I/q mismatch compensation in an ofdm receiver in presence of frequency offset
A receiver for an orthogonal frequency division multiplex radio signal in which a carrier frequency is modulated by subcarrier signals (S1) coded with data. Analogue signal processing means (3 to 12) produces baseband analogue signals (IRx, QRx) in phase quadrature and analoguetodigital converters (13, 14) convert the analogue signals to phase quadrature digital signals (x1(n), xq(n)). The digital signal processor includes the OFDM demodulator (15) and mismatch compensation (17, 18). The mismatch compensation (17, 18) combines each of the reproduced subcarrier signals (R1) with a limited number of the reproduced subcarrier signals (R1Rk) according to respective frequency offset coefficient (1, k) that is a function of an estimated value of the offset (fc) of the reference frequency relative to the carrier frequency.
This invention relates to an orthogonal frequency division multiplex (‘OFDM’) receiver with compensation for frequency offset. As used herein, ‘receiver’ refers to apparatus capable of receiving a radio signal, whether or not it is also capable of transmitting a radio signal.
BACKGROUND OF THE INVENTIONOFDM transmissions are widely used. Examples of their use include Digital Video Broadcast (DVB), Digital Audio Broadcast (DAB), and wireless broadband transmission standards such as IEEE 802.11a, ETSI/BRAN/Hiperlan2 and ARIB/MMAC/HiSWAN.
An OFDM transmitter/receiver (‘transceiver’) includes an analogue signal processing part, the RF frontend, and a digital signal processing (DSP) part, also referred to as the baseband digital IC. In the reception direction, the function of the RF frontend is to convert the OFDM signal from the RF frequency (e.g. 5 GHz in IEEE802.11a) to baseband, and to generate the inphase (I) and quadrature (Q) components of the baseband signal. The digitised I and Q signals are then processed by the DSP unit. There are two basic architectures to generate I and Q digital signals:

 The first architecture, which is not utilised by the present invention, is known as digital I/Q generation. In digital I/Q generation, the RF signal is converted to a low intermediate frequency (IF) (for example 20 MHz in IEEE802.11a) and sampled at a relatively high frequency (e.g. greater than 40 MHz in IEEE802.11a) by a single Analogue to Digital Converter (ADC). The single digitised signal is then processed by the DSP unit, which digitally generates the I and Q signals and processes them. A drawback of this architecture is that it is power consuming and increases the complexity of the DSP. Also, many I/Q compensation techniques only deal with a mismatch that remains constant over the whole frequency band of the signal and only give good results for narrow band signals.
 The second architecture is called analogue I/Q generation. An example of a transceiver in conformity with the IEEE802.11a standards with this architecture as disclosed in our copending European Patent Application N° EP 01401631.5 filed 20 Jun. 2001 is shown in
FIG. 1 and the signals appearing in operation at various stages of the receiver are shown inFIG. 2 .
An analogue I/Q receiver may include a first downconversion stage, which converts the RF signal to an intermediate frequency, as shown in
The transceiver illustrated in
Analogue I/Q generation has been found to be more difficult to implement than digital I/Q generation, however, because avoiding signal impairment (such as crosstalk between the subcarriers especially, for example) has required high quality matching between the I and Q signal paths. The analogue treatment of the I/Q signals is sensitive to mismatch. Such mismatch arises from slight differences in the values and behaviour of active and passive elements found in the I and Q signal paths, even though great care is taken in the design and layout of these elements in a symmetrical way during the design of the system and/or circuit. Mismatches are even more pronounced when the effects of thermal drift are taken into account.
The present invention enables signal impairments to be reduced without requiring such high quality matching between the I and Q signal paths. The overall solution combines the advantages of a high quality signal and a low power consumption and circuit area.
SUMMARY OF THE INVENTIONThe present invention provides a receiver for an orthogonal frequency division multiplex radio signal as described in the accompanying claims.
BRIEF DESCRIPTION OF THE DRAWINGS
Our copending European Patent Application N° EP 01401631.5 filed 20 Jun. 2001 describes a digital method for mismatch compensation of the I and Q paths of an OFDM transmitter or receiver implementing analogue I/Q generation.
