WO2003017607A2

WO2003017607A2 - Iterative calculation of coefficients for a multicarrier equaliser

Info

Publication number: WO2003017607A2
Application number: PCT/CA2002/001251
Authority: WO
Inventors: Octavian V. Sarca
Original assignee: Redline Communications Inc.
Priority date: 2001-08-14
Filing date: 2002-08-13
Publication date: 2003-02-27
Also published as: EP1419628A2; WO2003017607A3; AU2002322889A1

Abstract

An multicarrier adaptive channel equalizer is provided, which has a coefficient adaptation branch that updates the filter coefficients based upon a weighted average of previous coefficient values and decisions for the current symbol. An override or nulling of the weighting factor is provided to mitigate the effects of noise on the current symbol. An embodiment of the OFDM adaptive channel equalizer requires only one coefficient memory port, and only two multipliers as in a non-adaptive equalizer thereby allowing a cost effective hardware implementation.

Description

OFDM DECISION-FEEDBACK ADAPTIVE CHANNEL EQUALIZER

Field of the Invention

The present invention relates to decision feedback adaptive channel equalizers and is particularly concerned with equalizers for orthogonal frequency division multiplexer (OFDM) channels

Background of the Invention

It's well known to provide channel equalization in transmission systems to compensate for channel response. Theoretically, by determining the channel response, and then multiplying the received signals by the inverse of the channel response, its effects can be eliminated. In radio transmission systems this is accomplished by transmitting a known reference signal and determining from the received reference signal, a channel response. The inverse channel response is then applied to subsequently received signals.

In practice this approach only works for slowly varying channels that are not subjected to noise bursts during transmission of the reference signals. When channels vary more rapidly either more frequent reference signals are required or more complex equalizers are needed that can adapt the channel response between reference signals. However, use of additional reference signals reduces channel efficiency and more complex equalizers may lead to implementation difficulties and increased hardware costs.

Summary of the Invention

An object of the present invention is to provide an improved OFDM adaptive channel equalizer. According to an aspect of the present invention there is provided a method of OFDM channel equalization comprising the steps of: deriving a coefficient value from a received reference value; receiving subsequent symbols in dependence upon the coefficient value; and updating the coefficient value in dependence upon a weighted average of an optimal coefficient value for the current symbol and the previous coefficient value.

According to another aspect of the present invention there is provided a means for deriving a coefficient value from a received reference value; means for receiving subsequent symbols in dependence upon the coefficient value; and means for updating the coefficient value in dependence upon a weighted average of an optimal coefficient value for the current symbol and the previous coefficient value.

According to a further aspect of the present invention there is provided a first branch for deriving a channel coefficient value from a received reference value; a second branch for deriving an equalized output value in dependence upon the coefficient value and for adapting a previous channel coefficient value in dependence upon a weighted average of the coefficient value and an optional coefficient value for a current equalized output value and received signal; and an output for the equalized output value.

Brief Description of the Drawings

Fig. 1 illustrates in a functional block diagram a known OFDM transceiver;

Fig. 2 illustrates in a functional block diagram a known OFDM non-adaptive channel equalizer;

Fig. 3 illustrates in a functional block diagram a known OFDM adaptive channel equalizer; Fig. 4 illustrates in a functional block diagram an OFDM decision-feedback adaptive channel equalizer in accordance with an embodiment of the present invention; and

Figs. 5a and 5b graphically illustrate weighting factor nulling criteria for

QPSK and 16 QAM for the equalizer of Fig. 4.

Detailed Description of the Preferred Embodiment Referring to Fig. 1 there is illustrated in a functional block diagram a known

OFDM transceiver. The OFDM transceiver 10 includes an OFDM transmitter 12, an OFDM receiver 14, a data input 16, a data output 18 and an input/output port 20. The OFDM transmitter 12 includes a scrambler 22, a channel coder 24, an interleaver 26, a symbol mapper 28, and inverse fast Fourier transform (IFFT) 30 and a guard interval insertion 32. The OFDM receiver 14 includes a synchronization and guard interval stripper block 34, a fast Fourier transform (FFT) 36, an equalizer & sheer 38, a deinterleaver 40, a channel decoder 42, and a descrambler 44.

