WO2007045843A2

WO2007045843A2 - Correction of quadrature mismatch

Info

Publication number: WO2007045843A2
Application number: PCT/GB2006/003837
Authority: WO
Inventors: Ediz Cetin; Izzet Kale; Richard Charles Spicer Morling
Original assignee: University Of Westminster
Priority date: 2005-10-17
Filing date: 2006-10-17
Publication date: 2007-04-26
Also published as: GB2431322A; GB2431322B; GB0521056D0; WO2007045843A3; WO2007045843A8

Abstract

A method for analog front-end impairment compensation in OFDM Multi- Carrier Systems based on adaptive blind-source-separation techniques, which may be used to improve the IQ phase and gain mismatches in zero-IF type as well as low-IF type receivers. The method comprises resolving a received signal into its component carrier signals and demixing the I and Q channels of each resolved carrier in a demixing stage. In the demixing, signal components of a resolved carrier are input to first and second cross coupled adaptive filters, whose coefficients are updated by the outputs of the demixing stage.

Description

COMMUNICATIONS SYSTEM

Field of Invention

The present invention relates to communications systems, principally wireless systems but also including wired systems employing quadrature modulation and demodulation.

Background of the Invention

The use of quadrature modulation and demodulation is a common communication technique. Data communication systems modulate data onto in- phase (I) and quadrature (Q) components of a baseband signal and then mix those baseband signals with I and Q components of a Radio Frequency (RF) carrier to broadcast the modulated data. The Q signal is ninety degrees out of phase with the I signal. In the receiver the reverse process is carried out, first receiving the broadcast signal, then downconverting to recover the I and Q components of the modulated baseband signal, and then recovering the data from those I and Q components.

Receiver architectures that utilize IQ-signal processing are vulnerable to mismatches between the I and Q channels. This can happen at several stages in the receiver; the RF splitter used to divide the incoming RF signal equally between the I and Q paths may introduce phase and gain differences. The differences in the length of the two RF paths can result in phase imbalance. The quadrature 90° phase-splitter used to generate the I and Q Local-Oscillator (LO) signals that drive the I and Q channel mixers may not be exactly 90°. Furthermore, there might be differences in conversion losses between the output ports of the I and Q channel mixers. In addition to these, filters and Analog-to-Digital-Converters (ADCs) in the I and Q paths are not perfectly matched. The effects of these impairments on the receiver's performance can be detrimental. The IQ-imbalances can be characterized by two parameters: the amplitude mismatch,α_ε, and the phase orthogonality mismatch, φ_ε, between the I and Q branches. The amplitude-imbalance, β in decibels is obtained from the amplitude mismatch α_εas: £ = 201og₁₀[l+0.5or, /l-0.5α_β] (1 )

The Quadrature Demodulator receiver model of Figure 1 incorporates IQ- imbalances as impaired LO signals. An input signal s(t) is mixed with a local oscillator signal ή_o in quadrature channels. The mixed signal is subject in each channel to gain and low pass filtering (LPF).

Figure 2 demonstrates the effects of varying the IQ phase and gain mismatches on the raw Bit-Error-Rate (BER) vs. Signal to noise ratio (Eb/No) in systems using (a) 32-PSK and (b) 256-QAM modulation formats. As can be observed the IQ-imbalances degrade the system's BER performance greatly. This degradation in performance is not desirable and must be compensated. In order to ensure correct symbol detection RF impairments must be compensated for before the symbol decision takes place.

In the papers "Adaptive Compensation of Analog Front-End I/Q Mismatches in Digital Receivers", Cetin E., Kale I., Morling R.C.S., IEEE International Symposium on Circuits and Systems, (ISCAS 2001), vol. 4, pp. 370-373, May 2001., "Adaptive Self-Calibrating Image Rejection Receiver", Cetin E., Kale I., Morling R.C.S., IEEE International Conference on Communications (ICC 2004), vol. 5, pp. 2731-2735, June 2004., "On the structure, convergence and performance of an adaptive I/Q mismatch corrector" by: Cetin, E.; Kale, I.; Morling, R.C.S., IEEE Vehicular Technology Conference (VTC 2002 Fall), vol. 4, pp. 2288-2292, September 2002, there is discussed single ended zero-IF and Low-IF I/Q channel wireless systems. Such systems require a high image rejection ratio, but methods employing test tones have proved unsatisfactory. The papers propose a blind (unsupervised) technique that does not require pilot tones, but instead employ a blind adaptive algorithm. It is recognized that mismatch errors create cross-correlation between the orthogonal I and Q channels. In order to remove the cross-correlation, adaptive filters are cross-coupled between the I and Q channels. The coefficients of the filters are updated by the outputs of the filters, and the filters do not require an initial training phase.