The method that is described in our copending European Patent Application allows the compensation of the I/Q mismatch in an OFDM receiver when there is no offset or a negligible offset between the transmitter and the receiver carrier frequency. However, when the carrier frequency offset becomes bigger, the data transmission quality degrades very quickly (i.e. higher bit error rate) and it becomes necessary to implement a different compensation method.
The embodiment of the present invention shown in
The two digital signals x_{I}(n) and x_{Q}(n) from the ADCs 13 and 14 are sent to the DSP which implements the functions described below:

 As in most OFDM systems, the receiver (and transmitter) generates its clock signals internally from its own crystal (not shown). During data transmission, the receiver is tuned nominally to the same channel as the transmitter from which it is receiving the signals but a frequency offset equal to the difference between the transmitter and the receiver carrier central frequency usually appears. When the digital I and Q signals enter the DSP, this carrier frequency offset is compensated in the time domain by a frequency offset compensation circuit 16.
 The OFDM demodulator 15 performs a fast Fourier transform (‘FFT’) by converting the time domain signals to the frequency domain and recovers the subcarriers that were transmitted.
 An I/Q mismatch compensation block 17 removes crosstalk between subcarriers, which is generated by the mismatch between the analogue components of the I and Q channels, especially the IF mixers, the low pass filter and the analogue to digital converter. The I/Q mismatch compensation method is described in more detail below.
 The signals then pass to a phase offset compensation and equalisation circuit 18.
 Finally, the data is decoded from the compensated subcarriers in a circuit 19.
Crosstalk between subcarriers is generated In operation of an OFDM receiver without compensation after the conversion from IF to baseband by the mismatch between the analogue components of the I and Q channels and increases the data transmission error rate. This crosstalk will now be described by equations and the method to remove this crosstalk in order to improve the data transmission quality will then be analysed.
In the time domain, an OFDM signal is the sum of K sinusoidal waveforms, that is to say the subcarriers that carry the data to be transmitted. Each subcarrier's amplitude and phase is represented by the complex element S_{k }with k=[−K/2, −1, . . . , +1, +K/2].
The OFDM signal is transmitted after being up converted with a central carrier frequency f_{c}. Then, each subcarrier is sent at the frequency f_{c}+kf_{s}/N, with N being the size of the inverse fast Fourier transform (‘IFFT’) being used for the OFDM modulation and f_{s }the sampling frequency. The signal at the transmitter antenna is represented by the following equation:
In the receiver, after down conversion to baseband, the subcarriers are placed symmetrically around the DC frequency if there is no carrier frequency offset as shown in the upper part of
In a receiver system working with analogue I/Q generation, the essential path mismatch originates from the analogue part imperfections. Putting G_{k }as the common gain for the I and Q path, α_{k }and θ_{k }as the low pass filter and A/D converter amplitude mismatch and phase mismatch, they are function of the subcarrier k frequency. The IF mixers have an amplitude mismatch a and a phase mismatch θ that are independent of the subcarrier frequency. Then, the I and Q samples at the output of the receiver A/D converters are modelled as follows:
We consider α and α_{k }to be small compared to 1. Then, for subcarrier k, the overall amplitude mismatch can be represented by a_{k }and the overall phase mismatch by φ_{k}, with a_{k}=α+α_{k }and φ_{k}=θ+θ_{k}. The I and Q samples at the output of the receiver A/D converters can be described as follows:

 where n=−E . . . −1,0,1 . . . N−1, E being the length of the cyclic extension and φ_{0 }the phase offset for the first sample.
The complex signal before the frequency offset compensation is:
The offset frequency compensation is implemented in the time domain by multiplying the received samples by
where n=0,1 . . . N−1.
The demodulation is performed by a fast Fourier transform on the useful samples. The received subcarrier R_{I }is described by the following equations:
Equation 5 shows that each received subcarrier R_{I }is the sum of the transmitted subcarrier S_{I }multiplied by the coefficient I_{I }and the channel gain G_{I }plus a crosstalk (righthand term), which is dependent on all the other subcarriers.
