GB2334187A

GB2334187A - Quadrature modulator with automatic compensation

Info

Publication number: GB2334187A
Application number: GB9802368A
Authority: GB
Inventors: Rene Pihl Jepsen; Paul Anthony Howard
Original assignee: Motorola Ltd
Current assignee: Motorola Solutions UK Ltd
Priority date: 1998-02-04
Filing date: 1998-02-04
Publication date: 1999-08-11
Also published as: GB9802368D0

Abstract

Automatic compensation is provided for DC offset or gain assymetry of the modulator. The modulator comprises an amplitude envelope detector (205) connected to the modulator output. The amplitude signal at the output of the envelope detector (205) is fed to a sampler (207) and the samples are subsequently processed in an error generator (209) deriving an error signal. The error signal is filtered in a filter (211) before being applied to a compensator (201, 203) in each quadrature channel which modifies a characteristic of the modulator in response to the error signal. Typically an error signal is obtained by dividing the samples into two or more categories according to the corresponding data values and generating a difference between the samples in the two categories. Sampling of the amplitude signal is determined by a controller (213) in order to sample in a fixed relation to the timing of the baseband I- and Q-channels.

Description

MODULATOR WITh AUTOMATIC COMPENSATION AND METHOD mEREF)R Field of the Invention This invention relates to a modulator for modulating a data signal to be transmitted on typically a radio channel.

Bt to the Invention Many techniques for modulating a data signal to be transmitted over a channel as for example a radio channel is known. One technique uses quadrature modulation by pulse shaping the data at baseband on individual quadrature channels before mixing the channels with carriers 90" out of phase and adding the two quadrature modulated signals. This technique provides advantages in lower power consumption and complexity in comparison to for example a technique using direct synthesis at higher frequencies.

However, a major problem with baseband pulse shaping is that the DC offset and the asymmetry between the quadrature channels must be kept very low for sufficient performance. A DC offset on any of the quadrature channels will cause carrier leakage and amplitude variations of the modulated signal. Any component variations in the modulator can thus cause a significant degradation of the modulator. Conventionally, the problem of DC offset has been addressed by AC coupling the modulator but this has significant disadvantages as it can only be used for data signals with no DC bias and furthermore does not remove any DC offsets in the upconversion oscillators.

Another method for reducing the degradation caused by component variations has been to individually calibrate each modulator. This is a very expensive and time consuming activity. Furthermore, a calibration performed at the factory will not compensate for component variations due to temperature variations or ageing of the components and an additional degradation must be accepted, or a frequent and very costly recalibration must be carried out.

A modulator with automatic calibration which can compensate for component variations due to drift will thus be very advantageous.

Summary of the Invention The invention seeks to provide a modulator with automatic compensation for variations in a characteristic of the modulator, such as the DC offset or gain asymmetry, and a method therefor.

According to the invention, there is provided a modulator with automatic compensation for generating a modulated signal from a baseband signal generated by pulse shaping in a first and second quadrature channel of said modulator in response to a data signal, comprising: an amplitude envelope detector generating an amplitude signal from said modulated signal, a sampler for sampling said amplitude signal thereby generating a plurality of samples, a controller for said sampler determining sampling instants in relation to a symbol timing of said baseband signal, an error generator for generating an error signal from said plurality of samples, a filter generating a filtered error signal from said signal, and a compensator adjusting a first characteristic of said modulator in response to said filtered error signal.

Specifically the characteristic of the modulator can be a DC offset or a gain asymmetry between the quadrature channels. The sample instants can be determined by zero crossings of the baseband signals in the individual quadrature channels and the error signal can be generated as a function of the sample values corresponding to these zero crossings. The invention is applicable to many modulation formats including Minimum Shift Keying (MSK) and Quadrature Amplitude Modulation (QAM).

According to a different aspect of the invention, there is provided a method for compensation of a modulator for generating a modulated signal from a baseband signal generated by pulse shaping in a first and second quadrature channel of said modulator in response to a data signal, comprising the steps of: generating an amplitude envelope signal from said modulated signal, sampling said amplitude envelope signal thereby generating a plurality of samples, determining sampling instants in relation to a symbol timing of said baseband signal, generating an error signal from said plurality of samples, generating a filtered error signal from said signal, and adjusting a first characteristic of said modulator in response to said filtered error signal.

