WO2004110008A1

WO2004110008A1 - Digital time alignment in a polar modulator

Info

Publication number: WO2004110008A1
Application number: PCT/US2004/014218
Authority: WO
Inventors: Thomas E. Biedka; Wayne S. Lee; Gary L. Do
Original assignee: Tropian, Inc.
Priority date: 2003-06-04
Filing date: 2004-05-07
Publication date: 2004-12-16
Also published as: US20060029153A1; US20040247041A1; EP1629650A1; JP2006526948A; KR20060059885A; CN1799237A; KR101006298B1; EP1629650A4; US7042958B2; CN1799237B

Abstract

Methods of and apparatus for digitally controlling, with sub-sample resolution, the relative timing of the magnitude (101a) and phase (101b) paths in a polar modulator. The timing resolution is limited by the dynamic range of the system as opposed to the sample rate. The methods and apparatus of the invention use a digital filter (111) to approximate a sub-sample time delay. Various techniques for approximating a sub-sample time delay using digital signal processing may be used to achieve the approximation. Ideally, the filter will have an all-pass magnitude response and a linear phase response. In practice, the magnitude may be a low-pass and the phase may not be perfectly linear. Such deviation from the ideal response will introduce some distortion. However, this distortion may be acceptably small depending on the particular signal being processed.

Description

DIGITAL TIME ALIGNMENT IN A POLAR MODULATOR

FIELD OF THE INVENTION [0001] The present invention relates generally to power modulators. More specifically, the present invention relates to time alignment between the amplitude and phase paths of a polar modulator.

BACKGROUND OF THE INVENTION [0002] In a polar modulator, the amplitude modulation and phase modulation are processed separately before being combined to create the desired signal. The processing of the magnitude and phase modulations are different, which can introduce a relative time offset between the magnitude and phase signal components. If this relative time offset is not corrected, the desired signal will not be accurately generated, and distortion will result. Typically this distortion will result in an increase in out-of-band signal energy.

[0003] One approach to solving this problem is to process the signal with a sample rate that is fast enough so that the required time offset correction can be achieved with a simple digital register. This may be a reasonable approach if the signal bandwidth is low. However, for high bandwidth signals, such as CDMA and WCDMA, the relative magnitude/phase timing may need to be accurate to within one nanosecond or less. Such timing resolution would require a sample rate on the order of 1 GHz, which is not practical. [0004] Another approach is to use analog delay elements. These delay elements would be placed in the magnitude and/or phase paths after D/A conversion. Unfortunately, this approach has several drawbacks, including: (1) the delay may vary with time, with temperature, and from part to part; (2) the delay provided by the analog delay elements may need to be calibrated, which is a drawback in a low-cost manufacturing environment; and (3) the analog delay elements may be difficult to integrate into an integrated circuit, which increases the number of required components.

SUMMARY OF THE INVENTION [0005] Methods of and apparatus for digitally controlling, with sub-sample resolution, the relative timing of the magnitude and phase paths in a polar modulator are disclosed. Being completely digital, there is no manipulation necessary to timing of system signals. The timing resolution is limited by the dynamic range of the system as opposed to the sample rate. A digital filter is employed to approximate a sub-sample time delay. Various techniques for approximating a sub-sample time delay using digital signal processing may be used for this purpose. The filter will ideally have an all-pass magnitude response and a linear phase response. In practice, the magnitude may be low-pass and the phase may not be perfectly linear. Such deviation from the ideal response will introduce some distortion. However, this distortion may be acceptably small depending on the particular signal being processed. Thus, fine-resolution time alignment among signals can be achieved without any explicit delay manipulation of clock or data signals. Lower development and production costs are an immediate benefit. [0006] Further aspects of the invention are described and claimed below, and a further understanding of the nature and advantages of the inventions may be realized by reference to the remaining portions of the specification and the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a portion of a polar modulator;

FIG. 2 is a plot showing, in relation to an EDGE transmit mask power spectral density (PSD) of various signals including a signal generated from time-misaligned magnitude and phase components, a signal generated from magnitude and phase components time aligned using a linear digital filter, and a signal generated from magnitude and phase components time aligned using an all-pass filter;

FIG. 3 is a plot corresponding to that of FIG. 1 showing power at 400 kHz offset versus magnitude-phase time offset for the various signals; and

FIG. 4 is a plot comparing power at 400 kHz offset versus delay for linear and all-pass filters, respectively.

DETAILED DESCRIPTION OF INVENTION [0007] Referring now to FIG. 1 , there is shown of a portion of a polar modulator in accordance with an embodiment of the present invention. A signal to be transmitted is

represented in polar form as magnitude information p and phase information θ. Optionally, one

of the magnitude path and the phase path may be delayed relative to the other by one or more sample periods in delay blocks 101a, 101b to achieve coarse time alignment.