Orthogonal Frequency Division Multiplexing (OFDM) is a frequency multiplexing scheme that employs a combination of frequency division multiplexing and quadrature modulation and demodulation to effect high speed wireless data transfer. At the OFDM transmitter, the output of the quadrature modulator is frequency division multiplexed for transmission to an OFDM receiver. Inverse Fast Fourier Transform (IFFT) and Fast Fourier Transform (FFT) are used as the modulator and demodulator at the transmitter and receiver respectively. OFDM is a widely recognized and a standardized modulation technique due to its ability to cope with multipath environments. It is used in Wireless Local Area Networks (WLAN - IEEE 802.11x), Wireless Metropolitan Area Network (WMAN - IEEE 802.16), Digital Audio Broadcast (DAB) and Digital Video Broadcast (DVB-T/S/H) applications to name a few. Reception of this signal format however, as in all quadrature modulation and demodulation systems, is very sensitive to amplitude, phase, and frequency errors emanating from the RF front-end. Hence, the IQ-mismatches in the downconversion process lead to severe performance degradation, necessitating the use of area and power hungry discrete off-chip components. What is more, the IQ-mismatches corrupt the long training symbol, which is defined in the standards and used for channel estimation, resulting in inaccurate channel estimates leading to further performance degradation. In the industry today a lot of effort is being spent in developing cost and power efficient, highly integrated OFDM receivers.

International Patent Application WO 2005/029798 discloses a system for adaptive IQ imbalance correction for multicarrier systems, wherein the receiver includes an inverse demultiplexer for removing IQ imbalance resulting from the demultiplexing operation, and a frequency offset correction to correct IQ imbalance resulting from frequency offset errors. This is carried out prior to Fourier Transformation. Subsequent to Fourier Transformation, residual IQ imbalance is removed in an equaliser that is employed to make a decision on the symbol that is being received. The equaliser employs adaptive filters, whose weights are updated by the symbol decision. A training signal that may be used to correct for multipath further complicates the construction of the equaliser. Therefore, the disclosed technique for removing IQ imbalance is a supervised technique and complex, but is not error free.

US Patent Application US 2003/0231726 discloses a system for frequency domain compensation of OFDM signals with IQ imbalance. An adaptive filter technique is employed in the channel equaliser stage subsequent to Fourier Transformation, for mismatch correction, which relies on the existence of a reference pilot signal. This technique is therefore a supervised technique, and a mechanism for unsupervised operation is not disclosed. Further the IQ imbalance may corrupt the reference pilot signal, resulting in inaccuracies.

Summary of the Invention

In general, the present invention is concerned with a communications system employing a plurality of frequency division multiplexed carriers, each carrier having respective I and Q components.

In a first aspect, the invention provides, a method of removing I/Q- mismatches in a received signal comprising a plurality of frequency division multiplexed carriers, each carrier having respective I and Q components, the method comprising: resolving the received signal into I and Q signal components, resolving the I and Q signal components into carrier signal components; and demixing said I and Q signal components in a demixing stage, said demixing comprising providing selected said signal components to said demixing stage, said demixing stage including first and second adaptive filters, whose coefficients are updated by the outputs of the demixing stage, the outputs of the demixing stage representing an IQ mismatch corrected carrier..

