On-signal quadrature modulator calibration

[0001] 1. Field of the Invention

[0002] The present invention relates to calibration of a transmitter to remove distortions, and more particularly to on-signal quadrature modulator calibration using an output signal detector to remove DC, phase and amplitude imbalances.

[0003] 2. Description of the Related Art

[0004] Many modern communications systems combine in-phase (I) and quadrature phase (Q) components in a transmitted signal. The I and Q components can be generated and combined in the digital domain without distortion. The combined digital signal must then be converted to an analog signal prior to up-conversion and transmission, which generates an undesired image that is difficult and expensive to eliminate. Alternatively, the I and Q components may be separately converted to analog signals, which are then combined in the analog domain. The analog processing components are not perfectly matched, so that several types of distortion are introduced into the signal path. These distortions include DC offsets and amplitude and phase imbalances. For example, after conversion to analog, an I DC offset is added to the I signal channel and a separate Q DC offset is added to the Q signal channel. One or more mixer stages are used to up convert each of the analog signals to a radio frequency (RF) level suitable for transmission. The local oscillator (LO) splitter carrier signals and the mixer stages are not perfectly matched resulting in amplitude and phase imbalances that cause distortions in the transmitted signal.

[0005] The distortions, caused by DC, amplitude and phase imbalances, degrade the performance of digital communication systems including systems employing digital signal processing (DSP) techniques. Amplitude and phase imbalances cause variations in constellation plots that degrade communication effectiveness. Large gain or phase imbalances can effectively disable reliable communications, particularly higher density modulation techniques used to achieve higher data transmission rates.

[0006] In many legacy systems, distortions have been measured and calibrated at the time of manufacture. Imbalances, however, can drift over time due to various factors including, for example, temperature and aging. Static compensation at the time of manufacture does not address the dynamic factors that affect radio operation during use. Several dynamic compensation techniques are known. Some methods use off-line signals, such as training signals or the like, which require taking the system off-line for calibration. Such off-line methods degrade overall system efficiency by consuming valuable time to perform calibration and compensation. Vital packets may be missed during the calibration process since the antenna(s) must be disconnected or otherwise disabled to prevent inadvertent transmissions. Some calibration techniques negatively impact the wireless medium, such as undesirable signal splattering. It is desired to provide an automatic and relatively simple imbalance compensation solution that operates on the actual communication signals (e.g., on-signal) without having to go off-line or perform separate calibration cycles.

[0007] An on-signal calibration system according to an embodiment of the present invention calibrates in-phase (I) and quadrature phase (Q) signals of a transmitter to remove distortions in the radio frequency (RF) output signal of the transmitter. The calibration process is performed during transmission and does not require separate calibration cycles. The transmitter generates I and Q values and converts, modulates and combines the I and Q values into the RF output signal for transmission. The calibration system includes a detector, a sampler, a selector, an imbalance estimator, and an IQ corrector. The detector senses the RF output signal and provides a detection signal indicative thereof. The sampler samples the detection signal and provides digital samples. The selector selects from among the digital samples that correspond to predetermined ranges of the I and Q values, or otherwise predetermined selection boxes at predetermined phases. The imbalance estimator determines at least one imbalance estimate based on selected digital samples. The IQ corrector corrects the I and Q values using at least one imbalance estimate.

[0008] In various configurations, the detector is an envelope detector, the sampler is an analog to digital converter (ADC) and the selector selects digital samples. The digital samples that are selected depend upon the particular imbalance targeted. For the I channel DC offset, the selected digital samples may correspond with first and second selection boxes symmetrically located on either side of an I/Q origin on the I-axis of a plot of the I and Q values. For the Q channel DC offset, the selected digital samples may correspond with third and fourth selection boxes symmetrically located on either side of the I/Q origin on the Q-axis. For the amplitude imbalance, a combination of the digital samples of the first and second selection boxes and a combination of the digital samples of the third and fourth selection boxes may be used. For the phase imbalance, the selector may select the digital samples that correspond with selection boxes symmetrically located on either side of an I/Q origin on a 45 degree axis of a plot of the I and Q values and selection boxes that are symmetrically located on either side of the I/Q origin on a 135 degree axis.

[0009] The calibration system may include a power circuit that determines digital power values, where the selector selects from among the digital power values. In one embodiment, the imbalance estimator determines at least one imbalance estimate based on a ratio of selected digital power values. Alternatively, the calibration system may include a magnitude circuit that determines digital magnitude values, where the selector selects from among the digital magnitude values. In this latter case, the imbalance estimator determines the at least one imbalance estimate based on a ratio of the selected magnitude values.

[0010] A transmitter according to an embodiment of the present invention includes a baseband processor providing I and Q signals, an RF quadrature modulator, and an output signal detector. The RF quadrature modulator coverts the I and Q signals into an RF output signal. The output signal detector senses a characteristic of the RF output signal indicative of the at least one imbalance and outputs a characteristic signal. The baseband processor includes an I/Q corrector, a hit detector, a selector, and an imbalance estimator. The I/Q corrector corrects the I and Q signals using at least one imbalance metric. The hit detector generates gate signals indicative of predetermined ranges of the I and Q signals. The selector selects portions of the characteristic signal based on the gate signals. The imbalance estimator uses selected portions of the characteristic signal to determine the at least one imbalance metric.

[0011] In various embodiments, the output signal detector is an envelope detector and the characteristic signal is an envelope signal. The baseband processor includes a core that generates I and Q digital values and a sampler that samples the envelope signal and that provides digital samples. In one embodiment, the hit detector provides the gate signals indicative of a plurality of symmetric selection boxes at predetermined phases of the I and Q digital values, and the selector selects from among the digital samples based on the gate signals.

[0012] A method of on-signal calibration of an RF quadrature modulator according to an embodiment of the present invention includes detecting the RF output signal and providing a detection signal, sampling the detection signal and providing digital samples, selecting from among the digital samples corresponding to predetermined ranges of the I and Q digital values, estimating at least one imbalance metric using selected digital samples, and calibrating the I and Q signals using the at least one imbalance metric. In various embodiments, the detection signal is an envelope signal and the predetermined ranges of the I and Q digital values correspond with selection boxes at selected phases, such as, for example, 0, 45, 90, 135, 180, 225, 270, and 315 degrees.