In operation, the data input at 16 is first scrambled by the scrambler 22 to ensure that the modulated signal will exhibit a flat spectrum similar to white noise. The scrambled data is passed through the channel coder 24 that performs forward error correction (FEC) encoding. The encoded data in interleaved by the interleaver block 26 to ensure frequency and eventually time diversity and mapped into an OFDM symbol (onto different carriers) by symbol mapper 28. The OFDM symbol, which is list of complex (IQ modulated) values on different carriers, is then converted from the frequency domain to the time domain by the inverse fast Fourier transform

30. The IFFT 30 output is pre-pended with a guard interval (GI) or a cyclic prefix by GI Insertion 30 and sent to RF conversion and amplification and to antenna via the input/output port 20.

The OFDM receiver 14 implements the reverse operations in the reverse order of the OFDM transmitter 12. The GI stripper 34 removes the GI, the FFT 36 converts from time to frequency domain, the equalizer & slicer 38 performs symbol de- mapping. In addition to these functions, the receiver ensures time and frequency synchronization with the transmitter in the synchro block 34 and compensates for the channel impairments by use of the equalizer 38.

Referring to Fig. 2 there is illustrated a known OFDM non-adaptive channel equalizer. An OFDM channel equalizer is required to invert the channel frequency response. The OFDM spreads data on several carriers thus equalization is needed on each carrier separately. OFDM with cyclic prefix (also called guard interval) avoids inter-symbol interference by setting the duration of the cyclic prefix longer than the channel impulse response. In such a case, the channel equalizer is reduced to one-tap filter on each carrier. The OFDM non-adaptive channel equalizer 50 includes an input 52 for received complex values X_m>n and an output 54 for received sliced data

Y_m,_n, where m indicates the index of the OFDM symbol and n the index of the subcarrier. A received reference value Xo,_n is applied to a complex invertor 56, while remaining received complex values are applied to a complex multiplier 58. The output from the inverter 56 is applied to a second complex multiplier 64 having the other input set to the known transmitted reference value Yo,_n and the output to the coefficient C₀,_n is stored in memory 60. The complex multiplier 58 takes the remaining received complex values Cι,_n, C₂,_n, . and multiplies them with the coefficient C₀,_n from memory 60. The output of mixer 58 is applied to slicer 62, the output of which is applied to the equalizer output 54. The output from the inverter 56 is applied to a second mixer 64 having input for a known transmitted reference value

Y₀,n and outputs for coefficients Co,_n

There are two techniques to send reference information for the equalizer. One is scattered pilots. Each reference signal (called pilots) is sent on a subset of carriers in each symbol. Positions of the pilots change from symbol to symbol such that, on each carrier, a reference is sent periodically. The second technique is preamble or midamble technique in which reference signals are sent on all carriers in a special symbol. That is a set of reference signals that precludes the data-carrying OFDM symbols (preamble) and/or that are inserted periodically between data-carrying OFDM symbols (midamble). Embodiments of the invention may be applied to both methods thus, for simplicity, the description concentrates on what happens on a single subcarrier denoted here by n. In operation, on each carrier, the reference signals (pilots or pre/mid-ambles) are used to initialize the channel equalizer. If Xo,_n is the received complex reference value on carrier n and Yo_,n is the known transmitted reference value on the same carrier, then the channel equalizer coefficient is initialized with Co,_n ⁼Yo,_n/Xo,_n in a non-adaptive channel equalizer as shown in Fig. 2. The equalizer coefficient initialized during reference symbols on the upper branch is used unchanged to correct the following data symbols on the lower branch until another reference symbol is received. In such an equalizer, the sheer input for data symbol m and carrier n is computed as Co,_nXm,n and the sheer output is Y_m)n = T(Co, Xm,n)-

The known non-adaptive channel equalizer has several drawbacks. Noise from the reference symbol adds to the noise of data symbols through the equalizer coefficient. A noise burst during one reference symbol affects the overall signal to noise ratio (SNR) for all data symbols until the next reference symbol. Also, in variable channels, reference symbols must be sent more often, creating a significant overhead, hence reducing channel efficiency.

Referring to Fig. 3 there is a known adaptive channel equalizer using a least means squared algorithm. The OFDM channel equalizer of Fig. 3 adds further components to the second branch in order to update the coefficients for every received data symbol based on the difference between the output and input of the sheer 62. The lower branch contains the subtracter 76 that calculates the difference between inputs 72 and 74. The output of subtracter 76 is applied to a complex multiplier 78 with connection 80 from the input to provide an output at 82 applied to a μ weighting function 84, the output of which is summed in adder 86 with the previously stored coefficients via 88 to provide updated coefficients, stored via 90.

