a second step, be utilized by the well known advanced linear or non-linear precoding algorithms.

One of the differences of the current invention, compared to prior art references

[4] [5] [6], is that the MIMO communications is two-way and that no attempt to obtain knowledge of the "true" MIMO channel coefficients are made. Instead, by the use of linear precoding, we try to make the uplink and downlink MIMO channels equal, so that the uplink channel estimates can be used directly in optimization of the downlink.

The optimal transmit filters and the optimal receive filters may be a linear precoding matrix as in [1] or it may be a combination of a precoding matrix and a feedback filter, giving nonlinear precoding as in [8]. The calculation of optimal transmit and receive filter is coupled in the sense that the channel matrix must be known to the filter design algorithm, and the optimal filters for both the transmitter and the receiver sides are obtained simultaneously.

When the MIMO channel matrix is the same for uplink and downlink, or equivalently, in receiving mode and transmitting mode, large savings in computation power can be made since the calculations to find the optimal transmit filter also directly give the optimal receive filter. Hence, this optimization calculation need only be done once.

Furthermore, compared to prior art of category two, there is no need to quantize the precoding matrix to be close to a precoding matrix in the allowed set of precoding matrices in the codebook. Also, there is no need to feed back a precoding index which may be subject to errors when transmitted over the wireless channel.

To create the equal MIMO channels in the uplink and the downlink, linear precoding is performed in both uplink and downlink, as in expression (2). See Figure 2.

In order to implement the arrangement shown in Figure 1, each of the two transceivers will estimate the channel characteristics using the received data, and these channel characteristics will be used as a linear precoding on the transmitter side, preferably after a magnitude adjustment by multiplication with a scalar number.

An example is given in Figure 3. The uplink channel is estimated, and a scalar is found as a normalization constant. This uplink channel is then Hermitian transposed and used as a linear precoding in the downlink transmission. A similar approach is performed for the downlink channel. The equivalent system can be seen in Figure 4 where it can be seen that c_ULH_DLH^_L = [k • c_DLH_ULH_D ^* _L ) where & is a scalar constant.

When the equivalent MIMO system has been obtained as in Figure 4, these equivalent uplink and downlink channels can be estimated. Due to the special construction in this invention, for both ends in the link, the transmit channel is straightforwardly obtained as the Hermitian transpose of the estimated received channel. Hence, advanced linear or non-linear transmit precoding techniques that require channel knowledge can be applied on this equivalent system. Also, other techniques requiring knowledge of the channel can be applied on the equivalent system. Such techniques include adaptive modulation, adaptive coding and adaptive power control. Brief Description of the Drawings

Embodiments exemplifying the invention will now be described, by means of the appended drawings, on which

Fig. 1 illustrates a property of the invention to transform two different communication channels into two almost identical channels;

Fig. 2 illustrates the concept of precoding of the invention;

Fig. 3 illustrates an example of precoding the up and down link channels of the invention;

Fig. 4 illustrates the, due to precoding, obtained equivalent MIMO system;

Fig. 5 illustrates the estimation of communication channels using orthogonal reference signals;

Fig. 6 illustrates the use of an inner precoder of the invention.

Embodiments of the Invention

Assume a MIMO system consisting of two transceivers, denoted A and B operating over a wireless link, each transceiver equipped with N antennas for transmitting and receiving of information. The communication between A and B and in the opposite direction, between B and A, occurs at different carrier frequencies and are thus not interfering with the transmission between A and B.

To facilitate channel estimation to support the communication, 2N orthogonal reference signals are defined, denoted pilots. The 2N pilots are divided into two groups of N pilots, denoted "Pilot 1" and "Pilot 2". It should be noted that it is not a strict requirement for the reference signals to be orthogonal, but orthogonal reference signals enhances performance. It is preferred that reference signals at least have low cross correlation properties. When reference signals are orthogonal, they may e.g. be orthogonal in time and/or in frequency and/or in code.

To acquire the estimates of the MIMO channels, the set of N orthogonal pilot sequences (one for each transmit antenna) in Pilot 2 are used, see Figure 5. These pilots are used to estimate the "true" MIMO channels respectively, denoted H^ and

^HBA -

When these channels has been estimated as H _AB and H_BA respectively at each end A and B of the communication link, using the set of orthogonal reference signals in Pilot 2, normalization constants are found to compensate for the path loss. They can for example be calculated as

1 , 1

^CA = H - H and ^CB =

H BA H AB \

where ||»| is the Frobenius norm of the channel. The normalization is optional, but implies a significant improvement in system performance.