[0008] Considering first the phase path, phase modulation is performed by a digital phase modulator 103. Digital to analog conversion (105) then follows, and the resulting analog signal is applied to a voltage controlled oscillator (VCO) 107. An output signal of the VCO is applied to an RF input of an RF power amplifier 109. Preferably, the RF amplifier is operated in compression, or switch mode.

[0009] Considering next the magnitude path, the magnitude information is applied to a delay filter 111 for fine, sub-sample-time delay adjustment, as described in greater detail hereinafter. Digital to analog conversion (110) then follows, and the resulting signal is applied to an amplitude modulator, or power driver 113. The amplitude modulator receives as an input a supply voltage Vbatt and produces a voltage Vpa that is applied to a supply terminal of the power amplifier 109. Responsive to the output signal of the VCO and to the voltage Vpa, the power amplifier produces the desired output signal, which is applied to an antenna 115, for example, for transmission. It should be mentioned here that, although the delay filter is shown in FIG. 1 as being in the magnitude path, it could equally well be located in the phase path.

[0010] As illustrated in the foregoing FIG. 1, the relative timing of the magnitude and phase can be controlled by delaying one path relative to the other. Further details of the delay filter 111 will now be considered. Without loss of generality, consider delaying the magnitude p

by τ, where τ is some fraction of a sample interval. Stated mathematically, the magnitude p is known only at the discrete time instants kT, where k is an integer and T is the sampling period. An accurate estimate is to be found of the magnitude at some time instant kT + AT where 0 < A <

1. For the remainder of this disclosure, the magnitude p at discrete time instant kT will be

denoted by p(k).

[0011] One approach is to use linear interpolation. Linear interpolation can be viewed as a non-recursive filtering operation with two filter coefficients. The filter output will provide an estimate of the magnitude at time instant kT + AT. If the filter input is denoted by x and the filter output by y, linear interpolation is defined as

y(k) = x(k) + A[x(k + l) - x(k)]

[0012] Linear interpolation has been described here so that only one multiplication is needed, which is important in a practical implementation. Note that the resolution of the delay is only limited by the number of bits used to represent the delay.

[0013] Another approach is to use a first-order all-pass digital filter. This filter is of the infinite-impulse-response (ILR) type, and is defined by the recursion

y(k) = c[x(k)- y(k-l)]+x(k-l) where

1 - Δ c = ^■

1 + Δ

[0014] Still another approach is to use a finite-impulse-response (FLR) filter designed to have nearly linear phase and a nearly flat passband. Such a filter may be designed, for example, according to a minimum mean-square error (MMSE) criterion.

[0015] The effectiveness of digital sub-sample delay compensation will now be demonstrated through simulation. A simulation procedure was used to generate a PAM signal with 16 samples per symbol, calculate the magnitude and phase, time shift the magnitude by one sample, and decimate the magnitude and phase by two. The magnitude and phase then have 8 samples per symbol, and the relative magnitude-phase timing offset is exactly 1/2 sample. Without loss of generality, an EDGE signal was for this demonstration.

[0016] The effect of the 1/2 sample timing offset on the PSD of the reconstructed EDGE signal is shown in FIG.2, where it is seen that the magnitude-phase time offset causes the signal to violate the EDGE spectral mask (from 3GPP TS 45.005 v5.1.0 for DCS 1800 mobile station with output power < 30 dBm). If the magnitude-phase time offset is corrected with linear interpolation, it can be seen that the corrected signal meets the spectral mask. However the margin to the mask at 400 kHz offset is only about 3 dB. If a first-order all-pass filter is used to compensate for the time-offset, the resulting margin to the mask at 400 kHz offset is nearly 17 dB. This shows that the all-pass filter is better able to compensate for the magnitude-phase time offset. The relative performance of linear interpolation and all-pass filtering will depend on the sample rate. If the signal is processed at a sufficiently high sample rate, there will be very little difference between linear interpolation and all-pass filtering.

[0017] To determine the ability of linear interpolation and all-pass filtering to correct a range of time-offset values, the following simulation was performed. An EDGE signal was generated with 64 samples per symbol, and the following steps were performed: calculate the magnitude and phase, time shift the magnitude by zero to eight samples, then decimate the magnitude and phase by 8. The magnitude and phase then have 8 samples per symbol, and the relative magnitude-phase timing offset varies from 0 to 1 sample in 1/8 sample steps. FIG. 3 shows the relative power of the PSD at a 400 kHz offset for the uncorrected signal and corrected signals as a function of the time offset. It is clear that in all cases a time offset of 1/2 sample is the worst case. The all-pass filter is able to compensate for a broad range of time offsets with little variation in performance.