In operation, the sheer 62 input is calculated as C_m-ι_)nX_m>n> where coefficients

C_m,_n vary from symbol to symbol. If Y_m,_n ⁼T(C_m-ι,_nXm_)n) is the sheer output for the symbol m in carrier n then the coefficient C_m,_n is calculated from C_m-ι,_n and the difference (Y_m,_n-C_m.i,„X_ra,n).

An adaptive equalizer can track variable channels. It also has advantages in fixed channels since it takes the initial estimate of the channel response obtained from the reference symbol and refines it during the data symbols. The effective signal noise (SNR) continuously improves during the data symbols. The adaptation improves system immunity to noise in general and especially to noise bursts that occur during a reference symbol. A typical adaptive equalizer such as shown in Fig. 2 employs an LMS-like algorithm for adaptation. In such an algorithm, the coefficients are updated based on the rule C_n,_]n=C_m.ι,„+μ(Y_m,n-C_m,„X_m,n)X_m,„.

The cost of introducing the LMS-adaptation is significant since it requires an additional multiplier, two adders, a constant multiplication and an additional access to the coefficient memory. The multiplier brings most of the added cost in a hardware implementation. Also, C_m-ι_;n and X_m)n are used in two different places in the equalizer structure. In a pipeline hardware architecture this double usage requires either additional registers or double accesses to the input and coefficient memories to compensate for the pipeline delay. LMS-like algorithms can become unstable for large values of μ, thus μ has to be much smaller than 1 (e.g. typically less than 0.01).

Since equalizer tracking speed is proportional to μ, stability concerns limit the tracking speed of the equalizer and consequently the performance improvements could be provided by the equalizer.

Referring to Fig. 4 there is illustrated an OFDM channel equalizer in accordance with an embodiment of the present invention. The OFDM channel equalizer 100 includes an input 52 and an output 54, an upper branch including an inverter 102 and a multiplier 104 and a coefficient output 106 and a lower branch including a multiplier 108 and an input from coefficient memory 110. Output from the multiplier 112 is applied to a sheer 114 whose output is coupled to the equalizer output 54. Input to the sheer 114 is subtracted in the subtractorl 18 from the output from the slicer 114. The output of the subtracter 118 is applied to a μ weighting function 120 that also receives output via 122 from the slicer 114. The weighted output is applied to a second adder 126 to which is applied the output from the multiplier 108 via 124. The output from the second adder 126 is applied to a multiplexer 128 whose output 132 is applied to the multiplier 104.

In operation, the adaptive equalizer of Fig. 4 employs the update rule C_m,_n = C_m-i,_n+ μ(Ym,_n/Xm,n-C_m.ι_;n). This rule calculates the updated coefficient value as a weighted average between the previous coefficient value and the optimal coefficient for the current symbol. Note that μ can take any value between 0 and 1 without compromising stability. For μ = 0, the coefficient value is not updated at all, while for μ=l, the updated coefficient value is equal to the optimal coefficient for the current symbol and does not depend upon the previous value. To increase the tracking speed of the equalizer μ should be increased close to 0.5. This improves the immunity of the equalizer to noise bursts during the reference symbol, improves the overall signal to noise ratio and the ability to track channel changes. However it could make the equalizer sensitive to noise bursts during data symbols. To avoid this problem, the adaptive equalizer of Fig. 4 does not update a coefficient when the corresponding received data symbol is too noisy. Received symbols are considered too noisy when they do not fall within the given neighbourhood of one of the ideal constellation points. An example of such neighbourhoods for QPSK and 16QAM are depicted in Fig. 5 a and 5b, respectively. When the slicer 114 detects the symbols that are too noisy, for example, point 140 in Fig. 5a or points 142 and 144 in Fig. 5b, it forces μ = 0 (i.e. no update for equalizer coefficients). With this feature, the proposed equalizer of Fig. 4 can use a μ value close to 0.5 thereby achieving a much higher tracking speed than the LMS-like algorithms of Fig. 3.

The modified update rule can be rewritten as C_rn,n⁼(Cm-ι_>nX_mjn+μ(Y_m-ι,n-C_m- ι,nXm,n))/Xm,n. This leads to an efficient hardware implementation as shown in Figure

4. The adaptive equalizer in accordance with the embodiments of the present invention uses one multiplier less than the LMS equalizer of Fig. 3 and has a simpler architecture because C_m-ι_>n and X_m,_n are used only once in the structure. This further reduces the number of registers. Also, there is only one port that writes into the coefficient memory so an additional multiplexer is saved.