Since the phase φ_{1 }is known, it can be set to zero for simplification. Furthermore, it is known that the effect of a clock frequency offset on the frequency domain symbol is a rotation of subcarrier k by a phase that depends on the subcarrier frequency and is denoted by φ_{VPE}(k). It can be verified that in the presence of I/Q imbalance, carrier and clock frequency offset, Equation 5 is modified as follows:
The I/Q mismatch that is represented in these equations is compensated as follows in this embodiment of the present invention.
Based on equation 7, and the assumption that the crosstalk term remains small compared to the lefthand term, the compensation is implemented by subtracting the crosstalk from the received signal. Then, the corrected signal Z_{I }is defined by the equation below:

 where R_{I }is the received signal and S_{I }is the original transmitted signal. As shown in equation 9, the transmitted subcarrier S_{I }is recovered by calculation from Z_{I }by doing a phase shift of −φ_{VPE }followed by a division by I_{I}G_{I}. The phase compensation and the equalisation digital treatments in the circuit 18 that follows the I/Q mismatch compensation block 17 in the baseband digital IC implement these calculations.
If we assume that the amplitude and phase mismatch a_{k }and φ_{k }are small compared to 1, the second order term in I_{k }can be neglected and the ratio J_{k}/I_{k }can be simplified as follows:
The corrected signal Z_{I }can then be rewritten as:
When there is negligible carrier frequency offset (Δf_{C}˜0) the coefficient λ_{l,k }is equal to 1 for k equal to I and 0 otherwise. Then Equation 8 can be simplified and we obtain the same compensation as in the receiver described in our copending European Patent Application N° EP 01401631.5:
Z_{k}=R_{k}−(a_{k}+j.tan(φ_{k})){overscore (R)}_{−k} Equation 11
The complex term A_{k }or the simplified terms a_{k }and tan(φ_{k}) are called the I/Q mismatch coefficients and are calculated during a calibration procedure; various calibration procedures are available and a suitable procedure is described below.
As shown in the upper part of
When the carrier frequency offset becomes significant crosstalk is generated by all the other subcarriers. However, since the function sin (πN.x)/sin (πx) decreases quickly when x increases, λ_{l,k} also decreases quickly with increasing values of
It has been shown by simulation that it is possible to obtain sufficient compensation while reducing the complexity of the implementation by compensating the crosstalk of a limited number of subcarriers only. For instance, three or even two subcarriers (out of 64 in Hiperlan 2) are enough in some cases.
As shown in the lower part of
being smaller than a chosen maximum value. For instance the criterion
can be selected.
The corrected signal Z_{I }becomes:
A calibration procedure is used to calculate the mismatch coefficients a_{k }and tan(φ_{k}), for each subcarrier k that are used to implement the I/Q mismatch compensation. The calibration procedure is preferably executed once only, after power up of the system and is carried out before normal data transmission in order to avoid adding any overhead to the normal operation. The calibration method may be changed depending on the actual values of the analogue components mismatch in the various components of the RF frontend. The calibration method described below is similar to that described in our copending European Patent Application N° EP 01401631.5.
In
The first training signal S_{t1 }is used to measure the crosstalk coefficient for the negative subcarriers (k=−1 to −K:2). S_{t1 }is a time domain signal made of at least one symbol. It can either be stored as a set of time domain samples or be obtained by OFDM modulation of a plurality of stored frequency domain components D_{k}=B_{k}e^{jP}_{k }mapped on subcarriers of frequency (f_{c}+k f_{s}/N). In a preferred embodiment, S_{t1 }comprises a single OFDM symbol corresponding to the modulation of K/2 nonzero components D_{1 }to D_{K/2 }such that B_{k}=1, P_{k}=0, and thus D_{k}=1 for all k from 1 to K/2.
After insertion of the cyclic extension, the I and Q components are obtained and, for the purposes of this embodiment of the present invention, are used to train the receive path (although it would also be possible to use them to train the transmit path). Denoting by R_{k }and R_{−k }the FFT outputs corresponding to subcarriers k and −k as described above, the crosstalk coefficients for subcarrier −k, a_{−k }and tan(φ_{k}) are obtained by the following complex operations:
and where the asterisk * denotes complex conjugation.