Brief Description of the Drawings An embodiment of the present invention is described below, by way of example only, with reference to the accompanying drawings, in which: FIG. 1 is an illustration in block diagram form of a quadrature modulator according to prior art.

FIG. 2 is an illustration in block diagram form of a modulator in accordance with the present invention.

FIG. 3 illustrates the result of DC offset on a GMSK signal in the I-Q plane.

FIG. 4 illustrates preferred measurement points for a GMSK signal in the I-Q plane.

FIG. 5 illustrates an example of measurement points for a GMSK signal in the presence of a DC offset.

FIG. 6 illustrates an example of DC offset compensation according to the invention.

FIG. 7 illustrates the result of gain asymmetry on a GMSK signal in the I Q plane.

FIG. 8 illustrates an example of gain asymmetry compensation according to the invention.

FIG. 9 illustrates some preferred measurement points for a 16 QAM signal.

Detailed Description of a PreiZened Embodiment FIG. 1 illustrates in block diagram form a quadrature modulator 100 with baseband pulse shaping according to known techniques. A data signal is fed to a data controller 101 controlling two pulse shapers 103, 105 in two quadrature channels (in the following denoted the I-channel and the Qchannel respectively). In the shown example, the data controller 101 and pulse shapers 103, 105 are implemented digitally and the rest of the modulator is implemented using analog techniques. The pulse shapers 103,105 are therefore followed by Digital to Analog (D/A) converters 107,109.

The analog signals in the I- and Q- channel are fed to quadrature mixers 111,113 before being added together in an adder 115. A local oscillator 117 generates an upconversion signal fed to the I-channel mixer and through a 90C phase shifter 119 fed to the Q-channel mixer. The adder 115 is followed by a transmitter chain 121 upconverting, amplifying and filtering the signal before it is fed to the antenna 123.

The operation of the data controller 101 and the pulse shapers 103, 105 depend on the nature of the modulation. For a Quadrature Phase Shift Keying (QPSK) signal the data controller simply splits the data signal into two data streams which are fed to two pulse shapers. The pulse shapers are implemented as filters with an impulse response equal to the pulse shape of each data symbol. For many Continuous Phase Frequency Shift Keying (CPFSK) signals the signal in the I- and Q- channel depend on the data in the other channel and thus on more than one data symbol. In this case, the data controller 101 and pulse shapers 103,105 are typically implemented as a shift register storing the necessary data symbols, followed by a look-up table, typically implemented in Read Only Memory (ROM), generating the corresponding signal values for the I- and Qchannels.

A significant problem in baseband quadrature modulators is the degradation caused especially by the DC offset in the individual I- and Qchannel, and the gain asymmetry between the channels. These characteristics are typically minimised by calibration. However, due to drift of the components caused by ageing and temperature variations, a degradation will occur over time, or frequent recalibration of the modulator must be carried out. The calibration and recalibration are very time consuming and expensive.

FIG. 2 illustrates in block diagram form an example of an embodiment of a modulator according to the present invention. The modulator of FIG. 2 corresponds to the modulator of FIG. 1 with additional circuitry for compensation of variations in a characteristic of the modulator, such as for example DC offset or gain asymmetry. The output of the transmitter chain 121 is fed to an amplitude envelope detector 205 which removes the high frequency carrier leaving the amplitude variations at the unmodulated signal frequency (i.e. the frequency range corresponding to the baseband signal in the I- and Q channel). The amplitude signal at the output of the envelope detector 205 is fed to a sampler 207 and the samples are subsequently processed in an error generator 209 deriving an error signal.

The error signal is filtered in a filter 211 before being applied to a compensator in each quadrature channel which modifies a characteristic of the modulator in response to the error signal.