[0031] The following description is presented to enable one of ordinary skill in the art to make and use the present invention as provided within the context of a particular application and its requirements. Various modifications to the preferred embodiment will, however, be apparent to one skilled in the art, and the general principles defined herein may be applied to other embodiments. Therefore, the present invention is not intended to be limited to the particular embodiments shown and described herein, but is to be accorded the widest scope consistent with the principles and novel features herein disclosed.

[0032] The inventors of the present application have recognized the need for calibrating the modulator of a transmitter. They have therefore developed an on-signal quadrature modulator calibration system and method that calibrates the modulator in real-time while transmitting actual signals without the need for separate calibration cycles, as will be further described below with respect to FIGS. 1-13.

[0033] The present disclosure describes an on-signal quadrature modulator calibration scheme which uses a detector that senses information from the RF output signal that is used for calibration. The detection signal is fed back and processed to calculate one or more imbalance metrics, which are then used to calibrate the I and Q signals. DC, amplitude and phase imbalances are measured and calibrated. In one embodiment, a simple envelope detector is used as calibration feedback. Alternatively, a power detector is used. Only detector feedback monotonicity is necessary, and the present scheme converges more quickly with increased feedback linearity. The detection signal is sampled and the resulting digital samples, or derived power/magnitude values, are selected and sorted according to predetermined ranges of the I and Q values generated by the baseband processor. The predetermined ranges correspond to symmetrical selection boxes at optimal locations to measure the imbalances. An imbalance estimator determines one or more imbalance metrics which are used for calibration as further described herein.

[0034] A calibration scheme according to embodiments of the present invention operates during packet transmission rather than requiring separate calibration cycles. In particular embodiments specific to the IEEE 802.11 standard, the calibration scheme operates for 2 megabits per second (Mbps) DSSS payloads, 5.5 and 11 Mbps CCK payloads, and for OFDM packets. The envelope is preferably processed at 20 megahertz (MHz) for OFDM (IEEE 802.11a) and at 22 MHz for IEEE 802.11b. In the 802.11 schemes, the calibration is not performed during 1 Mbps DSSS preambles, headers or payloads since BPSK does not utilize the full I/Q plane. This is relatively inconsequential, however, since BPSK is significantly more robust than higher data modulations so that calibration is less critical.

[0035] FIG. 1 is a simplified block diagram of a transmitter 100 including on-signal quadrature modulator calibration implemented according to an exemplary embodiment of the present invention. The present invention is illustrated using transmitters implemented according to the IEEE 802.11 family of standard(s) (including, for example, 802.11a, 802.11b and 802.11g) in a zero-intermediate frequency (ZIF) quadrature modulator (where ZIF is otherwise known as direct conversion), although it is understood that the present invention is applicable to other communication methods. The transmitter 100 includes a baseband processor 101 and an RF quadrature modulator 103. Only applicable portions of the transmitter are shown, in which other portions (e.g., receiver, media access control (MAC), antenna, etc.) are omitted as not necessary for providing a complete description of the invention.

[0036] The baseband processor 101 develops an in-phase digital signal I_D and a quadrature-phase digital signal Q_D. In an ideal system, the I_D signal is provided directly to an I-channel digital to analog converter (I DAC) 107 for converting to analog format, shown as I_A. The analog I_A signal is then provided to one input of a mixer 113, which receives a modulating carrier signal cos(ω_ct) at its other input. A local oscillator (LO) 125 generates an LO signal, which is provided to a 0/90 degree oscillator splitter 127. The oscillator splitter 127 generates the cos(ω_ct) signal and another carrier signal −sin(ω_ct) signal as known to those of ordinary skill in the art. The term ω_C=2πf_C measured in radians, where f_C is the LO carrier frequency. The output radio frequency (RF) signal in the I signal path, I_RF, is provided to one input of a summing junction 115, which provides an RF output signal (OUT) at its output. A power amplifier 129 amplifies the OUT signal to a signal AOUT, which is amplified to the appropriate power level for transmission through the selected transmission medium.

[0037] The Q_D signal is processed in a similar manner. In the ideal system, the Q_D signal is provided directly to a Q-channel digital to analog converter (Q DAC) 117 for converting to analog format, shown as Q_A. The analog Q_A signal is then provided to another LPF 119 within the quadrature modulator 103, which provides a filtered Q channel signal to one input of another mixer 123. The mixer 123 receives the carrier signal −sin(ω_ct) at its other input from the oscillator splitter 127. The mixer 123 outputs a second RF signal in the Q signal path, shown as Q_RF. The Q_RF signal is provided to the other input of the summing junction 115, which combines the I_RF and Q_RF signals to generate the OUT signal. In the 802.11 configuration illustrated, the OUT signal is amplified by the power amplifier 129 to AOUT, which is transmitted in a wireless medium via an antenna (now shown).

[0038] The quadrature modulator 103 is not ideal. An adder 111 is inserted in the I channel between the I DAC 107 and the mixer 113 denoting the addition of an undesired DC offset signal I_DC. Similarly, another adder 121 is inserted in the Q channel between the Q DAC 117 and the mixer 123 denoting the addition of an undesired DC offset signal Q_DC. The respective carrier signals are intended to be exactly 90 degrees out of phase and to have identical amplitudes. Real world devices, however, cause amplitude and phase imbalances which result in signal distortions. Amplitude and phase imbalances (distortions) are represented as modifying the −sin(ω_ct) signal resulting in a distorted signal −(1+δ)sin(ω_ct+θ), where the delta term “δ” denotes an amplitude imbalance and the theta term “θ” denotes a phase imbalance. Since the I and Q channels carry relative signals, the amplitude and phase imbalances are accurately represented in the Q channel as affecting the sinusoidal term at the mixer 123. The DC offsets, however, are illustrated as separate DC offsets added into the respective I and Q signal paths. At least one source of the amplitude and phase imbalances is the oscillator splitter 127, which generates the sine and cosine carrier waveforms with amplitude and phase discrepancies. The mixers 113 and 123 are also imperfect generating additional imbalances.