The linear precoding matrices are now non-obviously selected as

w_A = C_AH;_A

W_B = c_BH_M'

at A and B respectively. These precoding matrices are then applied at A and B respectively to obtain the equivalent channels, after precoding as

G_BA = H_BAW_B ^HuHu'

respectively. It should be noted that instead of the Hermitian transpose channel estimates //^• and ft' used above, the ordinary transposes TJ^T and fj^τ are also applicable. In fact, starting from the ordinary channel estimates transposes fj^τ and /f ^r , it is trivial that the Hermitian transpose would be obtained by simply taking the complex conjugate of each element of these channel estimates. If the channel estimation using "Pilot 2" has good performance, these two MIMO channels are now complex conjugate transposes, alternatively the ordinary transpose if f{^τ , Ff^τ are used, of each other, within a multiplication of a scalar constant.

AB BA

^GAB = k - ^GB^*A

The second set of orthogonal pilot sequences denoted "Pilot 1 ", which all are orthogonal to the members in the set of pilot sequences "Pilot 2" in Figure 5, are used to estimate these G_AB and G_BA on each side of the wireless link respectively. The estimates are denoted G_AB and G_BA . As before, orthogonality is not a strict requirement, but improves perfoπnance.

The update of the precoding matrices W _A and W_B must follow the channel dynamics. If the method is applied to an OFDM system, where pilots may be orthogonally separated in time and/or frequency, the matrices must be calculated with a frequency separation in the order of the channel coherence bandwidth and a time separation in the order of the channel coherence time.

Now, advanced precoding techniques such as eigenvalue decomposition with waterfilling of power over eigenmodes or non-linear precoding such as Tomlinson- Harashima precoding can be applied at the transmitter since at A, the A to B channel which is necessary for performing this precoding, is obtained as the Hermitian transpose of the estimated B to A channel, i.e. G_AB = k • G_BA where k is a constant.

Similarly, at B, the B to A channel, is obtained as the Hermitian transpose of the estimated A to B channel, i.e. G_BA = k' • G_A ^* _B where k' is a constant.

The input output relation of a linear precoded MIMO system is (from (2))

y = HWx + n

and when the receiver also use a linear filter R to estimate the transmitted vector x , we have

x = RHWx + n In [1] different optimizations for R and W are presented. As an example one application is given where the singular value decomposition is used for linear precoding, the square MIMO channel matrix H is decomposed as

H = UAV^*

where the columns in U and V contains the orthogonal and normalized left and right singular vectors respectively and the diagonal matrix Λ contains the singular values on its diagonal. If the transmit precoding matrix is selected as W = F and the receive filter is selected as R = U^* , an equivalent system can be written as

x = U^*UAV^*Vx + U^*n = Λx + U^*n

where it can be seen that the transmitted symbols in the vector x are completely decoupled (interference free) in the estimated vector x due to the diagonal Λ . This application require that the channel is known to both the transmitter and the receiver to be able to calculate the linear precoding matrices R and W . Furthermore, if the uplink and downlink channels are different, the computationally expensive singular value decomposition of the channel, H = UAV^* , must be performed for the uplink and downlink channel separately. In the present invention, this need only be computed once since the uplink and downlink MIMO channel matrices are equal.

On the other hand, in the present invention, a linear precoding of the transmitted symbols must be performed as in Figure 2 to create an equivalent channel in uplink and downlink. After the "outer" precoding, for instance using the singular value decomposition described above, the precoded symbols must be precoded again using an "inner" precoder, to create the equivalent uplink and downlink channels, see Figure 6. So a drawback with the present invention compared to prior art is that an "inner" precoding structure is needed. However, the inner precoding involves a matrix to vector multiplication which has much lower computation complexity than a singular value decomposition, see a comparison in Table 1.

The singular value decomposition has a complexity that grows in the cubic of the number of antennas JV, while the growth for a matrix-vector multiplication is only square. Therefore, exchanging singular value decomposition with a matrix-vector multiplication implies great savings in computational complexity. Table 1 Number of operations needed to perform singular value decomposition and a matrix- vector multiplication where N is the size of the square matrix.

REFERENCES

[1] A. Scaglione, P. Stoica, S. Barbarossa, G. B. Giannakis, and H. Sampath, "Optimal designs for space-time linear precoders and decoders", IEEE Trans. Signal Processing, vol. 50, pp. 1987-2006, July 1999.

[2] I. E. Telatar, "Capacity of Multi-Antenna Gaussian channels", European Transactions on Telecommunications, vol. 10, no. 6, pp. 585-595, Nov./Dec. 1999.