The second training signal S_{t2 }is used to measure the crosstalk coefficient for the positive subcarriers (k=1 to K:2). S_{t2 }comprises a single OFDM symbol corresponding to the modulation of K/2 nonzero components D_{−1 }to D_{−K/2 }such that B_{−k}=1, P_{−k}=0, and thus D_{−k}=1 for all k from −1 to −K/2. In the same way as the first training signal, the crosstalk coefficients for subcarrier k, a_{k }and tan(φ_{k}) are obtained by the following complex operations:
In more detail, the subcarriers numbers of indexes I, k_{1 }and k_{2 }from the OFDM demodulation FFT circuit 17 are selected by a selection circuit 26 for each value of the index I in turn as defined by a counter 27. In this preferred implementation, the values λ_{I,k1 }and λ_{I,k2 }(see equation 6) are precomputed and stored in a lookup table 28 for the various values of/and of the frequency offset, although it would also be possible to compute them directly as and when needed. The complex numbers A(k)=J_{k}/I_{k}=a_{k}+j tan(φ_{k}) have also been computed during the calibration phase and are stored in another lookup table 29. Then for each FFT subcarrier output of index I, complex multiplications are performed of A_{k1 }by λ_{I,k1 }in a multiplier 30 and of A_{k2 }by λ_{I,k2 }in a multiplier 31. Complex conjugations of R_{k1 }and R_{k2 }are performed in circuits 32 and 33 and the results are multiplied by the outputs of multipliers 30 and 31 in multipliers 34 and 35 respectively. Two complex subtractions of R_{k1 }and R_{k2 }from R_{I }are performed in a sum circuit 36 to obtain the compensated symbol Z_{I}.
The number of operations per symbol depends on the chosen maximum value, but the resulting complexity is typically small compared to other blocks like FFT 15 and in any case much smaller than digital I/Q generation complexity.
In order to evaluate the effect of the I/Q mismatch compensation of the above embodiments of the invention, simulations of packet error rate (PER) were performed on a Hiperlan2 simulator with a carrier frequency offset ranging from −310 kHz to +310 kHz, which is a range wider than the ETSI specifications (40 PPM between transmitter and receiver clock, corresponding to +/− 200 kHz with a 5 GHz channel central frequency).
The results from simulations that implement the I/Q mismatch compensation algorithm described by Equation 12 (with max_value equal to 1) are shown in
It will be noted that when the carrier frequency offset is close to zero, the algorithm described by Equation 12 and that described by Equation 11 produce the same improvement in data transmission and the PER is substantially lower than that obtained without I/Q mismatch compensation.
However, when the carrier frequency offset becomes bigger (more than a few tens of kHz) the data transmission quality becomes much better for the algorithm described by Equation 12 than that described by Equation 11. Above a frequency offset of 70 kHz, it can be seen that the algorithm described by Equation 11 can even degrade the data transmission quality compared to no I/Q mismatch compensation at all. This is explained by the fact that the compensation method calculates the crosstalk to be removed from the received subcarrier using a single symmetric subcarrier despite the fact that the crosstalk was generated by several other subcarriers.
Claims
1. A receiver for an orthogonal frequency division multiplex radio signal in which a carrier frequency is modulated by subcarrier signals (S1) coded with data, the receiver comprising analogue signal processing means for combining a received radio frequency signal with at least one reference frequency signal and producing analogue signals (IRx, QRx) in phase quadrature whose frequencies are functions of frequency differences between said radio frequency signal and said reference frequency signal, analoguetodigital converter means for converting said analogue signals to phase quadrature digital signals (xI(n), xQ(n)), and digital signal processing means including demodulator means for reproducing said subcarrier signals, mismatch compensation means for combining each of said reproduced subcarrier signals (R1) according to a mismatch coefficient (A(k)) with another of said reproduced subcarrier signals (R−1) whose subcarrier frequency is symmetrical therewith and compensating for offset (Δfc) of said reference frequency relative to said carrier frequency, and decoding means for recovering data from the reproduced and compensated subcarrier signals, characterised in that said compensation means is arranged to combine each of said reproduced subcarrier signals (R1) with a plurality of said reproduced subcarrier signals (R−1, Rk) according to respective values of said mismatch coefficient (A(k)) and according to a respective frequency offset coefficient (λl, k) that is a function of an estimated value of said offset (Δfc) of said reference frequency relative to said carrier frequency.