[0039] In the illustrated embodiment, an envelope detector 131 senses the OUT signal and provides a corresponding analog envelope E_A signal to an analog-to-digital converter (ADC) 133 in the baseband processor 101. The envelope detector 131 extracts instantaneous amplitude information of the RF envelope of the OUT signal. The envelope detector 131 is relatively simple and easy to build in a low-cost manner, and may include a diode or rectifier or the like, such as a hot-carrier diode or full-wave rectifier. In one embodiment, the envelope detector 131 outputs absolute-value amplitude information so that the E_A signal is always a positive value.

[0040] The ADC 133 samples and converts the analog E_A signal into a corresponding stream of digital samples E_D, which is provided to a quadrature modulator calibration circuit 135. In one embodiment, the ADC 133 is configured to sample the E_A signal at the appropriate sample rate corresponding to the signal type being transmitted, such as, for example, 20 MHz for OFDM signals or 22 MHz for IEEE 802.11b 5.5 and 11 Mbps CCK signals. The quadrature modulator calibration circuit 135 provides a set of distortion estimate signals to an IQ correction circuit 105 provided within the baseband processor 101. The distortion estimate signals include estimates of the distortion I_DC, Q_DC, δ, and θ, shown as IDC_EST, QDC_EST, δ_EST, and θ_EST, respectively. The IQ correction circuit 105 receives and modifies the digital I_D and Q_D signals into calibrated IC_D and QC_D signals prior to conversion to analog format by the I and Q DACs 107 and 117, respectively.

[0041] FIG. 2 is a more detailed block diagram of the IQ correction circuit 105 shown in mathematical format. The IQ correction circuit 105 includes a combiner 201 that multiplies the Q_D signal by a phase term sin(θ_EST)/cos(θ_EST) and an adder 203 in the I signal path which subtracts the resulting value Q_D sin(θ_EST)/cos(θ_EST) from the I_D values. The output of the adder 203 is provided to the positive input of another adder 205, which subtracts the IDC_EST value to remove the I channel DC offset. The output of the adder 205 provides the IC_D digital signal. In the Q signal path, a first combiner 207 multiplies the Q_D values by a phase term 1/cos(θ_EST), and a second combiner 209 that multiplies the output of the combiner 207 by an amplitude term 1/(1+δ_EST). The output of the combiner 209 is provided to the positive input of another adder 211, which subtracts the QDC_EST value to remove the Q channel DC offset. The output of the adder 211 provides the QC_D digital values.

[0042] FIG. 3 is a block diagram of another transmitter 300 including on-signal quadrature modulator calibration implemented according to an exemplary embodiment of the present invention. Similar elements and components as those of the transmitter 100 are given the same numeric designations. The transmitter 300 is substantially the same as the transmitter 100 except that the envelope detector 131 removed and a low-pass filter 301 is coupled directly to the power amplifier 129 for sensing an envelope of the AOUT signal, which is provided as an envelope signal E to the LPF 301. The LPF 301 filters the E signal and provides the E_A signal to the ADC 133. The power amplifier 129 is implemented similar to many off-the-shelf amplifier devices and includes an output sensing device, such as a rectifier or diode or the like. In this manner, the built-in sensing device of the power amplifier 129 is employed to sense the AOUT signal rather than adding a separate envelope detector for detecting the OUT signal. A possible disadvantage of employing the output sensing device of the power amplifier 129 is that power amplifiers tend to be non-linear devices, so that it may incorporate additional distortion.

[0043] In yet another embodiment, the envelope detector 131 is replaced with a power detector (not shown) that provides a power signal which is proportional to the square of the corresponding envelope signal. The power signal may be generated from either the OUT signal or the AOUT signal and a LPF is used to filter the power signal. Depending upon the particular configuration, the power signal may be used directly or squared prior to further processing by the baseband processor 101. In the general case, an output signal detector senses a characteristic of the output signal and provides a characteristic signal, which incorporates the imbalance information used for purposes of calibration.

[0044] FIG. 4A is a block diagram of an exemplary circuit 400 employed to estimate the I channel DC offset I_DC, where the circuit 400 comprises a portion of the quadrature modulator calibration circuit 135 according to one embodiment of the present invention. The I_D and Q_D digital values are provided to a digital delay block 401, which outputs delayed versions thereof, shown as I_DD and Q_DD, respectively. The I_DD and Q_DD delayed digital values are provided to a HIT/BIN detection block 403, which selects E_D samples according to a predetermined selection criterion based on the I_DD and Q_DD values.

[0045] FIG. 4B is a graph diagram of a random modulation signal 405 plotted on an I/Q graph illustrating the selection criterion employed by the detection block 403 according to one embodiment. A pair of symmetrical selection boxes M0 and M180 are shown super-imposed on the graph on either side of the graph origin O along the I axis. The selection boxes M0 and M180 are approximately equidistant from the origin and have approximately the same size and shape. In particular, box M0 is at 0 degrees on the positive I side and box M180 is at 180 degrees on the negative I side. The selection criterion is that samples falling within the boxes M0 and M180 are used for I_DC calibration and the samples falling outside the box are rejected for purposes of I_DC calibration.

[0046] The selection box M0 is located at phase 0 degrees and the selection box M180 is located at phase 180 degrees along the I axis. The height and width of the selection boxes corresponding to predetermined ranges of the I and Q digital values to optimize a determination of the I DC offset. The width of each of the selection boxes M0 and M180 along the I axis is determined to ensure that a sufficient number of samples are collected given expected real-time sample values over time within a predetermined amount of error based on expected or otherwise experimentally-determined levels of distortion. The height of the selection boxes M0 and M180 above and below the I axis in the positive and negative Q directions is selected sufficiently small to ensure valid samples intended to be located on or near the I axis within a predetermined amount of error based on the expected levels of distortion. The width and height of the selection boxes M0 and M180 are not shown to scale; the height may be chosen to be significantly smaller compared to the width.

[0047] In general, it is determined that the average value of the magnitudes of the E_D samples corresponding to the selection box M0 will be approximately equal to the magnitude of the E_D samples corresponding to the selection box M270, if I_DC is negligible or zero and if the boxes are approximately the same size and shape and are symmetrically located. A discrepancy between the average values is proportional to the amount of I_DC of the signal.