[3] U. Erez and S. ten Brink, "A Close-to-Capacity Dirty Paper Precoding Scheme", IEEE transactions on information theory, vol.51, No.10, October 2005, p.3417-3432.

[4] T.Marzetta and B.Hochwald, "Fast transfer of channel state information in wireless systems", submitted to IEEE Transactions on Signal Processing, 2004, http.7/mars.bell-labs.com/papers/channel_estimation/FDDvsTDD.pdf

[5] IEEE802.16e-04/422, "Improvements to the uplink channel sounding for OFDMA", IEEE BWA WG, 2004-04-11, http://www.ieee802.org/16/tge/contrib/C80216e-04_422.pdf

[6] 3GPP Rl -050516, "Additional details on DCFB for obtaining MIMO channel information at Node B", Motorola, Seoul, Korea, November 2005.

[7] R. F. H. Fischer, C. Windpassinger, A. Lampe and J. B. Huber, "Space-time transmission using Tomlinson-Harashima precoding", Proc. of 4 ITG Conference on Source and Channel Coding, pp. 139-147, January 2002. [8] O.Simeone, Y-Bar-Ness and U.Spagnolini, "Linear and nonlinear preequalization/equalization for MIMO Systems with long term channel state information at the transmitter", IEEE Transactions on wireless communications, p. 373-378, Vol3. No.2, March 2004.

[9] P.Vandenameele, L. van Der Perre, B.Gyselinckx, "An SDMA algorithm for High-Speed WLAN Performance and complexity", Global Telecommunications Conference, pp. 189-194, November 1998.

Claims

1. Method for determining communication channel characteristics in a multiple input, multiple output (MIMO) communication system, the system comprising at least one transceiver A having a plurality of antennas and at least one transceiver B having a plurality of antennas, said system being arranged to use at least a first communication channel H_AB from transceiver A to transceiver B and a second communication channel H _BA from transceiver B to transceiver A, the method comprising sending N reference signals, from a first set of reference signals, on said first communication channel H_ΛB , using N antennas of the plurality of antennas, one reference signal on each antenna, receiving in transceiver B the reference signals, sent from transceiver A, on N antennas of transceiver B of the plurality of antennas, and obtaining in transceiver B the channel H_M characteristics estimate H_AB on the basis of the N reference signals received, characterised in that the method further includes sending N reference signals, from the first set of reference signals, on said second communication channel H_BΛ , using N antennas of the plurality of antennas, one reference signal on each antenna, receiving in transceiver A the reference signals, sent from transceiver B, on N antennas of transceiver A of the plurality of antennas, obtaining in transceiver A the channel H_BΛ characteristics estimate H_BA on the basis of the Preference signals received, using the transpose H\_A as a precoding matrix for the first channel H_M and thereby creating a channel G_AB = H^ H\_Λ, using the transpose H' '_M as a precoding matrix for the second channel H_BA and thereby creating a channel G_BA = H_BA H^TM, sending N reference signals, from a second set of reference signals, on the created channel G_AB , using N antennas of the plurality of antennas, one reference signal on each antenna, sending N reference signals, from the second set of reference signals, on the created channel G_BA , using N antennas of the plurality of antennas, one reference signal on each antenna, receiving in transceiver A the reference signals, sent from transceiver B on channel G_BA , on N antennas of transceiver A of the plurality of antennas, receiving in transceiver B the reference signals, sent from transceiver A on channel G_AB , on TV antennas of transceiver B of the plurality of antennas, obtaining in transceiver A the channel G_BA characteristics estimate G_BA on the basis of the Preference signals received, obtaining in transceiver B the channel G_M characteristics estimate G_AB on the basis of the Preference signals received, obtaining in transceiver A the estimate G_AB as Q ₌ h . G^T where k is a constant, and obtaining in transceiver B the estimate G_BA as Q — fc' . _Q ⁷ where k' is a constant.

2. Method according to claim 1, wherein the channel estimates H_AB and H_BA are normalised to compensate for path loss.

3. Method according to claim 1 or 2, wherein the transposes H^r _ω and H\ are additionally complex conjugated so that channels G_M and G_BA becomes

G_M = H_MHl = Hβ\_A and Q_M = H_BΛW_M = H_ε H\_m-

4. Method according to claim 1, 2 or 3, wherein the transposes fr and fr are additionally complex conjugated so that the obtained channel estimates G_AB and G_BA becomes _{G = k} . _Q ^r _{= JC} . Q- and Q = k' - G^τ = k' - G' ■

5. Method according to any of claims 1-4, wherein the first set of reference signals are orthogonal, the second set of reference signals are orthogonal and the first set of reference signals are orthogonal to the second set of reference signals.