In a second aspect, the invention provides apparatus for removing I/Q- mismatches in a received signal comprising a plurality of frequency division multiplexed carriers, each carrier having respective I and Q components, the apparatus comprising:

I and Q resolving means for resolving the received signal into I and Q signal components, carrier resolving means for resolving said I and Q signal components into carrier signal components and a demixing stage arranged to receive as input signals selected said signal components, the demixing stage including first and second adaptive filters, and means for updating the coefficients of said adaptive filters by the outputs of the demixing stage, the outputs of the demixing stage representing IQ-mismatch corrected signals.

The system in accordance with the invention may be applied both to wireless systems and to wired systems. . Thus, this invention provides a method and apparatus that will reduce the effects of RF impairments in receivers for multi-carrier wired and wireless systems. This invention concerns the use of unsupervised algorithms to deal with analog front-end impairments in multi-carrier systems, namely (but not limited to) those employing OFDM modulation. For the purposes of the present specification, it will be understood that references to OFDM include all variants thereof, as will be apparent to those skilled in the art, including for example COFDM (Coded Orthogonal Frequency Division Multiplexing). In accordance with the invention, it is assumed that each resolved received carrier signal is orthogonal to other carrier signals, and that the I and Q values of each resolved component are orthogonal to each other, but subject to cross- correlation, as more particularly explained below. The de-mixing stage of the invention removes this cross-correlation, but without requiring an initial pilot training phase. The present invention therefore provides a blind, unsupervised technique, therefore simplifying implementation, and being inherently more accurate than systems relying on an initial training phase where the pilot signal may be inaccurate.

The invention may provide digital-signal-processing-based compensation schemes through multi-channel unsupervised signal processing techniques either in the time or frequency domains or a mixture of the two to combat these undesirable impairments. The invention may provide a separate processing stage, independent of other signal processing stages, which may be inserted at an appropriate point in the processing chain.

Use of the technique of this invention allows complexity in analog circuitry and associated costs to be reduced at the expense of additional (increased) digital signal processing, yielding an overall economic system solution. Application of these unsupervised signal processing techniques to multi-carrier wireless receivers eliminates the need for discrete off-chip components resulting in simpler, lower cost and low power receivers with enhanced performance. These will subsequently manifest themselves in simpler RF front- ends and relaxed ADC analog circuit requirements and resulting in a major step towards "true" integration of "low-power single-chip radio receivers".

Analog front-end impairments greatly limit the performance of transceivers. A practically realizable low-power digital unsupervised compensation structure is proposed, in a preferred embodiment, based on two digital filters to alleviate performance degradation. A digital compensation structure and adaptive coefficient update algorithm for determining the digital compensation filter coefficients have been developed. It has been found through extensive testing and observation, that a sufficiently high accuracy (and low BER) may be obtained merely by using the sign of the output signal to update the coefficients of the adaptive filter. Thus the construction of the filter is further simplified

In the presence of IQ-mismatches the received signal can be expressed in the frequency domain by a complex scaling of the desired frequency component and by the emergence of an interfering component at the image frequency. Related to an OFDM transmission system, where the sub-carrier demodulation is done by an FFT process with its circular indexing the desired frequency component is assumed at frequency index k. The interfering component then appears at the image frequency index N-k where N is the overall number of carriers.

With this invention we assume that sub-carriers are not correlated with each other. This is the case with OFDM modulation. Signals are orthogonal if they are mutually independent of each other. Orthogonality is a property that allows multiple information signals to be transmitted perfectly over a common channel and be successfully detected. OFDM achieves orthogonality in the frequency domain by allocating each of the separate information signals onto different sub-carriers. OFDM signals are made up from the sum of sinusoids, with each corresponding to a sub-carrier. The baseband frequency of each sub-carrier is chosen to be an integer multiple of the inverse of the symbol time, resulting in all sub-carriers having an integer number of cycles per symbol. As a consequence the sub-carriers are orthogonal to each other and hence independent, i.e. no correlation. Loss of this will result in great performance degradation.