Typically, an error signal is obtained by dividing the samples into two or more categories according to the corresponding data value and generating a difference between the samples in the categories. The compensator will depend on the characteristic which is being compensated. For example, an error signal may be derived for the DC offset in one of the quadrature channels by dividing samples into two categories depending on whether they are expected to have a positive value in the given quadrature channel or not, and determining the difference between these categories. The compensator will then increase or decrease a DC bias of the channel in response to this error signal. The sampling of the amplitude signal is determined by a controller 213 in order to sample in a fixed relation to the timing of the baseband I- and Q-channels, for example by instigating a sample whenever a symbol in the I- or Q channel reaches maximum amplitude or performs a zero-crossing. FIG. 2 shows the controller to derive the sampling instants in response to the data signal prior to the data controller but it can equally well derive the instants in response to the baseband signals anywhere in the quadrature channels.

FIG. 2 shows the compensators to precede the D/A converters but they can be elsewhere in the quadrature channels depending on the specific embodiment. Specifically they can be implemented as analog compensators following the D/A converters. Depending on the embodiment the feedback loop consisting of the amplitude envelope detector 205, the sampler 207, the error generator 209, the filter 211 and the compensators 201,203 can be implemented by analog means but will preferably include an Analog to Digital (A/D) converter, preferably in connection with the sampler 207. The example also shows the compensation to be performed in both quadrature channels but it may be limited to only one of the quadrature signals. This will specifically be applicable to a modulator using only one of the quadrature channels, such as a Binary Phase Shift Keying (BPSK) modulator, to which this invention is also applicable. The amplitude envelope detector will preferably be fed an RF (Radio Frequency) amplitude signal from the output of the transmit chain but may alternatively be fed a signal from an intermediate stage or from the input of the transmit chain.

The filtering performed in the feedback loop can alternatively be performed prior to generating the error signal for example by individually filtering the samples of each category of samples. In the extreme situation where the stability and noise level of the feed back loop permits the filter may simply consist in a delay element.

In the following, a specific embodiment is described for a Gaussian Minimum Shift Keying (GMSK) modulator, which is the modulation format used in the Global System for Mobile telecommunication (GSM).

Initially an embodiment for compensating a DC offset characteristic of a modulator be described followed by a description of compensation for gain asymmetry. These two characteristics are typically the largest sources of degradation from a quadrature modulator and significantly improved performance can be achieved by removing these error sources.

FIG. 3 illustrates the effect of a DC offset on a GMSK signal. A GMSK signal is a constant amplitude signal and ideally corresponds to a circle 301 centered on point (0,0) in the I-Q plane 300. However, a DC offset in each of the quadrature channels will move the circle from being centered on point (0,0) as illustrated in FIG. 3 where a DC offset of 0.1 in the I channel and -0.2 in the Q channel causes the GMSK signal to span a circle 303 centered on (0.1,-0.2). This will result in amplitude variations and carrier leakage of the modulated signal.

According to the invention, the modulated signal is sampled in a specific relation to the symbol timing of the baseband signals. For the GMSK case the sampling is preferably performed when the signal aligns completely with either the I- or Q-channel as shown in FIG. 4, where four measurement points 401,403, 405, 407 corresponding to the four possible alignments with a quadrature channel are indicated. For a normal Minimum Shift Keying (MSK) signal this timing corresponds to the instant when one symbol ends and the next begins. An error signal is generated from the synchronised samples preferably by dividing the samples into categories according to the data value of the baseband signal. For a DC offset, the samples can be divided into four categories according to the data values corresponding to which of the measurement points they correspond to. For each of the quadrature channels the samples are thus divided into two categories of samples corresponding to a negative baseband amplitude and a positive baseband amplitude value. The four categories would correspond to samples taken at measurement point 401, 403, 405 and 407 respectively. The samples in each category are preferably filtered individually for example by generating an average value over a past number of samples.

FIG. 5 illustrates an example of a GMSK signal 509 in an I-Q plane 500 in the presence of a DC offset. Due to the DC offset the envelope amplitude will be different for the four measurement points and thus of the filtered values of the four categories of samples. The illustration shows an example where samples at 501 has an amplitude of 0.8, samples at 503 an amplitude of 1.1, samples at 505 an amplitude of 1.2 and samples at 507 an amplitude of 0.9.