6. Method according to any of claims 1-5, wherein the first and second set reference signals are orthogonal in time.

7. Method according to any of claims 1-6, wherein the first and second set reference signals are orthogonal in frequency.

8. Method according to any of claims 1-7, wherein the first and second set reference signals are orthogonal in code.

9. Method according to any of claims 1-8, wherein knowledge of channel estimates G_AB , G_BA in any of transceivers A and B is used for communication properties improvement purposes.

10. Method according to claim 9, wherein knowledge of channel estimates G_AB , G_BA is used for adaptive modulation.

11. Method according to 9 or 10, wherein knowledge of channel estimates G_M ,

G_BA is used for adaptive coding.

12. Method according to any of claims 9-11, wherein knowledge of channel estimates G^ , G_BA is used for adaptive power control.

13. Method according to claim 12, wherein the power control is performed using a water filling method.

14. Method according to any of claims 8-13, wherein knowledge of channel estimates G_AB , G_BA is used for precoding.

15. Method according to claim 14, wherein the precoding is of linear type.

16. Method according to claim 15, wherein the precoding is of non-linear type.

17. A transceiver configured to perform the method according to any of claims 1 -

16.

18. MIMO communication system comprising at least one transceiver A having a plurality of antennas and at least one transceiver B having a plurality of antennas, said system being arranged to use at least a first communication channel H _AB from transceiver A to transceiver B and a second communication channel H_BA from transceiver B to transceiver A₅ whereby the transceiver A is arranged to send N reference signals, from a first set of reference signals, on said first communication channel H₄₃ , using N antennas of the plurality of antennas, one reference signal on each antenna, the transceiver B is arranged to receive the reference signals, sent from A, on N antennas of the plurality of antennas, the transceiver B is arranged to obtain the channel H _AB characteristics estimate H_AB on the basis of the N reference signals received, characterised in that the transceiver B is arranged to send N reference signals, from the first set of reference signals, on said second communication channel H_BA , using N antennas of the plurality of antennas, one reference signal for each antenna, the transceiver A is arranged to receive the reference signals, sent from B, on N antennas of the plurality of antennas, the transceiver A is arranged to obtain the channel H_BA characteristics estimate H_BA on the basis of the N reference signals received,

the transceiver A is arranged to use the transpose

as a preceding matrix for the first channel H_M and thereby creating a channel G_M = H_AB H^rω,

• the transceiver B is arranged to use the transpose H^TM as a precoding matrix for the second channel H_BA and thereby creating a channel G_BA — H_BΛ H^τ _^, the transceiver A is arranged to send N reference signals, from a second set of reference signals, on the created channel G_AB , using N antennas of the plurality of antennas, one reference signal on each antenna, the transceiver B is arranged to send N reference signals, from the second set of reference signals, on the created channel G_BA , using N antennas of the plurality of antennas, one reference signal on each antenna, the transceiver A is arranged to receive the reference signals, sent from transceiver B on channel G_BA , on N antennas of transceiver A of the plurality of antennas, the transceiver B is arranged to receive the reference signals, sent from transceiver A on channel G^ , on N antennas of transceiver B of the plurality of antennas, the transceiver A is arranged to obtain the channel G_BA characteristics estimate G_BA on the basis of the reference signals received, the transceiver B is arranged to obtain the channel G_M characteristics estimate G_AB on the basis of the reference signals received,

the transceiver A is arranged to obtain the estimate AB as *> ^" ^, where k is a constant,

the transceiver B is arranged to obtain the estimate BA _as ^yjru ^{~ κ " 0}^, where

"^• is a constant.

19. MIMO communication system according to claim 18, wherein the transceivers A and B are arranged to additionally complex conjugate transposes

and H^TM _} so that channels G_AB and G_BA becomes ■*» ^{— /7} ^ -" M ^{~ n} _*>"^■ IΛ and G_u = HjF_m = Hjϊ_M ^' .

20. MIMO communication system according to claim 18 or 19, wherein the transceivers A and B are arranged to additionally complex conjugate transposes fr and _Qi- , so that the obtained channel estimates G_AB and G_BA becomes

G = * • W = k - G- and G = /t'.5⁷ = yt' - G;-

21. MIMO communication system according to claim 18, 19 or 20, wherein the system is arranged to use knowledge of the estimates G_M , G_BA , in any of transceivers A and B for communication properties improvement purposes.