Quadrature circuits are commonly used in transceivers to enable communication of signals as symbols. To this end, quadrature circuits permit quadrature modulation and demodulation of a digital data stream into an in-phase (I) oscillation signal and a quadrature (Q) oscillation signal that is ninety degrees out of phase from the in-phase signal. A composite RF signal that will be transmitted onto the wireless medium can then be obtained from combination of the in-phase and quadrature signals, or alternately, the in-phase and the quadrature signals are derived from demodulation of a RF signal received over the wireless medium. A transceiver typically comprises a transmitter block and a receiver block and each block has a respective in-phase branch for processing the I signal and a quadrature branch for processing the Q signal. These in-phase and quadrature branches of a transmitter and/or receiver operate in parallel and any mismatch or imbalance among the symmetrical components of one of the two pairs of I/Q branches may cause gain mismatch and phase error between the two branches, which in turn may cause the I and Q signals to be non- orthogonal.

IEEE 802.11 specifies an Orthogonal Frequency Division Multiplex (OFDM) scheme that employs a combination of frequency division multiplexing and quadrature modulation and demodulation to effect high-speed wireless data transfer. In high data rate systems using OFDM, the information data is spread over multiple orthogonal sub-carriers whose combined spectrums cover the entire allocated channel. The observed imbalance in the I/Q branches degrades the performance of quadrate transceiver systems by causing leakage from one sub band to another.

A first base-band gain and phase mismatch observed between the two branches relies in imperfections of the high frequency components used. The mismatch experimentally appears as a shift in the frequency response of the base-band filter present in each I/Q branch. Such mismatch is mostly caused by high frequency local oscillator and mixer non-idealities. Solutions have thus been sought by the industry to compensate for it and a conventional compensation solution uses analog means.

A second sort of gain and phase mismatch is also experimentally observed at the edge of the band around the cut-off frequency. In OFDM systems, the information content

of the sub-bands near the edge of the band is however as important as the sub-bands closer to the center and this observed gain and/or phase error also needs compensation. This high frequency error may be caused by base-band components asymmetry between the I and Q branches.

Therefore, in view of the foregoing problems, there is a need for systems and methods that compensate for transceiver component mismatch. These systems and methods would preferably compensate for gain and phase errors between branches of a quadrature modulated signal and at the edge of sub-band cut-off frequency.

The inventors have realized that high frequency ripples commonly observed at the edge of the band are exacerbated. They further realized that such exacerbation is in part caused by an asymmetry between the two base-band filters of the in-phase and quadrature branches and that this asymmetry could be cured by a test estimation of the phase and gain errors introduced by the two filters before operation of the transceiver. A method for correcting a component non-ideality in a quadrate transceiver is therefore proposed. First, a waveform signal at a selected frequency is applied to a first filter of an in-phase branch of the transceiver, which results in an in-phase filtered test signal. The same wavefonn signal shifted by ninety degrees is applied to a second filter of a quadrature branch of the transceiver, which outputs a quadrature filtered test signal. An error parameter is thereafter measured between the filtered test signals and a correction signal is determined therefrom for the selected frequency. The correction signal may be used for compensating a pair of quadrature signals whose spectrum comprises the selected frequency. The compensation is done when the two quadrature signals are applied to the filters and results in compensated filtered signals.

The residual error near the band edge may be corrected via loop back tests on the respective in-phase and quadrature filters of the receiving or transmitting branches using predetermined waveform signals and the resulting correction parameters are further stored in a digital signal processing unit for pre-distorting and post-distorting incoming or out- going in-phase and quadrature signals passed through the respective in-phase and quadrature branches when the transceiver actually operates.

Thus, when the transceiver is transmitting over the wireless medium, the pair of quadrature signals I and Q are pre-distorted in the transmitting branch before they are actually passed through the respective in-phase and quadrature filters of the transmitting branch. For a given sub-band, the I and Q signals for the sub-band spectral components may

be pre-distorted using the error parameter measured and stored in the DSP for one of these spectral components of the sub-band. In an embodiment, the error parameter used for compensating the spectral components of a given sub-band is that previously obtained during loop tests for a frequency centrally located in the spectrum of that sub-band.

When the transceiver is receiving over the wireless medium, the pair of quadrature signals I and Q are post-distorted in the receiving branch after they have been passed through the respective in-phase and quadrature filters of the receiving branch. For a given sub-band, the I and Q quadrature signals for the sub-band spectral components may be post- distorted using the error parameter measured and stored in the DSP for the receiving branch and corresponding to a frequency of one of the sub-band components.

The invention is explained in further detail, and by way of example, with reference to the accompanying drawings wherein: Fig. 1 illustrates a block diagram of an exemplary transceiver according to one embodiment of the invention; and, Fig. 2 illustrates an exemplary phase-shift detector according to an embodiment of the invention.