Brief Description of the Drawings

A preferred embodiment of the invention will now be described with reference to the accompanying drawings wherein: -

Figure 1 is a schematic view of a prior art Quadrature Demodulator; Figure 2 shows the effects of IQ-imbalances on BER of (a) 32-PSK and (b) 256-QAM modulated signals;

Figure 3 shows diagrammatically a way of appreciating the effects of IQ- imbalances on OFDM Signals, which appreciation underlies the present invention;

Figure 4 schematic block diagram of a receiver in accordance with the invention for an OFDM system;

Figure 5 shows schematically preferred configurations for removing IQ mismatch in accordance with the invention; Figure 6 shows in more detail, preferred Frequency Domain configurations for removing IQ-mismatch in accordance with the invention;

Figure 7 is a schematic circuit diagram of a preferred de-mixing unit according to the invention for removing IQ mismatch between I and Q channels of a single carrier of an OFDM system; and Figure 8 is a schematic circuit diagram of an alternative de-mixing unit according to the invention for removing IQ mismatch between I and Q channels of a single carrier of an OFDM system.

Description of Preferred Embodiments

Referring to figure 4, an OFDM receiver 10 in accordance with the invention comprises an input stage 12, comprising low noise amplifier, a down converter for converting to low-IF or zero-IF (or any other variant), and an analog to digital converter. The output of stage 12 is in the form of two channels I and Q, constituting the received signal. The I and Q signals make up a complex signal, with a complex envelope u(t). In the presence of IQ-impairments, the overall resulting signal (I + j*Q) contains not only the desired signal u(f) but also the undesired complex conjugate of it u*(t). These I and Q channels are processed in an FFT stage 14 that performs a complex FFT to resolve them into the separate carriers of the OFDM signal. The separate carriers are provided by selecting different outputs or frequency bins of the FFT. Each of these bins corresponds to one OFDM carrier. A demixing stage 16, comprising a preferred embodiment of the invention, compensates for IQ-mismatch, as will be described. Further stages are provided of a channel equaliser 18, demapper 20, deinterleaver 22, and decoder 24. Referring back to Figurei , the non-ideal quadrature-demodulator LO signal can be expressed as: f_w = 2(l+05a_ε)cos(2rf_wt + 0.5φ_ε)-j2(l-0.5a_ε)sm(2πf_LOt-0.5φ_ε) (2)

This can be re-written as: f_w __g2e-J™<?e _{) + e}-J⁽Ww»_(gιe-J0-5<Ps _{+ g2}e^j0-^5φ° ) (3)

where gi=(1+0.5α_ε) and c/₂=(1-0.5α_ε). Quadrature mixing the incoming signal, s(k) with the non-ideal LO and Low-Pass Filtering results in the base-band signal Γ_/Q(/C). Hence, the complex baseband equation depicting the IQ-imbalance effects on the ideal received OFDM signal in the time domain is given as: rig (£) = Si [U₁ (k) cos(p_ε / 2) + u_Q (k) sin(p_ε / 2)]

= h₁u(k) + h₂u^*(k) where (•)* is the complex conjugate.

As can be observed output not only consists of the desired signal, scaled by complex factor hi, but also the conjugate signal scaled by fø.

In the frequency domain the influence of IQ errors can be expressed as follows: Let the transmitted OFDM symbol (in the frequency domain) be OFS, then IFFT(OFS) is the incoming time domain signal at the receiver (IFFT: Inverse FFT). Applying IQ-imbalances of (2) and taking the FFT leads to: r_1Q(f) + h₂.[IFFnOFS)f} _(g)

ΠQ( f ) is the received OFDM symbol and (OFS)_1n is the transmitted OFDM symbol in the frequency domain, mirrored over the carriers. (OFS)_m(i) = (OFS)_mom__{i+2 N)} N being the number of the sub-carriers in the OFDM symbol, l≤/≤N and mod is the modulo operation. Referring to Figure 3, which shows the situation diagrammatically, Carrier 1 is the leftmost carrier, for /=1. In the absence of the IQ- imbalances only the carriers from O - (D are present. In the presence of the IQ-imbalances however these sub-carriers are interfered with by the sub-carriers from O - O i.e. conjugate of t/(t). Thus, u(k) = sub- carrier 1 + sub-carrier 8, and u*(k)= sub-carrier 8 + sub-carrier 1

It should be noted that the signal corruption depicted in Figure 3, can be cast in the time-domain as dictated in equation (4). In the time-domain equation (4) can be recast as: r_]Q(k) = Hu_1Q(k) (6) where H is the unknown non-singular mixing matrix which is determined by the phase and gain errors and UIQ(/C) is the transmitted signal. Given the received vector ΓIQ(/C), the source separation problem comprises the recovery of the original signals in an unsupervised way by finding a de-mixing matrix W hence recovering the sources: c_/e(*) = Wr_/β(*)

= WKu_IQ(k)

Both W and H are 2-by-2 matrices.