According to the invention an error signal is generated from these four pluralities of samples. As the DC offset is individual in the I- and Qchannel and must be individually compensated in the two channels, two error signals are generated. The error signal for the I-channel is generated as the difference between the two pluralities of samples corresponding to alignment with the I-channel, i.e. measurement points 503 and 507. Letting xn denote samples taken at measurement point n the error signals are thus given as e1 = x,, - x507 = 1.1 - 0.9 = 0.2 = = x50l - x505 = 0.8 - 1.2 = -0.4 The error signals thus independently derives values for the I and Q channel offset. The derived error signals can be filtered before being applied to the compensators, or the filtering of the samples prior to the error signal generator may be sufficient and the error signals can be fed directly to the compensators. This technique thus enables independent error signals for DC offset in each of the two quadrature channels to be derived from the combined modulated signal at the output of the transmitter.

The compensators will add a DC bias into each of the quadrature channels in response to the error signal. For the given example, the compensators can simply increase or decrease a DC bias in response to the error signal.

This corresponds to including an integrator in the feedback loop as is well known from regulation loops. The loop must be ensured to be stable and have the desired dynamic characteristics as is well known in the art.

FIG. 6 illustrates in block diagram form a specific embodiment as just described. For brevity only the essential elements needed for the description are included, and the data controller, pulse shapers, phase shifter and local oscillator are therefore not shown. The pulse shaped signals in the Iand Q-channel are converted from analog to digital form by DIA converters 601, 603 before being quadrature modulated in two mixers 605, 607 and slimmed together in an adder 609. In this example, the amplitude envelope detector 611 precedes the transmit chain and follows the adder 609 directly.

A sampler 613 samples the output of the amplitude envelope detector 611, converts the analog samples into digital and divides the samples into four categories according to the data value of the baseband signal. The samples corresponding to the I-channel are fed to one error generator 615 and the samples corresponding to the Q-channel are fed to another error generator. The two error signals are fed to two integrators 619, 621 and the integrated error signals are fed to two adders 623, 625 introducing a DC bias into each of the quadrature channels compensating for the DC offset of the channels.

The exact timing of the samples depend on the specific embodiment and the modulation format. For example, a GMSK signal as used in GSM will only have a zero crossing when the previous data symbol and the following data symbol are different. When these symbols are identical the signal phase will not reach closer than 30 degrees from a zero crossing. In this case, the sampling can be performed regardless of the data values and the deviation from a zero crossing can either be compensated for in the calculation of an error signal, or it can automatically cancel out due to the averaging of samples performed in the loop. Alternatively, sampling can be performed only when a zero crossing does occur corresponding to the preceding and following symbol being different. It is also possible to simply discard samples not corresponding to zero crossings when generating the error signal. As the loop typically will be much slower than the data rate this is likely to be fully sufficient. The sampling and the division of samples into different categories may thus depend on the data values of more than one data symbol and may specifically depend on the data transitions as well as the data pattern.

The control of the sampling instants can simply be obtained by having a zero-crossing detector in each of the quadrature channels generating a sampling strobe which is fed to the sampler. Alternatively the modulator can directly generate sampling strobes in reference to the internal timing, for example by generating a sampling strobe when a new data value is fed to the pulse shapers or with a constant delay relative to this. In this way, sampling strobes can be generated corresponding to any relation to the timing of the baseband signal and is not limited to zero-crossings.

The invention is also applicable to compensating a gain asymmetry characteristic of the modulator. As illustrated in FIG. 7, a gain asymmetry between the quadrature channels will for a GMSK signal cause the circle spanned by the signal in the I- and Q-plane 700 to transform into an ellipse 709. In the example illustrated, the gain in the Q-channel is significantly larger than the gain in the I-channel resulting in an very elongated ellipse. In agreement with the previously described example for DC offset compensation, four measurement points are generated from the zerocrossings. In this case the samples are divided into two categories depending on whether the samples correspond to being aligned with the I or Q channel. An error signal is generated as the difference between these two categories: e = x701705 - x703,707 = 1.2 - 0.8 = 0.4 The positive sign of this error signal indicates that the gain of the Qchannel is higher than the gain of the I channel and the magnitude is an indication of how much higher. The error signal can thus be used for controlling the gain of preferably one of the quadrature channels thereby compensating any imbalance between the channels.