[0048] The detection block 403 operates as a selector with a gating function in which it asserts a hit signal H1 for each time the I_DD and Q_DD values define a point that when plotted is located within either of the selection boxes M0 and M180. The detection block 403 asserts a bin signal B1 that identifies one of the selection boxes M0 and M180 in which the hit occurred. A bin de-multiplexer 407 selects each E_D sample when indicated as a “hit” by the H1 signal asserted by the detection block 403. The bin de-multiplexer 407 sorts selected samples into two “bins”, including a first bin corresponding to samples for the selection box M0 and a second bin corresponding to samples for the selection box M180 as indicated by the B1 signal.

[0049] As previously described, the I_D and Q_D digital values are converted to analog signals which are processed and filtered within the quadrature modulator 203 to develop the RF output signals. The selected output signal is monitored by the envelope detector 131 or the like to provide the E_A signal, which is then sampled by the ADC 133 to provide the E_D digital samples. The approximate delay of the processing loop between the I_D and Q_D values and the corresponding E_D samples is compensated by the digital delay block 401, which outputs the delayed I_DD and Q_DD values. In this manner, the timing of the I_DD and Q_DD values approximately corresponds to the timing of the E_D samples at the bin de-multiplexer 407. The amount of delay depends upon the particular configuration and implementation of the baseband processor 101 and the quadrature modulator 103 and the particular signal being transmitted. In one embodiment specific to an OFDM signal samples at 20 MHz, for example, it was determined using correlation results that the delay is approximately 7 sample times (e.g., approximately 350 nanoseconds (ns) when each sample time is approximately 50 ns) between the I_D, Q_D values and the corresponding E_D samples. In one embodiment, for example, the delay box 401 is implemented using a first-in, first-out (FIFO) set of 7 shift registers in which each register is sufficiently large to store a pair of I_D, Q_D values. Thus, the I_D, Q_D values are delayed by 7 samples and the corresponding I_DD and Q_DD values provided to the detection block 403. It is appreciated that other delay methods may be employed, including more accurate delay methods to arrive at a more accurate correlation, as further described below.

[0050] The bin de-multiplexer 407 outputs a first set of selected samples E_D0 corresponding to the selection box M0 and outputs a second set of selected samples E_D180 corresponding to the selection box M180. The E_D0 samples are filtered or otherwise averaged by a first average block 409 for providing a first sample average EA_D0 for the selection box M0 and a second sample average EA_D180 for the selection box M180. The EA_D0 and EA_D180 values are used to derive the IDC_EST value, which is an estimate of the I channel DC offset. In particular, one-half of the difference between the first and second sample averages E_AD0 and EA_D180 is determined at calculator block 413, which outputs the IDC_EST value. The IDC_EST value is an estimate of the I channel DC offset I_DC, so that the IDC_EST value is subtracted from the I signal path in the IQ correction circuit 105 to remove the I channel DC offset. Each time a hit occurs, the corresponding E_DO or E_D180 value is updated, the corresponding EA_DO or EA_D180 value is updated, and the IDC_EST value is updated to track the I DC offset.

[0051] FIG. 5A is a block diagram of an exemplary circuit 500 employed to estimate the Q channel DC offset Q_DC, where the circuit 500 comprises a portion of the quadrature modulator calibration circuit 135 according to one embodiment of the present invention. The circuit 500 is substantially the same as the circuit 400 except modified to measure Q_DC rather than I_DC. In this case, the detection block 403 is replaced with detection block 503, which operates in a similar manner but with a modified selection criterion specific to measuring Q_DC. Again, the I_DD and Q_DD values are provided to the detection block 503, which outputs a hit signal H2 and a bin signal B2 collectively used to select E_D samples.

[0052] FIG. 5B is a graph diagram of the random modulation signal 405 plotted on the I/Q graph illustrating the selection criterion employed by the detection block 503. A pair of symmetrical selection boxes M90 and M270 are shown super-imposed on the graph on either side of the graph origin O along the Q axis. The selection boxes M90 and M270 are approximately equidistant from the origin and have approximately the same size and shape. In particular, box M90 is at 90 degrees on the positive Q side and box M270 is at 270 degrees (or −90 degrees) on the negative Q side. The selection criterion is that samples falling within the boxes M90 and M270 are used for Q_DC calibration and the samples falling outside the box are rejected for purposes of Q_DC calibration.

[0053] The selection box M90 is located at phase 90 degrees and the selection box M180 is located at phase 180 degrees along the Q axis. The height and width of the selection boxes corresponding to predetermined ranges of the I and Q digital values to optimize a determination of the Q DC offset. @In one embodiment, the selection boxes M90 and M270 are substantially the same size and shape as the selection boxes M0 and M180, except rotated by 90 degrees. For example, the width of the selection boxes M90 and M270 along the Q axis is determined to ensure that a sufficient number of samples are collected given expected real-time sample values over time within a predetermined amount of error based on expected levels of distortion. The “height” of the selection boxes M90 and M270 on either side of the Q axis in the positive and negative I directions is selected sufficiently small to ensure valid samples intended to be located along the Q axis within a predetermined amount of error based on the expected levels of distortion.

[0054] In general, it is determined that the average value of the magnitudes of the E_D signal samples corresponding to the selection box M90 will be approximately equal to the magnitude of the E_D signal samples corresponding to the selection box M270, if Q_DC is negligible or zero and if the boxes are approximately the same size and shape and are symmetrically located. A discrepancy between the average values is proportional to the amount of Q_DC of the signal.

[0055] The H2 and B2 signals are provided to a bin de-multiplexer 507, which operates in substantially the same manner as the bin de-multiplexer 407 by selecting each sample of the E_D signal when indicated as a “hit” by the H2 signal asserted by the detection block 503. The B2 signal identifies the corresponding selection box M90 or M270 in a similar manner as the B1 signal. The bin de-multiplexer 507 also sorts selected samples into two “bins”, including a first bin corresponding to samples for the selection box M90 and a second bin corresponding to samples for the selection box M270. The bin de-multiplexer 507 outputs first selected samples E_D90 corresponding to the selection box M90 and outputs second selected samples E_D270 corresponding to the selection box M270. The E_D90 samples are filtered or otherwise averaged by a first average block 509 for providing a first sample average EA_D90 for the selection box M90 and a second sample average EA_D270 for the selection box M270. The EA_D90 and EA_D270 values are used to derive an estimate of Q_DC. In particular, one-half of the difference between the first and second sample averages EA_D90 and EA_D270 is determined at calculator block 513, which outputs the Q channel DC estimate QDC_EST, which is an estimate of the Q channel DC offset. The QDC_EST value is subtracted from the Q signal path in the IQ correction circuit 105 in an attempt to remove the Q channel DC offset. Each time a hit occurs, the corresponding E_D90 or E_D270 value is updated, the corresponding EA_D90 or EA_D270 value is updated, and the QDC_EST value is updated to track the Q DC offset.

