Description

CHANNEL ESTIMATION SELECTION BETWEEN DIFFERENT SCHEMES

The present invention relates to a method and receiving apparatus for estimating a channel in a communications system.

In a communications system, data is transmitted over channels in the form of space/time coded symbols.

Enhanced UTRA (E-UTRA) is the long term evolution (LTE) of the Universal Mobile Telecommunications System (UMTS) currently standardized in the 3rd Generation Partnership Project (3GPP) .

In a receiver of an Orthogonal Frequency Division Multiplex (OFDM) -based mobile radio transmission system (E-UTRA, WiMAX etc.) an estimation of the channel transfer function has to be performed. This estimate is used by the equalizer to remove the influence of the channel on the transmitted symbols.

To support the channel estimation, pilot symbols are transmitted on the sub-carriers at certain time steps. These pilot symbols are known at the receiver.

Fig. 1 (prior art) shows the time arrangement of OFDM symbols for the case of the E-UTRA system. The figure shows a time sequence of eight OFDM symbols, building a sub frame, i.e. a unit of data to be transmitted. Each OFDM symbol consists of multiple orthogonal frequency carriers. The sub frame consists of short two blocks SB-I, SB-2 containing pilot symbols and six long blocks LB-I to LB-6 containing data. Each of the OFDM symbols LB-I to LB-6 and SB-I, SB-2 has a preceding cyclic prefix time interval CP to avoid inter-symbol interference .

For the time instants where SB-I and SB-2 are received the current channel transfer function can be estimated by well known methods like Maximum Likelihood Estimation or Minimum Mean Square Error Estimation [I].

To detect and reconstruct the data symbols which are transmitted at LB-I to LB- 6, estimates of the channel transfer function for the corresponding time instants are necessary. These intermediate channel estimates corresponding to the channel states at time instants LB-I to LB-6 have to be derived from the channel estimates calculated at the time instants of SB-I and SB-2 using the transmitted pilot symbols.

This task of deriving channel estimates at the intermediate positions in time has been quite well researched for the case when the pilot symbols are identical, i.e. both SB are concentrated over the sub-carriers allotted to the user (case 1) .

The E-UTRA specifications [4] however include the possibility of the case when one of the SB is spread over the entire system spectrum so as to allow channel dependent frequency domain scheduling (case 2) .

In case 1, both SB, i.e. both transmitted pilot symbols, are concentrated over a chunk. In this case, the following solution is known. Given the sub frame structure in Fig.l, there exist 2-D FIR Wiener filters that are optimal in sense of minimizing the mean square error (MSE) [2] . These methods require in general channel scattering functions. By accepting channel estimates at pilot sub-carriers as their input, these methods yield channel estimates over the entire time and frequency grid.

In order to reduce receiver complexity, it is possible to adopt a cascaded interpolation approach. In this approach, first a minimum mean-square error (MMSE) based frequency do-

main interpolation is performed to arrive at optimal channel estimates at pilot symbol positions. Next, a suitable interpolation along time is carried out to yield channel estimates for the data symbols. Linear interpolation respectively extrapolation is sufficient. This approach has been shown to perform reasonably well as compared to the joint 2-D interpolation [3] along with significantly reduced complexity.

In case 2, one of the SB, i.e. one of the two transmitted pilot symbols, is spread over the entire spectrum.

As mentioned above, there exist no analytical solutions for this case when one of the SB is spread over the entire spectrum. It is important to note that owing to the power constraints and consequent decreased effective Signal-to-noise ratio (SNR) at relevant data sub-carriers, the spread SB leads to significant loss for channel estimation at data sub- carriers .

It is an aim of the invention to address the problems discussed above.

Said problems are solved by the features mentioned in the independent claims. Preferred embodiments of the invention are described in the dependant claims.