However, in the frequency domain, as can be observed from (3) we operate at the sub-carrier level. As observed from (5) in the presence of IQ-errors there is inter-carrier-interference. The invention is based on de-correlating the carriers to eliminate the IQ-errors. With this approach, W and H are 1-by-N matrices.

Figures 5(a) and 5(b) depict the different configurations of the invention for the time and frequency domain approaches.

In Figure 5(a), for time domain demixing, a demixing stage 50 is provided that accepts the I and Q components, in digital form, of the received signal from ADCs 52 of unit 12. Demixing takes place and the demixed signals are applied to FFT 14.

In Figure 5(b), for frequency domain demixing, the demixing stage 16 is placed after the FFT 14, as in Figure 4, to receive the resolved sub carriers.

By way of example, Figure 6(a) shows in more detail the demixing stage 16 of Figure 5(b) for the frequency domain. The FFT demodulation provides 8 outputs 1 - 8, each representing a different frequency bin (any number of outputs may be provided in practice). These outputs are provided to four demixing units PE1-4, each demixing unit being as shown in Figure 7. Outputs 1 and 8 , 2 and 7, 3 and 6, 4 and 5 are mirror image frequency signals. Such mirror image frequencies are positive and negative values of the sub-carrier frequency. The amplitude (u) of each frequency bin is a complex value, and the amplitude of the respective mirror image frequency bin is the complex conjugate (u*) of that amplitude. This is a known property of complex Discrete Fourier Transforms.

As stated above, in the presence of IQ-mismatches the received signal can be expressed in the frequency domain by a complex scaling of the desired frequency component and by the emergence of an interfering component at the image frequency. Related to an OFDM transmission system, where the sub- carrier demodulation is done by an FFT process with its circular indexing the desired frequency component is assumed at frequency index k. The interfering component then appears at the image frequency index N-k where N is the overall number of carriers. Referring to Figure 3, in the absence of the IQ- imbalances only the carriers from (D - ® are present. In the presence of the IQ- imbalances however these sub-carriers are interfered with by the sub-carriers from © - O i.e. conjugate of t/(t). Thus, u(k) = sub-carrier 1 + sub-carrier 8, and u*(k)= sub-carrier 8 + sub-carrier 1. Thus for N frequency bins, the information for decorrelating N/2 carriers is provided. As per Figure 6(a) we only require N/2 demixing units PE, each PE representing a demixing unit as shown in Figure 7.

Figure 7 shows a preferred implementation of a demixing unit, comprising cross-coupled filters for resolving IQ-mismatch. Figure 7 shows in detail the manner in which the adaptive filter system uses both u and u*, where (•)* is the conjugate operation, to generate the corrected signals. The adaptive system includes cross-coupled adaptive filters.

The received signal r(k) and its complex conjugate r*(k) are fed into cross- coupled adaptive filters. The adaptive coefficient update block determines a new de-correlation-matrix that, when used to generate another corrected signal, further reduces the magnitude of the error signal. The output of the coefficient update block is then provided back to the adaptive-filter system which then replaces its de-correlation-matrix as provided by the coefficient-update block. This new de-correlation-matrix is then used to perform inverse filtering, these estimates are then subtracted to yield the estimated or reconstructed signals ci(/c) and C₂(k). The process continues until the magnitude of the error signal reaches a minimum or a pre-defined threshold. The error signal thus functions as a feed-back signal for adjusting the de-correlation-matrix.