[0056] FIG. 6A is a block diagram of an exemplary circuit 600 employed to estimate the amplitude distortion delta term “δ”, where the circuit 600 comprises a portion of the quadrature modulator calibration circuit 135 according to one embodiment of the present invention. The I_DD and Q_DD values from the delay block 401 are provided to a HIT/BIN detection block 603, which selects E_D samples according to a predetermined selection criterion based on the I_DD and Q_DD values.

[0057] FIG. 6B is a graph diagram of the random modulation signal 405 plotted on an I/Q graph illustrating the selection criterion employed by the detection block 603. In this case, two pairs of symmetrical selection boxes are shown, including a first pair 605a and 605b positioned on either side of the origin O along the I axis and another pair 607a and 607b above and below the origin along the Q axis. In one embodiment, the selection boxes 605a,b and 607a,b are configured in substantially the same manner as the selection boxes M0, M180 and M90, M270, respectively, as previously described. The detection block 603 asserts a hit signal H3 and a bin signal B3 in a similar manner as the H1, B1 and H2, B2 signals when a E_D sample falls within any of the selection boxes. For the circuit 600, the B3 signal indicates only one of two bins, including a first bin for hits within either of the selections boxes 605a and 605b and another bin for hits within either of the selection boxes 607a and 607b. In other words, a hit in either of the selection boxes 605a or 605b is sorted to a first bin and a hit in either of the selection boxes 607a or 607b is sorted to a second bin.

[0058] The I_DD and Q_DD values from the delay block 401 are also provided to an amplitude square block 609, which outputs amplitude squared values A²=I_DD²+Q_DD². In one embodiment, each of the I_DD and Q_DD values are separately squared and then added together to calculate A². It is appreciated, however, that the I_DD and Q_DD values are known according to a selected modulation scheme, so that every possible combination of A² may be pre-calculated and stored in a memory. In an efficient circuit implementation, for example, the amplitude square block 609 is configured as a lookup table in which each pair of I_DD and Q_DD values are used as index values applied to the lookup table, which outputs the corresponding pre-stored A² value. The E_D samples are also squared in a separate square block 611, which outputs corresponding E_D² values. The A² values and corresponding E_D² values are provided to a divide block 613, which divides each E_D² value by its corresponding A² value, and outputs corresponding power values P to a bin de-multiplexer block 615. The blocks 609, 611 and 613 collectively form a power circuit for determining the power values P.

[0059] The bin de-multiplexer block 615 selects the P values that are indicated as a hit by the H3 signal and sorts into two bins according to the B3 signal, and outputs the selected P values as P_I power values if the hit was within the selection boxes 605a or 605b or as P_Q power values if the hit was within the selection boxes 607a or 607b. The set of P_I power values are averaged by average block 617 and the set of P_Q power values are averaged by average block 619, and the averaged power values PA_Q and PA_I are provided to a power divide block 621, which calculates PA_Q/PA_I≈1+2δ_EST. As previously described, the δ_EST term is an estimate of the amplitude distortion term and “≈” denotes an approximation. The corresponding δ_EST term is determined by subtracting 1 from each PA_Q/PA_I value and then dividing by 2, where the resulting δ_EST terms are then provided to the IQ correction circuit 219 for amplitude compensation. It is noted that since the output values (1+2δ_EST) are provided in digital form, it is a relatively simple matter of determining the δ_EST term by decrementing a register by 1 and shifting to divide by 2.

[0060] FIG. 7 is a block diagram of another exemplary circuit 700 employed to approximate the amplitude distortion delta term δ according to an alternative embodiment of the present invention. The circuit 700 is an alternative to and somewhat simplified version of the circuit 600 for calculating the amplitude distortion delta term δ_EST. The delay block 401 and detection block 603 are used in a similar manner for generating the H3 and B3 signals. In this case, the I_DD and Q_DD values from the delay block 401 are provided to an amplitude block 701, which outputs amplitude values A=SQRT(I_DD²+Q_DD²), where “SQRT” denotes the square-root mathematical function. In one embodiment, each A value is calculated by separately squaring each I_DD value and Q_DD value, adding the results together and then taking the square root. In another, more efficient implementation, the amplitude block 701 is configured as a lookup table in which the I_DD and Q_DD values are used as index values applied to the lookup table, which outputs the corresponding pre-stored A value. The A values and corresponding E_D values are provided to a divide block 703, which divides each E_D value by its corresponding A value, and outputs corresponding magnitude values M to a bin de-multiplexer block 705. The blocks 701 and 703 collectively form a magnitude circuit for providing the magnitude values M.

[0061] The bin de-multiplexer block 705 selects the M values that are indicated as a hit by the H3 signal, sorts into two bins according to the B3 signal, and outputs the M values as M_I magnitude values for hits within either of the selection boxes 605a or 605b or as M_Q magnitude values for hits within the either of the selection boxes 607a or 607b as indicated by the B3 signal. The M_I values are averaged by average block 707 and the M_Q values are averaged by average block 709, and the averaged magnitude values MA_I and MA_Q are provided to a magnitude divide block 711, which calculates MA_Q/MA_I≈1+δ_EST. The corresponding δ_EST term may be determined by subtracting 1 from each MA_Q/MA_I value, where the resulting δ_EST terms are then provided to the IQ correction circuit 135 for amplitude compensation. Alternatively, the 1+δ_EST values may be used directly without further modification. The circuit 700 is potentially a more efficient implantation than the circuit 600 since the square block 611 is eliminated and the output values (1+δ_EST) may be used without modification.

[0062] The circuit 700 may be modified by replacing the amplitude block 701 with a selection block 713, shown in dashed lines. The selection block 713 selects the I_DD value when the hit occurs within the selection boxes 605a or 605b and selects the Q_DD value when the hit occurs within the selection boxes 607a or 607b. The selection block 713 is sufficiently accurate when the selection boxes 605a and 605b incorporate relatively small or otherwise negligible Q_DD values and when the selection boxes 607a and 607b incorporate relatively small or otherwise negligible I_DD values. The selection block 713 incorporates relatively simple logic and may be preferable to the logic calculation or lookup table implementations of the amplitude block 701.