2. A receiver as claimed in claim 1, wherein said compensation means is arranged to combine each of said reproduced subcarrier signals (R1) with a limited number only of said reproduced subcarrier signals (R−1, Rk).
3. A receiver as claimed in claim 1, wherein said compensation means is arranged to combine each of said reproduced subcarrier signals (R1) with only those of said reproduced subcarrier signals (R−1, Rk) for which a decision parameter N k  l N  2 Δ f c f s is less than a threshold value, where N is the number of subcarriers used for the OFDM modulation, l is the index of the subcarrier (R1) subjected to compensation, k is the index of the crosstalk subcarrier (Rk) considered for compensation, Δfc is the offset of said reference frequency relative to said carrier frequency and fs is the sampling frequency used for the OFDM modulation.
4. A receiver as claimed in claim 1, wherein said threshold value is at least approximately equal to 2.
5. A receiver as claimed in claim 1, wherein said threshold value is such that each of said reproduced subcarrier signals (R1) is combined with a number of said reproduced subcarrier signals (Rk) between 2 and 4.
6. A receiver as claimed in claim 1, wherein said frequency offset coefficients (λl, k) satisfy at least approximately the relation λ l, k = 1 N ⅇ j π ( N  1 ) ( k  l N  2 Δ f C f S ) sin ( π N ( k  l N  2 Δ f C f S ) ) sin ( π ( k  l N  2 Δ f C f S ) ) where N is the number of subcarriers used for the OFDM modulation, l is the index of the subcarrier (R1) subject to compensation, k is the index of the crosstalk subcarrier (Rk) considered for use in compensation, Δfc is the offset of said reference frequency relative to said carrier frequency and fs is the sampling frequency used for the OFDM modulation.
7. A receiver as claimed in claim 1, wherein said compensation means includes conjugate means for producing a complex conjugate value ({overscore (R)}−k) of said plurality of reproduced subcarrier signals (Rk) and multiplying means for multiplying said complex conjugate value ({overscore (R)}−k) by a compensation value that is a function of said frequency offset coefficient (λl, k).
8. A receiver as claimed in claim 1, wherein said compensation value is arranged to be a function of said mismatch coefficient (A(k)) as well as of said frequency offset coefficient (λl, k).
9. A receiver as claimed in claim 1, wherein said mismatch coefficient (A(k)) is arranged to be at least approximately equal to ak+j tan(φk), where ak is the amplitude mismatch and φk the phase mismatch that said analogue processing means presents to the phase quadrature channels (IRx, QRx) for a subcarrier signal Sk.
10. A receiver as claimed in claim 1, wherein said compensation means includes storage means for storing precomputed values of said frequency offset coefficient (λl, k) as a function of different values of said offset (Δfc) of said reference frequency relative to said carrier frequency for different values of l and k and for selecting one of said precomputed values as a function of an estimated value of said offset (Δfc) and of l and k.
11. Communication apparatus comprising a receiver as claimed in claim 1 and a transmitter for generating an orthogonal frequency division multiplex radio signal, wherein said compensation means comprises calibration means for passing an orthogonal frequency division multiplex signal generated by said transmitter to said receiver and for responding to signals reproduced by said receiver therefrom mismatch coefficient (A(k) in combining said reproduced subcarrier signals (R1) with said reproduced subcarrier signals (Rk).
Type: Application
Filed: Feb 18, 2003
Publication Date: Jul 14, 2005
Patent Grant number: 7443783
Inventors: Paul DeChamps (PrevessinMoens), Francois Bourzeix (Ornex), Sebastien Simoens (GifSurYvette Cedex)
Application Number: 10/504,456