As shown in Figure 7,. the I and Q components of the carrier, r(k), r*(k) are applied to the inputs of a demixing unit 70. Demixing unit 70 comprises first and second adaptive filters 72 in feed forward loops 74. Loops 74 are cross-coupled between the two channels, and are connected to summation points 76 in the channels, so that each input signal, as modified by the adaptive filter, is added to the other input signal. The outputs of the channels, ci(k), c₂(k) represent the outputs of the demixing unit, and are used to update the coefficients of the filters, as at 78. Referring to Figure 3, r(k) = sub-carrier 1 + sub-carrier 8, and r*(k)= sub-carrier 8 + sub-carrier 1. We can have the following: wi and W₂ are both complex, where wi and W₂ refer to complex adaptive filters. For the feed-forward case the source estimates, Ci(k) and C2(k), become:

C₂(K) = Qi₂ - w₂)r(*) + (l - wΛ>-^'(*)

When the filters converge, i.e. Wi = hi and W₂ = h₂ then the source estimates become:

For the feed-back case:

(10)

O₂ (k) = —^ [(K - W₂ )r(k) + (1 - W₂Zz₁ )r'(*)]

1 - W₁W₂ When the filters converge, i.e. wi = hi and W₂ = h₂then the source estimates become:

Furthermore, with the further frequency domain approach of Figure 6(b) a single demixing unit would suffice, demixing one carrier only - the other carriers are demixed by an estimation process, using the information from the demixed carrier. However, to improve estimation performance two demixing units are shown. By observing the outputs of PE1 and PE2 e.g. impulse response of the filters or the filter coefficients which depend on the IQ-impairments, by only utilising 4 carriers one can estimate the mixing matrix and create a de-mixing matrix using these elements. This results in reduced complexity. By using the first and last carrier IQ-impairments can be estimated over the received band. An alternative implementation for the demixing unit is placing the filters in the feedback loop. The structure is shown in Fig 8, where similar parts to those of figure 7 are denoted by the same reference numeral. Filters 72 are placed in feedback loops 80.

For the time domain demixing stage of Figure 5(a), the structure of the demixing unit is as shown in Figure 7, and the equations (8-11 ) above apply.

We have demonstrated that the coefficient update works by using only the polarity of the output signal, with the overall system performance not compromised in any way; this results in massively reduced complexity.

Furthermore, the operation of the adaptive filter will also be simplified as we will be operating using the sign of the derived coefficients.

Features of the invention, at least in the embodiments described above are as follows:

1- Method that eliminates RF impairments in communications receivers without the need for pilot/test tones. 2. Enhanced performance devices with reduced bill of material costs, which enables electronics manufacturers to cost-effectively design and market cheaper products. 3. Integration and elimination of large, power hungry analog components with their digital counterparts leading to more robust and power efficient products designed using non-specialist low-production-cost CMOS technology. 4. Method is applicable both to zero-IF and low-IF receivers.

5. Method applicable for both time and frequency domain correction.

6. Method for correction matrix estimation using the polarity of the data only rendering extremely hardware efficient solution. We have demonstrated through analytical simulation studies that this coefficient update block works with just polarity information with the overall system not compromised resulting in massively reduced complexity

7. Blind hence no need for training or pilot/test tones. However, training can be easily reconfigured to incorporate the pilot signal in OFDM signal processing to improve the performance. 8. The approach is modulation format and constellation size independent.

9. Easily integrated into standard signal-processing chains of receivers with little hardware/software overhead. The invention can be easily applied to existing systems, without requiring changes to installed broadcasting infrastructure. 10. Operates under multi-path and fading environments.

11. Both channels are recovered to high quality. With the low-IF version, one not only receives the desired channel but also the interferer which happens to be the adjacent channel. This can be put to use in base-station receivers to reduce complexity and cut cost. 12 Can be applied to any IQ signal processing block to eliminate impairments with minor modifications mainly to the input signal configuration. 13 Both transmitter and receiver imbalances are dealt with as a composite entity.

14. The present invention may be used in any of a number of wireless communication platforms e.g. WLAN (IEEE 802.11x), WMAN (IEEE 802.16), DAB, DVB-T, DVB-S and DVB-H applications to name a few.