[0063] FIG. 8A is a block diagram of an exemplary circuit 800 employed to estimate the phase distortion theta term “θ”, where the circuit 800 comprises a portion of the quadrature modulator calibration circuit 135 according to one embodiment of the present invention. The circuit 800 is similar to the circuit 600 in which similar components assume identical reference numbers. In particular, the delay block 401, the amplitude square block 609, the square block 611 and the divide block 613 operate in substantially the same manner in which the divide block 613 outputs power sample values P as previously described. The I_DD and Q_DD values from the delay block 401 are provided to a detection block 803, which selects samples according to a predetermined selection criterion.

[0064] FIG. 8B is a graph diagram of the random modulation signal 405 plotted on an I/Q graph illustrating the selection criterion employed by the detection block 803. Again, two pairs of selection boxes are shown, including a first pair 805a and 805b positioned along a 45 degree axis on either side of the origin O in symmetrical fashion, and another pair 807a and 807b positioned along a 135 degree axis on either side of the origin O in symmetrical fashion. More particularly, the selection box 805a is at a phase of 45 degrees, the selection box 807a is at a phase of 135 degrees, the selection box 805b is at a phase of 225 degrees (or −135 degrees), and the selection box 807b is at a phase of 315 degrees (or −45 degrees). In one embodiment, the selection boxes 805a,b and 807a,b are configured in a similar manner as the selection boxes 605a,b and 607a,b as previously described. Alternatively, the selection boxes 805a,b and 807a,b may have more of a square shape in which their heights and widths are more equal. In any event, the symmetrical selection boxes are selected to optimize a determination of the phase imbalance in the output signal.

[0065] The detection block 803 asserts a hit signal H4 and a bin signal B4 when a sample falls within any of the selection boxes 805a,b and 807a,b. The B4 signal indicates only one of two bins, including a first bin for either of the selections boxes 805a and 805b and a second bin for either of the selection boxes 807a and 807b. In other words, a hit in either of the selection boxes 805a or 805b is sorted to a first bin and a hit in either of the selection boxes 807a or 807b is sorted to a second bin as indicated by the B4 signal.

[0066] The P power values from the divide block 613 are provided to a bin de-multiplexer block 809, which receives the H4 and B4 signals. The bin de-multiplexer block 809 selects the P values that are indicated as a hit and sorts into two bins, and outputs selected P values as P₄₅ power values if the hit was within the selection boxes 805a or 805b or as P₁₃₅ power values if the hit was within the selection boxes 807a or 807b. The P₄₅ power values are averaged by average block 811 and the P₁₃₅ power values are averaged by average block 813. The averaged power values PA₄₅ and PA₁₃₅ are provided to a differential divide block 815, which calculates corresponding estimated phase distortion theta terms θ_EST=(PA₁₃₅−PA₄₅)/(PA₁₃₅+PA₄₅). The θ_EST terms are estimates of the phase distortion theta term θ, which are provided to the IQ correction circuit 105 for phase compensation.

[0067] FIG. 9 is a block diagram of an exemplary embodiment of an envelope detector 900 that may be used as the envelope detector 131. The envelope detector 900 includes a full-wave rectifier 901 receiving the OUT signal and providing an unfiltered envelope signal E to a low-pass filter (LPF) 903. The LPF 903 is configured with a wide enough frequency range to pass the desired envelope of the OUT signal, and outputs the E_A signal. As previously described, the E_A signal is sampled and converted to the digital E_D signal by the ADC 133. The sample rate of the ADC 133 is selected to correspond with the signal type being transmitted. Different sample rates may be used if rate-change filtering is applied in the transmit signal path as known to those of ordinary skill in the art.

[0068] FIG. 10 is a block diagram of an exemplary correlation circuit 1000 that is used in alternative embodiments to achieve more accurate time alignment using polyphase filters 1007. The amount of delay through the analog RF modulator and associated filters may not be precisely aligned with E_D samples from the ADC 215. In a worst case scenario, the E_D samples may be aligned in time in the middle of the corresponding I_DD and Q_DD delayed values. For example, rather than a 7 sample time delay, the delay may be a time corresponding to 6.5 or 7.5 sample times, which may lead to inaccurate results since the digital samples in the feedback compensation loop are not aligned. One solution is to configure the delay block 401 to be adjustable between minimum and maximum expected delay increments, set the delay increment to just above the actual delay when measured, and then add a fractional delay device in the feedback path using a filter or an interpolator or the like. The fractional delay device is able to shift the delay time by less than a full sample time to achieve more precise alignment. For example, if it is expected that the delay for a given configuration ranges between approximately 6 and 8 sample times give or take a sample, the delay block 401 is configured to be programmable within a range of sample times, such as between 5 and 9 sample times. The actual delay is measured and the delay block 401 is programmed to the incremental delay just greater than the actual delay. If the delay is measured at 6.5 sample times, for example, then the delay block 401 is programmed with a 7 sample delay. The fractional sample delay device in the feedback path is then adjusted and programmed with a fractional sample delay as close as possible to the difference between the actual delay and the programmed delay of the delay block 401, e.g., 0.5 sample time, so that the differential between the delays provides more precise timing alignment.

[0069] In the embodiment shown, the I_DD and Q_DD values from the delay block 401 are provided to an amplitude square block 1003, which operates in substantially the same way as the amplitude square block 609 for generating the amplitude square values A². The E_D samples from the ADC 215 are provided through a switch 1005 to the input of a selected one of multiple polyphase filters 1007. Each of the polyphase filters 1007 is configured with a different delay, where each delay is a fraction of a full sample time. As shown, for example, four separate polyphase filters POLY0, POLY1, POLY2 and POLY3 provide four different fractional sample delay amounts. It is appreciated that any number of polyphase filters may be employed depending upon the desired level of accuracy. The output of each of the polyphase filters 1007 provides shifted ES_D samples, which are provided to a square block 1009. The square block 1009 operates in a similar manner as the square block 611 for providing ES_D² values, which are provided to a correlator 1011. The correlator 1011 receives the A² values and outputs correlation results C. The correlation results C are monitored to find the best correlation results, which is usually in the form of the largest magnitude output.