FIG. 8 illustrates a specific embodiment for gain asymmetry compensation. Only the essential elements for describing the difference to the example of DC offset are included. The pulse shaped signals in the Iand Q-channel are converted from analog to digital form by D/A converters 801, 803 before being quadrature modulated in two mixers 805, 807 and sllmmed together in an adder 809. In this example, the amplitude envelope detector precedes the transmit chain and follows the adder 809 directly. A sampler 813 samples the output of the amplitude envelope detector 811, converts the analog samples into digital and divides the samples into two categories according to the data value of the baseband signal i.e. to which quadrature channel the current sample aligns. An error generator 815 generates an error signal as the difference between the samples corresponding to the I-channel and the samples corresponding to the Qchannel. The error signal is fed to an integrator 817 and the integrated error signal controls the gain of an amplifier 819 in one of the quadrature channels thereby compensating for a gain asymmetry.

It will be obvious to the skilled person that the comments on possible embodiments of the modulator with DC offset compensation are equally applicable to the modulator with gain asymmetry compensation.

It will be apparent, that the invention is applicable to other modulation formats than GMSK. One other possibility is Quadrature Amplitude Modulated (QAM) signals as for example Quadrature Phase Shift Keying (QPSK). Again the samples can be derived from zero crossings but due to the non-constant amplitude of the signal a compensation dependent on the data values of the baseband signal can be introduced. As an example, FIG.

9 illustrates a 16 QAM signal. A data transition from symbol 901 to symbol 903 will result in a zero crossing measurement point of 905 whereas data transition from symbol 907 to symbol 909 will result in a zero crossing measurement point of 911. The difference between the measurement point amplitude values can be compensated for example by deriving an amplitude value relative to an expected value or alternatively the amplitude signal can be scaled according to the difference between the measurement points. Depending on the characteristic compensated and the modulation format used, the filtering and the averaging performed in the feedback loop can be sufficient to generate a sufficient error signal regardless of these variations. The technique of compensating the measurements by comparing to expected values for the present data values is not limited to zero-crossing measurements. They are for example also applicable to measurements performed at sampling instants corresponding to the constellation points. Similarly the comparison to expected values derived from the baseband data signal can be applied to the individual samples or to the filtered and averaged values.

The implementation of the individual components in hardware is well known in the art and no detailed description is necessary. Preferably, the compensators, error generator, filters and the controller for the sampler are implemented digitally as a software program performing the algorithms described on a suitable processor such as a micro processor or a digital signal processor. The amplitude envelope detector is preferably implemented as a low pass filter with a cut off frequency significantly higher than the maximum frequency of the baseband signal and much lower than the modulation frequency of the local oscillator.

The invention thus provides a modulator with and method for automatic compensation for variations in a characteristic of the modulator.

Specifically according to the invention, simultaneous and independent compensation of DC offset in each of the quadrature channels and gain asymmetry between the channels can be achieved based on the modulated output signal. The invention thus enables a modulator with automatic compensation of DC offset and gain asymmetry while requiring no calibration.