Numerous modifications, variations and adaptations may be made to the particular embodiments of the invention described above without departing from the scope of the claims, which is defined in the claims.

Claims

What is claimed is:

1. A method of OFDM channel equalization comprising the steps of:

a) deriving a coefficient value from a received reference ^"value;

b) receiving subsequent symbols in dependence upon the coefficient value; and

c) updating the coefficient value in dependence upon a weighted average of an optimal coefficient value for the current symbol and the previous coefficient value.

2. A method of claim 1 further comprising the step of:

determining a quality of the current value and maintaining the previous coefficient value when the quality is below a predetermined threshold.

3. A method as claimed in Claim 2 wherein the step of updating the coefficient determines new coefficient value from C_mιn=C_m-ι_(n+ μ(Ym,n Xm_>n-C_m.|_jn) where Ym,n=T(Cm.ι_>nXm,n) for symbol m and carrier n,

X_m!n is a received complex value, Y_mιn is an equalized output value and

C_m-ι,_n is a previous coefficient value.

4. An OFDM adaptive channel equalizer comprising:

a) means for deriving a coefficient value from a received reference value;

b) means for receiving subsequent symbols in dependence upon the coefficient value; and c) means for updating the coefficient value in dependence upon a weighted average of an optimal coefficient value for the current symbol and the previous coefficient value.

5. An OFDM adaptive channel equalizer as claimed in Claim 4 further comprising means for determining a quality of the current value and maintaining the previous coefficient value when the quality is below a predetermined threshold.

6. An OFDM channel equalizer as claimed in Claim 5 wherein the means for updating the coefficient includes means for determining a new coefficient value from C_m,_n=C_m.|,„+ μ(Y_m,n/Xm,n-Cm-ι,n) where Ym₎n=T(C_m.ι,_nX_m,_n) for symbol m and carrier n, X_m)n is a received complex value, Y_m>n is an equalized output value and C_m-i,n is a previous coefficient value.

7. An OFDM adaptive channel equalizer comprising an input for received signals:

a first branch for deriving a channel coefficient value from a received reference value;

a second branch for deriving an equalized output value in dependence upon the coefficient value and for adapting a previous channel coefficient value in dependence upon a weighted average of the coefficient value and an optional coefficient value for a current equalized output value and received signal; and

an output for the equalized output value.

8. An OFDM adaptive channel equalizer as claimed in Claim 7 wherein the upper branch includes an multiplier.

9. An OFDM adaptive channel equalizer as claimed in Claim 8 wherein the upper branch includes an inverter.

10. An OFDM adaptive channel equalizer as claimed in Claim 9 wherein the upper branch includes an output port to a coefficient memory.

11. An OFDM adaptive channel equalizer as claimed in Claim 10 where the second branch includes a coefficient adaptation branch that shares the output port to the coefficient memory.

12. An OFDM adaptive channel equalizer as claimed in Claim 11 wherein the upper branch includes a multiplexer.

13. An OFDM adaptive channel equalizer as claimed in Claim 12 wherein the multiplexer has an output coupled to the multiplier of the upper branch.

14. An OFDM adaptive channel equalizer as claimed in Claim 13 wherein the multiplexer has a first input for a known reference signal.

15. An OFDM adaptive channel equalizer as claimed in Claim 14 wherein the multiplexer has a second input for output from the coefficient adaptation branch.

16. An OFDM adaptive channel equalizer as claimed in Claim 7 wherein the second branch includes a coefficient adaptation branch.

17. An OFDM adaptive channel equalizer as claimed in Claim 16 wherein the second branch includes a signal slicer.

18. An OFDM adaptive channel equalizer as claimed in Claim 17 wherein the coefficient adaptation branch includes a first adder having inputs coupled to an input and an output of the signal slicer.

19. An OFDM adaptive channel equalizer as claimed in Claim 18 wherein the coefficient adaptation branch includes a weighting function coupled to an output of the first adder.

20. An OFDM adaptive channel equalizer as claimed in Claim 19 wherein the slicer has a nulling output to the weighting function.

21. An OFDM adaptive channel equalizer as claimed in Claim 20 wherein the coefficient adaptation branch includes a second adder having inputs coupled to an output of the weighting function and an input to the signal slicer.

2. An OFDM adaptive channel equalizer as claimed in Claim 21 wherein the output of the second adder is coupled to an input of the multiplexer.