[0070] Several configurations are contemplated for implementing the correlation circuit 1000 within the transmitter circuit for enabling more precise timing control. In one configuration for integrated circuits (ICs), the switch 1005 and the polyphase filters 1007 are inserted in the baseband processor 101 between the ADC 133 and the quadrature modulator calibration circuit 135. The switch 1005 is externally controllable by the test system to enable selection of the polyphase filters 1007. In one embodiment, the correlator 1011 is also incorporated on-chip, such as the at the outputs of the square block 611 and the amplitude square block 609, and the output of the correlator 1011 is externally available to monitor timing results. Alternatively, the outputs of the square block 611 the amplitude square block 609 are externally available and connected to an external correlator (not shown). At manufacture time, the transmitter chip is tested to adjust timing in which the delay block 401 and the switch 1005 are controlled and the correlation results monitored. During test, the polyphase filters 1007 are selected one at a time to process multiple samples. The polyphase filter that provides the best correlation results (e.g., largest correlation values) is selected for the part under test. Once determined, the delay block 401 and the switch 1005 are programmed to achieve the best timing alignment. Such programming may be achieved using an standard method, such as, for example, fuse blowing or EPROM programming or the like.

[0071] FIG. 11 is a block diagram of another exemplary correlation circuit 1100 that is used to achieve more accurate time alignment using a Farrow interpolator 1101. The correlation circuit 1100 is substantially similar to the correlation circuit 1000, except that the switch 1005 and the polyphase filters 1007 are replaced by the Farrow interpolator 1101. Similar components assume identical reference numbers. The Farrow interpolator 1101 receives a delay control signal DEL, which is varied to change the fractional sample delay amount through the Farrow interpolator 1101. Operation of the correlation circuit 1100 is similar to the correlation circuit 1000, in which the correlation results C output from the correlator 1011 are monitored while adjusting the DEL signal to determine optimal timing alignment. In this case, the DEL control signal is externally adjusted during test until optimal correlation results C are achieved, and the unit is programmed accordingly in a similar manner as previously described.

[0072] FIG. 12 is a block diagram of a transmitter 1200 including on-signal quadrature modulator calibration implemented according to an exemplary embodiment of the present invention illustrating closed loop operation. The transmitter 1200 includes a digital portion performed within a baseband processor (e.g., the baseband processor 210) and an analog portion performed within an analog quadrature modulator (e.g., the quadrature modulator 203). The lines of demarcation between the digital and analog portions includes the I DAC 101 and the Q DAC 109 in the I and Q signal paths previously described, and the ADC 133 in the feedback path. The analog portion is substantially the same and includes the LPFs 109 and 119, an RF modulator 1205, the summing junction 115 and the envelope detector 131 for sensing the OUT signal. The RF modulator 1205 includes, for example, the local oscillator 125, the oscillator splitter 127 and the mixers 113 and 123.

[0073] The digital portion includes an optional fractional delay device 1207 which delays the E_D samples and provides shifted ES_D samples. As previously described, the fractional delay function may be performed by filtering or interpolation or the like. The shifted ES_D samples are provided to a sample selector block 1209, which selects the delayed ES_D samples based on a gating function defined by GATEn signals. The GATEn signals represents a combination of the hit and bin signals H1-H4 and B1-B4 previously described. The digital portion includes a kernel or core 1211, which provides the I_D and Q_D digital values to a rate changer filter 1213 and to the delay block 401. The core 1211 is implemented according to OFDM or CCK or the like. The rate changer filter 1213 is used to increase the rate of the digital values to a higher rate to facilitate improved operation of the DACs 107, 117. For example, the rate changer filter 1213 may up-sample 20 MHz signals (OFDM) or 22 MHz signals (CCK) to a selected higher rate, such as, for example, 44 MHz or 80 MHz as known to those having ordinary skill in the art. The I and Q outputs of the rate changer filter 1213 are provided to the IQ correction circuit 105 previously described.

[0074] The delayed I_DD and Q_DD digital values from the delay block 401 are provided to a HIT/BIN detection block 1215, which generates the GATEn signals. In this case, the detection block 1215 includes the combined functions of the detection blocks 403, 503, 603 and 803 in which it asserts the GATEn signals indicative of a hit within any of the operative selection boxes and identification of the corresponding selection box for which each hit occurred. It is noted that in one embodiment, the selection boxes M0 and M180 are configured with the same size, shape and location as the selection boxes 605a and 605b and that the selection boxes M90 and M270 are configured with the same size, shape and location as the selection boxes 607a and 607b, so that only the selection boxes 605a,b and 607a,b are used for detecting and correcting both DC and amplitude imbalances. It is also noted that the four selection boxes 605a,b and 607a,b represents six different bin values, including one bin value for each selection box for DC calibration, and one bin for the combined selection boxes 605a and 605b and another bin for the combined selection boxes 607a and 607b for amplitude calibration. The detection block 1215 also detects and reports the gating functions for the selection boxes 805a,b and 807a,b for detecting and calibrating phase imbalances.

[0075] The sample selector block 1209 outputs to imbalance estimators 1217 selected ES_D samples several group sample group signals based on the GATE_n signals, where the sample group signals include S0, S45/225, S90, S135/315, S180 and S270. The first sample group S0 includes each ES_D sample that is a hit within the selection box 605b. The second sample group S45/225 includes each ES_D sample that is a hit within either of the selection boxes 805a and 805b. A third sample group S90 includes each ES_D sample that is a hit within the selection box 607a. The fourth sample group S135/315 includes each ES_D sample that is a hit within either of the selection boxes 8075a and 807b. The fifth sample group S180 includes each ES_D sample that is a hit within the selection box 605a. The sixth sample group S270 includes each ES_D sample that is a hit within the selection box 607b.

[0076] In the embodiment shown, the imbalance estimators 1217 incorporate the functionality similar to the calculator blocks 413, 513 for determining IDC_ERR and QDC_ERR values, which are provided to integrators 1219. In the configuration shown, the average blocks 409, 411, 509, and 511 are not included in the imbalance estimators 1217. Instead, the IDC_ERR value is calculated as half the difference between the sum of the samples of the sample groups S0 and S180. In a similar manner, the QDC_ERR value is calculated as half the difference between the sum of the samples of the sample groups S90 and S270. The IDC_ERR value is updated for each new sample from the sample groups S0 or S180, and the QDC_ERR value is updated for each new sample from the sample groups S90 or S270.