Claims

Claims 1. A modulator with automatic compensation for generating a modulated signal from a baseband signal generated by pulse shaping in a first and second quadrature channel of said modulator in response to a data signal, comprising: an amplitude envelope detector generating an amplitude signal from said modulated signal, a sampler for sampling said amplitude signal thereby generating a first plurality of samples, a controller for said sampler determining sampling instants in relation to a symbol timing of said baseband signal, an error generator for generating an error signal from said plurality of samples, a filter generating a filtered error signal from said error signal, and a compensator adjusting a first characteristic of said modulator in response to said filtered error signal.
2. A modulator as claimed in claim 1 wherein said first plurality of samples are divided into at least a second and third plurality of samples and said error signal is generated as a first difference between said first and third plurality of samples.
3. A modulator as claimed in claim 2 wherein said second and third plurality of samples are filtered before said error signal is generated.
4. A modulator as claimed in claim 2 wherein said first plurality of samples are divided in response to corresponding values of said data signal.
5. A modulator as claimed in claim 2 or 4 wherein said error signal is generated in comparison to an expected value of said second and third plurality of samples.
6. A modulator as claimed in claim 2 or 4 wherein each sample of said first plurality of samples is compensated by comparison to an expected value of said each sample.
7. A modulator as claimed in any of the claims 2 to 6 wherein said first characteristic is a DC offset.
8. A modulator as claimed in claim 7 wherein said DC offset is in said first quadrature channel and said second plurality of samples comprises samples with a negative baseband amplitude value in said first quadrature channel and said third plurality of samples comprises samples with a positive baseband amplitude value in said first quadrature channel.
9. A modulator as claimed in claim 8 wherein said controller is operable to detect a zero crossing of said baseband signal in said second quadrature channel and to instigate a sampling when this occurs.
10. A modulator as claimed in claim 9 wherein sampling is only performed when said controller detects a zero crossing of said baseband signal in said second quadrature channel.
11. A modulator as claimed in claim 8 or 9 wherein said modulated signal is a Quadrature Amplitude Modulated (QAM) signal.
12. A modulator as claimed in claim 11 wherein each of said first plurality of samples are scaled according to an amplitude level of said baseband signal corresponding to each of said plurality of samples.
13. A modulator as claimed in claim 8 or 9 wherein said modulated signal is a Gaussian Minimum Shift Keying (GMSK) modulated signal.
14. A modulator as claimed in any of the claims 1 to 6 wherein said first characteristic is a gain asymmetry between said first quadrature channel and said second quadrature channel,
15. A modulator as claimed in claim 14 wherein said controller is operable to detect a zero crossing of said baseband signal in both said first and second quadrature channel and to instigate a sampling when this occurs.
16. A modulator as claimed in claim 15 wherein said second plurality of samples comprises samples corresponding to a zero crossing in said first quadrature channel and said third plurality of samples comprises samples corresponding to a zero crossing in said second quadrature channel.
17. A modulator as claimed in claim 16 wherein said gain asymmetry is adjusted by adjusting a gain of said first quadrature channel in response to said filtered error signal.
18. A modulator as claimed in claim 15, 16 or 17 wherein said modulated signal is a Gaussian Minimum Shift Keying (GMSK) modulated signal.
19. A modulator as claimed in claim 15 wherein said modulated signal is a Quadrature Amplitude Modulated (QAM) signal.
20. A modulator as claimed in claim 19 wherein each of said first plurality of samples are scaled according to an amplitude level of said baseband signal corresponding to each of said first plurality of samples.
21. A method for compensation of a modulator for generating a modulated signal from a baseband signal generated by pulse shaping in a first and second quadrature channel of said modulator in response to a data signal, comprising the steps of: generating an amplitude envelope signal from said modulated signal, sampling said amplitude envelope signal thereby generating a first plurality of samples, determining sampling instants in relation to a symbol timing of said baseband signal, generating an error signal from said plurality of samples, generating a filtered error signal from said signal, and adjusting a first characteristic of said modulator in response to said filtered error signal.
22. A method as claimed in claim 21 wherein said first plurality of samples are divided in response to corresponding values of said data signal.
23. A method as claimed in claim 22 wherein said first characteristic is a DC offset.
24. A method as claimed in claim 23 wherein said DC offset is in said first quadrature channel and said first plurality of samples is divided into a second plurality of samples comprising samples with a negative baseband amplitude value in said first quadrature channel and a third plurality of samples comprising samples with a positive baseband amplitude value in said first quadrature channel.
25. A method as claimed in claim 24 wherein said sampling instants are determined in response to a zero crossing of said baseband signal in said second quadrature channel.
26. A method as claimed in claim 25 wherein said modulated signal is a Gaussian Minimum Shift Keying (GMSK) modulated signal.
27. A method as claimed in claim 22 wherein said first characteristic is a gain asymmetry between said first quadrature channel and said second quadrature channel.
28. A method as claimed in claim 27 wherein said sampling instants are determined in response to a zero crossing of said baseband signal in said first or second quadrature channel.
29. A method as claimed in claim 28 wherein said modulated signal is a Gaussian Minimum Shift Keying (GMSK) modulated signal.
30. A modulator substantially as hereinbefore described with reference to the accompanying drawings.
31. A method for compensation of a modulator substantially as hereinbefore described with reference to the accompanying drawings.