Throughout the drawings, the same reference numerals indicate similar or corresponding features or functions.

This invention is premised on the observation that a substantial amount of I/Q channel mismatch can be introduced by the filters used in the transmitter to limit the bandwidth of the I/Q modulated output and by the filters used in the receiver to isolate the transmitted I/Q modulated signal. This internally produced I/Q mismatch is particularly acute in OFDM systems because of the filtering required at each of the plurality of quadrature modulation systems to avoid interference with each other, and because of the filtering required to isolate each of the frequency-division-multiplexed signals.

Fig. 1 shows a block diagram of an exemplary transceiver 300 in accordance with an embodiment of the invention. Transceiver 300 comprises a receiver 100, a transmitter 200 and a digital signal processing unit 400 for processing incoming and out-going communication and data signals.

Receiver 100 comprises a tunable front end 110, which in normal operation, receives an incoming signal transmitted over the wireless medium. The output of front end 110 is demodulated by a quadrature demodulator to provide quadrature output signals Q and I. For ease of illustration, only in-phase branch of the quadrature demodulator is described herein,

the quadrature branch being functionally equivalent but operating at an orthogonal phase provided by the quadrature phase generator 170. The quadrature branch comprises the same symmetrical components as the in-phase branch described herein. Output of front end 110 is demodulated by mixer 120 using an oscillating signal generated by local oscillator 350 generating thereby in-phase signal 1. The demodulated base-band in-phase signal is then filtered by base-band filter 130a. An adjustable gain amplifier 140 provides a base-band analog signal, which is converted into digital samples via the analog-to-digital converter ADC 160a for further processing by DSP 400.

In a similar fashion, transmitter 200 receives two digital streams for I and Q channel modulation and transmission. As with receiver 100, for ease of illustration, only one branch of the quadrate modulator is described herein, i. e. the quadrature branch. The other branch, i. e. the in-phase branch, is functionally equivalent. A digital-to-analog DAC converter 260b converts the samples of the digital input stream received from DSP 400 into an analog signal that is filtered by base-band filter 230b and provided to an adjustable gain amplifier 240. A mixer 220 provides the quadrature modulation of the quadrature signal I or Q with the oscillating signal generated by local oscillator 350. Adder 250 thereafter combines both quadrature modulated signals and amplifier 210 prepares the composite signal for transmission onto the wireless medium.

The quadrature branch of receiver 100 comprises base-band filter 130b symmetrical to base-band filter 130a of the in-phase branch. Both filters 130alb are designed to attenuate signals above a cut-off frequency and as it is known in the art, in addition to providing this frequency-dependent attenuation, filters may introduce a frequency-dependent phase shift.

If both filters 130a and 130b were identical, the phase shifts introduced by each filter 130a and 130b would be identical and the in-phase I and quadrature Q streams would stay in phase relative to each other. In addition, if both filters 130a, 130b are identical, both filters 130a/b would introduce the same gain and the amplitude of the two I and Q signals would remain the same relative to each other. A pair of identical filters 130a/b would not introduce any imbalance or phase-shift between the two branches.

However, if the pair of filters 130a, 130b is not identical, the I and Q streams that are nominally in-phase with each other may exhibit a gain imbalance or a phase-shift relative to each other, particularly at or near the cut-off frequency of the filters, where the substantial frequency-dependent attenuation and frequency-dependent phase shift are introduced in each I and Q branch. If the phase-shift difference between the filters is

substantial, the bit-stream output from the ADC 160a and ADC 160b of the respective I and Q channels will be out-of-phase with each other. The same consequence may be observed in the transmitter 200 if the pair of base-band filters 230a, 230b is non identical.

As noted above, OFDM systems are particularly susceptible to frequency-dependent gain imbalance and phase-shifts. Sharp cutoffs are required to minimize interference and to isolate transmitted signals and substantial gain and phase mismatch between the filters may consequently occur. Although each pair of filters 130a/b and 230a/b is designed to comprise identical filters in each I and Q channels, the fabrication process and the aging of the filters after prolonged use can introduce unpredictable variations in the filters'responses and such variations are enhanced near the cutoff frequency. Indeed, because of the required sharp filter response, minor shifts in the cut-off frequency can introduce substantial phase difference and gain imbalance.

To compensate for the phase shift and gain imbalance potentially introduced by the pairs of filters 130a/b and 230a/b, a compensation scheme has been added to transceiver 300. The compensation scheme enables pre-distorting a pair of signals before transmission by DSP 400 to transmitter 100 such that the applied pre-distortion compensates for the actual phase shift and gain imbalance introduced by the filters pair 130a/b. The compensation scheme also enables post-distorting a pair of digital signals received by DSP 400 from receiver 200 before processing by DSP 400 such that the applied post-distortion compensates for the phase-shift and imbalance previously introduced by the filters pair 230alb. In an embodiment, the pre-distortion and post-distortion are performed by DSP 400 which modifies the respective outgoing and received digital signals with digital compensation signals previously determined for a selected frequency.