The present invention provides a method for estimating a channel in a communications system. In the communications system, data is transmitted in at least one unit. The channel is estimated based on a first pilot symbol and on a second pilot symbol, resulting in a first pilot estimation output and a second pilot estimation output. For each received unit, the method comprises the following steps: In a first step, a first channel estimation is determined by utilizing said first and second pilot estimation output according to a first channel estimation scheme. In a second step, a second channel estimation is determined by utilizing said first and second

pilot estimation output according to a second channel estimation scheme. In a third step, the method selects between said first and said second channel estimation scheme to be applied based on a decision metric. The decision metric quantifies a distance between expected data and data equalized based on said first and second channel estimation.

Furthermore, the present invention provides a receiving apparatus for executing the presented method and an according communications system.

The proposed invention provides the advantage that the overall system performance is significantly improved. The invention applies equally well to both cases described above and leads to performance improvements in terms of bit error rate.

Preferred embodiments of the present invention will now be described with reference to the accompanying drawings in which :

Fig. 1: shows a sub frame structure for OFDM (prior art)

Fig. 2: shows a flow chart for channel estimation and equalization

Fig. 1 shows a sub frame structure for OFDM as described above .

Without loss of generality, in the following it is assumed that SB-I is concentrated over the chunk and SB-2 is spread over the entire spectrum for the purpose of scheduling.

Depending upon the channel time variance scenarios, there are two different competing schemes for channel estimation. Both schemes are described in the following.

The first scheme relates to relatively high channel time variance, e.g. to high vehicular speeds such as 120 km/hr or even 300 km/hr. Channel tracking is necessary, and an interpolation between SB-I and SB-2 is a must, as the channel variations are high due to the high vehicular speeds. For the above described case 2 with SB-I and SB-2 being spread over the entire system spectrum, this first scheme means that incorporation of the poor quality estimate from SB-2 is inevitable at high vehicular scenarios.

The second scheme relates to time invariant or slow time varying channels. For the second scheme, channel estimation performance can be improved as follows:

• For case 1 : averaging the channel estimates obtained from SB-I and SB-2. This leads to significantly increased performance at low SNRs where this averaging can lead to effectively reduced noise.

• For case 2 : ignoring the poor channel estimate obtained from SB-2. Because the channel stays almost the same during the entire sub frame, the accurate SB-I channel estimate can be used for equalization of all LBs without incurring any loss due to absence of channel tracking.

The inventive method comprises the step of making a decision between one of the two proposed schemes for each sub frame. Since the two schemes perform well only in exclusive high vehicular motion and low vehicular motion scenarios, a combination of the two approaches results in a significantly improved performance as compared to either one.

The decision strategy is based on the soft equalizer outputs and works as follows.

1. In a first step, one LB (e.g. LB-5) will be equalized using the channel estimate according to the first scheme, producing the output X _schem ei •

2. In a second step, the same LB will then be equalized using the channel estimate according the second scheme, producing Output X _sc h _e m _e 2 •

3. In a third step, a decision metric D is calculated, which represents a measure of the mean detection reliability of all bits of the received LB under the two schemes .

Any distance measure between the received and expected received bits of the at least one LB used in the first and second step can be used, or any measure which is proportional to the log-likelihood-ratio values of the bits. The decision metric that we use can be written as

hereby

• N is the number of bits

• x(n) is the soft output assigned to the nth bit

• y(n) is the expected received value assigned to the nth bit.

In another refinement of the method, instead of using a data symbol (e.g. LB-5) as input for the above described first and second step of equalizing, a pilot signal SB-I, SB-2 is used as input for the equalizer. In this case, the decision metric is advantageously modified as follows:

In this case, is it not necessary to use absolute values for x(n) and y(n) as the bits of the pilot signal SB-I, SB-2 are exactly known at the receiver. This refinement offers more robustness of the method with respect to a bad SNR.

The decision metric has to be modified according to the used modulation scheme in order to get the bitwise distance measures .

Under the assumption that the described distance measure D is proportional to the bit error probability, the decision will be done in this way:

• If D _schemel is larger than D _scheme2 the channel estimates derived from scheme2 will be used for equalization.

• If D _schemel is less than D _scheme2 the channel estimates derived from schemel will be used for equalization.

Inserting a factor F in the decision expression

L-* schemel ^ ' LJscheme2

allows, for instance, adjusting the switching behavior in accordance with the special system requirements. For example, F < 1.0 leads to a more frequent activation of the interpolation mode and a better performance at high mobile velocity, but to degradation at low velocities and low SNR.