15. The invention compensates for frequency dependent IQ errors as well as static IQ errors, since the correction takes place in the frequency domain.

Claims

1. A method of removing l/Q-mismatches in a received signal comprising a plurality of frequency division multiplexed carriers, each carrier having respective I and Q components, the method comprising: resolving the received signal into I and Q signal components, resolving the I and Q signal components into carrier signal components; and demixing said I and Q signal components in a demixing stage, said demixing comprising providing selected said signal components to said demixing stage, said demixing stage including first and second adaptive filters, whose coefficients are updated by the outputs of the demixing stage, the outputs of the demixing stage representing an IQ mismatch corrected carrier.

2 A method according to claim 1 , wherein the received signal is provided by an OFDM wireless system.

3. A method according to claim 1 or 2, wherein said resolving is carried out by means of a complex Fourier Transformation, and the outputs of the Fourier Transformation represent I and Q channels of each resolved carrier.

4. A method according to any preceding claim, wherein said demixing takes place in the time domain, prior to the resolution of said carriers, and comprising applying said I and Q components to said demixing stage.

5. A method according to any of claims 1 to 4, wherein said demixing takes place in the frequency domain subsequent to the resolution of said carriers, and comprising applying selected said carrier components to said demixing stage 6. A method according to claim 5, wherein demixing of a carrier is carried out by applying a first resolved carrier, and applying the carrier at a mirror frequency thereof, representing the complex conjugate of the first resolved carrier, to said demixing stage.

7. A method according to claim 5 or 6, wherein said demixing is applied to at least a first carrier, and demixing of remaining carriers is carried out by means of an estimation process.

8. A method according to claim 7, wherein said demixing is applied to first and second carriers.

9. A method according to any preceding claim, wherein the coefficients of the adaptive filters comprise the sign of the outputs of the demixing stage.

10. Apparatus for removing l/Q-mismatches in a received signal comprising a plurality of frequency division multiplexed carriers, each carrier having respective I and Q components, the apparatus comprising:

I I Apparatus according to claim 10, incorporated in an OFDM wireless system.

12. Apparatus according to claim 10 or 11 wherein said carrier resolving means comprises complex Fourier Transformation means, wherein each resolved carrier is represented in its I and Q values by a first frequency output, and the output at the mirror frequency thereof. 13. Apparatus according to claim 10 or 11 , wherein said demixing stage is coupled between said I and Q resolving means and said carrier resolving means, and demixing takes place in the time domain, prior to the resolution of said carriers.

14. Apparatus according to any of claims 10 to 12, wherein said demixing stage is coupled to the output of said carrier resolving means, so that demixing takes place in the frequency domain subsequent to the resolution of said carriers.

15. Apparatus according claim 14, arranged for applying a first resolved carrier, and for applying the carrier at a mirror frequency, representing the complex conjugate of the first resolved carrier, to said demixing stage.

16. Apparatus according to claim 14 or 15, wherein said stage is arranged to demix at least a first carrier, and including estimating means for demixing remaining carriers by means of an estimation process.

17. Apparatus according to claim 16, wherein said demixing stage is arranged to demix first and second carriers.

18. Apparatus according to any of claims 10 to 17, wherein said demixing stage comprises one or more demixing units, each unit being arranged to demix a respective signal component.

19. Apparatus according to any of claims 10 to 18, wherein a said demixing unit comprises a first input coupled through a first signal path that includes first summation means to a first output, a second input coupled through a second signal path that includes second summation means to a second output, said first filter being coupled from said first to said second signal path, and said second filter being coupled from said second to said first signal path.

20. Apparatus according to claim 19, wherein said first filter is coupled between said first input and said second summation means in a feed forward loop, and said second filter is coupled between said second input and said first summation means in a feed forward loop.

21. Apparatus according to claim 19., wherein said first filter is coupled between said first output and said second summation means in a feedback loop, and said second filter is coupled between said second output and said first summation means in a feedback loop. 22. Apparatus according to any of claims 10 to 21 , wherein the coefficients of the adaptive filters comprise the sign of the outputs of the demixing stage.