In an embodiment, the selected frequency is such that it is one of the components of the sub-band to which the outgoing or received signals belongs. In yet another embodiment, the selected frequency is the center frequency of the spectrum of the sub-band and each component of the sub-band received or transmitted by transceiver 300 is compensated using the digital compensation signals derived as explained hereinafter for that center frequency.

The digital compensation signals may have been obtained from prior testings of the gain and phase mismatches introduced by both pairs of filters for various selected frequencies. DSP 400 may store one or more look-up tables indicating the digital compensation signals for these various frequencies for which the gain and phase mismatches have been measured according to the invention.

In the embodiment of Fig. 1, the compensation signals are determined when transceiver 300 is neither receiving nor transmitting data over the wireless medium. The transceiver 300 is placed on a test mode in which the compensation signals are determined using back loop tests. In the test mode, loops including filters 130a/b and filters 230a/b are formed and test signals generated by DSP 400 are passed through each respective pair of filters 130a/b or 230a/b to determine the respective gain and phase mismatches associated with each pair. The compensation system comprises various switches 510,520, 530 and 540.

Gain and phase mismatch introduced by the receiver filters 130a and 130b are determined as follows. Switch 510 and 520 are positioned so that the output of DAC 260b of the transmitter quadrature branch is connected to the input of filter 130b of the receiver quadrature branch and so that the output of the DAC 260a of the transmitter in-phase branch is connected to the input of the base-band filter 130a of the receiver in-phase branch. Switch 530 and 540 are positioned so that a phase-shift detector 150 is connected to both branches of receiver 100. A detailed embodiment of phase-shift detector 150 is shown in Fig. 2.

During the test, DSP 400 provides a predetennined first square wave signal at a first frequency to filter 130a. The first frequency may be chosen close to the filters'130a/b cut- off frequency since, as explained above, the observed gain and phase mismatches are likely to be greater near the cutoff frequency of the filters. DSP 400 provides a similar second square wave signal shifted by ninety degrees relative to the first square wave signal to filter 130b. Both filtered square wave signals are then passed through phase-shift detector 150 which makes a measurement of the phase shift introduced by the filters pair 130a/b. The operation of phase-shift detector 150 is explained in further details below with reference to Fig. 2. The filtered square wave signals are further passed to ADC 160a and 160b and DSP 400, which measures the gain mismatch introduced by the pair 130a/b.

A compensation signal is therefore obtained for this first selected frequency and the compensation signal is stored in DSP 400 for later use. DSP 400 will use the compensation signal derived for the first frequency during operation of transceiver 300 for pre-distorting the I and Q signals of a component that is in a sub-band whose spectrum comprises the first frequency. Other measurements of the gain and frequency mismatches may be carried out for other frequencies using square wave signals at these other frequencies. Other compensation signals may then be derived for these other frequencies and will be used by

DSP to compensate I and Q spectral sub-band components that belong to the same sub-band as components at these other frequencies.

It is also within the scope of the invention to use other types of waveforms and the invention is by no way limited to the measurement of the gain and phase mismatch using square wave signals only. However, the third and fifth harmonics of a square wave signal are at a higher frequencies than the principal harmonic of the square wave and situated outside the passing band of the filters. These third and fifth harmonics will thus be attenuated when the square wave signal will be filtered. One skilled in the art can easily generate a square waveform from the DSP as opposed to other types of waveforms.

Depending on the positions of switch 530,540 phase detector 150 may receive either in-phase and quadrature signals after they have been filtered by base-band filters 130a/b of receiver 100 or phase detector 150 may receive in-phase and quadrature signals after they have been filtered by base-band filters 230a/b of transmitter 200. Referring now to Fig. 2, detector 150 comprises switches 151,153, 155,157 and multiplier 156. When the four switches 151,153, 155 and 157 are in the lower position, multiplier 156 receives the filtered in-phase I and quadrature Q signals and multiplier 156 outputs on line 152 a parameter or signal representative of the phase shift between the two filtered signals I and Q and, as a consequence, of the phase shift introduced by the filters pair 130a/b or 230alb.

When the four switches 151,153, 155 and 157 are in the upper position, detector 150 only transmits inputted filtered signals I and Q and the phase-shift is not measured. Switches 151,153, 155 and 157 may be placed in the upper positions when transceiver 300 is in normal communication reception or transmission mode.

Similar tests may be carried out in order to detect the phase-shift and gain imbalance introduced by the pair of filters 230a/b. In such case, switches 540 and 530 are positioned so that phase-shift detector 150 is connected to the output of filters 230a and 230b. Switches 520 and 530 are positioned so that DSP 400 and DAC 260a/260b provide the square wave signal and its shifted version to filters 230a and 230b, respectively.