Fig. 2 shows a flow chart for channel estimation and equalization according to the invention.

In the flowchart shown in Fig. 2 the overall channel estimation and equalization process is described for one sub frame according to the invention with the sub frame being structured as shown in Fig. 1. In this case, the pilot symbols on the second short block SB-2 are spread over the full bandwidth.

In a first step 1 of the method shown in Fig. 2, a first channel estimation based on the first pilot symbol received in the first short block SB-I is performed. The output of the first step 1 is a first estimation output H _SB1 .

In a second step 2, a second channel estimation based on the second pilot symbol received in the second short block SB-2 is performed. The output of the second step 2 is a second estimation output H _SB2 .

In a third step 3, one of the symbols transmitted in a long block LB-I to LB-6 is equalized using the first estimation output H _SB1 . In the case shown in Fig. 2, the symbol transmitted in the fifth long block LB-5 is equalized. In this case, the equalization accords to the above described second scheme as only the received symbol of the first short block SB-I is taken into account. The output of the third step 3 is a first equalized output X _L B5,scheme2 •

In a fourth step 4, an interpolation along time to yield channel estimates for the data symbols in the long blocks LB- 1 to LB-6 is performed, using the first estimation output H _SB1 and the second estimation output H _SB2 as input. The output of the fourth step 4 consists of a set of estimation outputs H _LB1 to H _LB6 with one estimation output per each long block LB-I to LB-6.

In a fifth step 5, the symbol transmitted in the fifth long block LB-5 is equalized again, this time using the estimation output according to the fifth block H _LB5 calculated in the fourth step 4. In this case, the equalization accords to the above described first scheme as an interpolation based on the first estimation output H _SB1 and the second estimation output H _SB2 is performed. The output of the fifth step 5 is a second equal i zed Output XLB5, schemel •

In a sixth step 6, a first decision metric D _schemel is calcu ^¬ lated based on the second equalized output X _LB S,schemei •

In a seventh step 7, a decision metric D _scheme2 is calculated based on the first equalized output X _LB S,scheme∑ •

In an eighth step 8, a decision is done in the following way:

• If D _schemel is larger than D _scheme2 equalization of the long blocks LB-I to LB-6 will be done based on the first estimation output H _SB1 (first variant of a ninth step 9a)

• If D _schemel is less than D _scheme2 equalization of the long blocks LB-I to LB-6 will be done based on the set of estimation outputs H _LB1 to H _LB6 (second variant of a ninth step 9b) .

In a further refinement of the method, Decision Directed Channel Estimation (DDCE) [3] is introduced. DDCE allows a further performance gain. In this refinement, the detected symbols at the output of the equalizer will be assumed as error free and processed in the same way as the known pilot symbols .

Considering the flow chart in Fig. 2 this means that DDCE would be applied to the output of the third block 3 and the fifth block 5, producing new channel estimates for schemel and scheme2. Thereupon the equalization processes, the third block 3 and the fifth block 5 will be repeated with new channel estimates. The calculation of the decision metric values, shown in the sixth block 6 and the seventh block 7, will be done with the new equalizer output.

The presented method allows overall performance gains with respect to the bit error rate. The combination of the two approaches presented as the first scheme and the second scheme results in a significantly improved performance as compared to either one. The presented method is of a very low complexity scheme with almost the same performance as known schemes without switching. The presented method performs especially well if there are differently strong pilot symbols.

The invention is applicable to any transmission system where intermediate channel estimates between pilot symbols or training sequences are used for equalization.

References :

[1] A Comparison of Pilot-Aided Channel Estimation Methods for OFDM Systems, M. Morelli, IEEE Trans, on Signal Proc, Vol.49, NO 12, 2001

[2] OFDM and MC-CDMA, L.Hanzo, Wiley, 2003, Chap.14.6 and 14.7

[3] OFDM and MC-CDMA, L.Hanzo, Wiley, 2003, Chap.15.2

[4] Physical Layer aspects for E-UTRA, 3GPP TR 25.814