[0077] The I_DD and Q_DD digital values from the delay block 401 are provided to the imbalance estimators 1217. In one embodiment, the imbalance estimators 1217 incorporate the functionality of blocks 609, 611 and 613 for calculating the A² values, for squaring each of the sample values of the sample groups S0 and S180, and dividing the squared samples by A² to generated normalized power values P_I. The imbalance estimators 1217 also calculate a normalized power value P_Q in a similar manner, and the ratio of P_Q/P_I is determined to derive 1+2δ_ERR values in a similar manner as previously described. Alternatively, the magnitude values are calculated and the ratio of M_Q/M_I is determined to derive 1+δ_ERR values. In another alternative embodiment, the amplitude imbalance delta error terms δ_ERR are approximated as P_Q−P_I≈δ_ERR. The 1+δ_ERR values or, alternatively, the δ_ERR values are provided to the integrators 1219. In a similar manner, the imbalance estimators 1217 calculate normalized power values P₄₅ using the digital samples from the sample group S45/135 and normalized power values P₁₃₅ using the samples from the sample group S135/315. The imbalance estimators 1217 approximate a phase imbalance theta error term θ_ERR as (P₁₃₅−P₄₅)/(P₁₃₅+P₄₅)≈sin(θ_ERR)≈θ_ERR. In an alternative embodiment, the phase imbalance theta term θ_ERR is approximated as (P₁₃₅−P₄₅)≈θ_ERR.

[0078] The IDC_ERR, QDC_ERR, δ_ERR (or 1+δ_ERR) and θ_ERR error values are further processed by the integrators 1219 to generate corresponding estimate values IDC_EST, QDC_EST, δ_EST and θ_EST, respectively. The integrators 1219 apply filtering to convert the error values to the estimate values before being applied to the IQ correction circuit 105 to ensure proper loop operation. In the embodiment shown, for example, the IDC_EST value is determined by the integrators 1219 and provided to the IQ correction circuit 105, where IDC_EST(n+1)=IDC_EST(n)+K_I*IDC_ERR, in which “n” is an index value that is incremented from one sample to the next, IDC_EST(n+1) is the updated estimate value, IDC_EST(n) is the previous estimate value, K_I is a constant or programmable multiplier that is less than one that determines the bandwidth or speed of convergence of the IDC_EST values, IDC_ERR is the value provided from the imbalance estimators 1217, and the asterisk “*” denotes multiplication. A QDC_EST value is determined by the integrators 1219 and provided to the IQ correction circuit 105 in a similar manner, where QDC_EST(n+1)=QDC_EST(n)+K_Q*QDC_ERR, in which QDC_EST(n+1) is the updated estimate value, QDC_EST(n) is the previous estimate value, K_Q is a constant or programmable multiplier that is less than one that determines the bandwidth or speed of convergence of the QDC_EST values, and QDC_ERR is the value provided from the imbalance estimators 1217. An δ_EST value is determined by the integrators 1219 and provided to the IQ correction circuit 105, where δ_EST(n+1)=δ_EST(n)+K_δ*δ_ERR, in which δ_EST(n+1) is the updated estimate value, δ_EST(n) is the previous estimate value, K_δ is a constant or programmable multiplier that is less than one that determines the bandwidth or speed of convergence of the δ_EST values, and δ_ERR is the value provided from the imbalance estimator 1217. An θ_EST value is determined by the integrators 1219 and provided to the IQ correction circuit 105 in a similar manner, where θ_EST(n+1)=θ_EST(n)+K_θ*θ_ERR, in which θ_EST(n+1) is the updated estimate value, θ_EST(n) is the previous estimate value, K_θ is a constant or programmable multiplier that is less than one that determines the bandwidth or speed of convergence of the θ_EST values, and θ_ERR is the value provided from the imbalance estimators 1217.

[0079] The distortion estimate values IDC_EST, QDC_EST, δ_EST, and θ_EST values are programmed with initial or default values. In one embodiment, the distortion estimate values are initially set to zero. Alternatively, the distortion estimate values are each set to corresponding default values at the time of manufacture that are predetermined to effectively establish initial static compensation. The feedback error loops measure estimated distortions and modify the distortion estimate values accordingly, which dynamically converge to more accurate values to reduce or otherwise eliminate DC offsets and amplitude and phase imbalances.

[0080] FIG. 13 is a simplified flowchart diagram illustrating the order of calibration processing according to an exemplary embodiment of the present invention. The method illustrated may be employed by any or all of the embodiments described herein, including the transmitters 100, 300, and 1200. It is appreciated that the selection boxes used for DC calibration are similar to the selection boxes used for amplitude compensation. It is determined that the DC offsets may negatively impact determination of amplitude and phase compensation, so that it is desired to first adjust DC to within a predetermined target range before compensating for amplitude and phase.

[0081] At first block 1301, after power up and/or reset and during transmission, IDC and QDC calibration begins while amplitude and phase compensation is temporarily suspended. Operation continues for IDC and QDC calibration until the IDC_EST, QDC_EST values converge to within a predetermined target range, as indicated at next query block 1303. Convergence means that the I and Q DC offsets are being resolved and reduced by compensation to eliminate the DC offsets in the OUT signal. The target range is determined arbitrarily or experimentally and selected so that the DC offsets are reduced to a small enough level to enable relatively accurate amplitude and phase calibration to begin. Upon convergence to within the target range, operation proceeds to block 1305 in which the compensation loop is operated for amplitude and phase distortions as well as the I and Q DC offsets. In this manner, the I and Q DC offsets are first removed to within an acceptable range and then all of the compensation loops are operated together for compensating DC, amplitude and phase distortions.

[0082] Although the present invention has been described in considerable detail with reference to certain preferred versions thereof, other versions and variations are possible and contemplated. Those skilled in the art should appreciate that they can readily use the disclosed conception and specific embodiments as a basis for designing or modifying other structures for providing out the same purposes of the present invention without departing from the spirit and scope of the invention as defined by the appended claims.