[0018] A final issue is the required resolution of the delay parameter for linear interpolation and all-pass filtering. To study this the following simulation was performed. An EDGE signal was generated with 8 samples per symbol, and 1/2 sample time-offset between magnitude and phase, and the time delay parameter of a linear interpolator and an all-pass filter was varied. The results of this experiment are shown in FIG. 4. At a delay value of 0.5 samples, nominal time-alignment is achieved. For the linear filter, power at a 400 kHz offset is more than 55 dB down. For the all-pass filter, power is more than 70 dB down. [0019] In practice, the parameters of the delay filter may be calibrated at design time and set so as to achieve the desired PSD characteristics. For different designs, the parameters of the delay filter may be different.

[0020] It will be appreciated from the foregoing description that both linear interpolation and a first-order all-pass filter can provide an effective means for controlling magnitude-phase timing to sub-sample accuracy. The all-pass filter provides superior performance at the cost of increased implementation complexity. Further improvements in performance can be achieved by using an FUR. filter with more than two taps to approximate the desired delay. Such a filter could be designed according to an MMSE criterion. All of these simulations presented have been fully validated by hardware implementation. This proves that sub-sample-time signal delay shifting can be fully implemented using strictly digital techniques, eliminating any need for delay manipulation of data or clock signals in the design. Or, equivalently, eliminating the tradeoff of being forced to accept time alignment resolution set by system clock frequencies (sample times).

[0021] Whereas the above is a complete description of the preferred embodiments of the invention, various alternatives, modifications, and equivalents may be used. Therefore, the above description should not be taken as limiting the scope of the invention as it is defined by the appended claims.

Claims

CLAIMSWhat is claimed is:

1. A method of achieving time alignment between an amplitude path signal and a phase path signal in a modulator having separate amplitude and phase paths, comprising: providing a digital circuit within at least one of the amplitude and phase paths; and adjusting at least one parameter of the digital circuit to achieve a relative delay between the amplitude and phase path signals that corresponds to a substantial time alignment of the amplitude and phase path signals.

2. The method of Claim 1 wherein the digital circuit comprises a linear filter, an IIR filter, an FIR filter, or an all-pass filter.

3. The method of Claim 1 wherein the digital filter comprises an LTR filter having more than two coefficients.

4. The method of Claim 1 wherein said relative delay is less than or equal to one sample interval associated with said modulator.

5. The method of Claim 1 wherein said digital circuit comprises an FIR filter having two or more coefficients.

6. The method of Claim 1 wherein no physical time delays are performed on either a system clock controlling the modulator or on data signals provided by either of said amplitude and phase paths.

7. A method of achieving time alignment between an amplitude path signal and a phase path signal in a modulator having separate amplitude and phase paths, comprising: providing a digital circuit within at least one of the amplitude and phase paths; and adjusting at least one parameter of said digital circuit to achieve a time alignment of the amplitude path and phase path signals that is less than or equal to a sample interval associated with said modulator.

8. The method of Claim 7 wherein said digital circuit comprises a digital filter.

9. The method of Claim 8 wherein said digital filter comprises a linear filter, an IIR filter, an FIR filter, or an all-pass filter.

10. The method of Claim 8 wherein said digital filter comprises an IIR filter having two or more coefficients.

11. The method of Claim 8 wherein said digital filter comprises an FLR filter having two or more coefficients.

12. A transmitter, comprising: separate amplitude and phase paths; an amplifier responsive to separate amplitude and phase signals for producing a communications signal; and a digital circuit within at least one of the amplitude and phase paths, wherein adjusting a parameter of the digital circuit achieves a relative delay between the amplitude and phase path signals that is less than or equal to a sample interval associated with digital circuitry of said transmitter.

13. The transmitter of Claim 12 wherein said digital circuit comprises a digital filter.

14. The transmitter of Claim 13 wherein said digital filter comprises a linear filter, an LTR filter, an FIR filter, or an all-pass filter.

15. The transmitter of Claim 13 wherein said digital filter comprises an IIR filter having two or more coefficients.

16. The transmitter of Claim 13 wherein said digital filter comprises an FLR filter having two or more coefficients.

17. An electrical circuit, comprising: an amplitude path having circuitry operable to amplitude modulate an input signal to produce an amplitude-modulated signal; a phase path having circuitry operable to phase modulate said input signal to produce a phase-modulated signal; and a digital delay circuit within at least on of said amplitude and phase paths, said digital delay circuit having one or more parameters that can be adjusted to correct for a timing misalignment between said amplitude-modulated and phase-modulated signals.

18. The circuit of Claim 17, further comprising an amplifier configured to receive said amplitude-modulated and phase-modulated signals.

19. The circuit of Claim 17 wherein said digital circuit is capable of effecting a reduction in the timing misalignment to within a fraction of a sampling interval.

20. The circuit of Claim 17 wherein no analog-implemented time delays are performed on either of said amplitude-modulated and phase-modulated signals.

21. The circuit of Claim 19 wherein no analog-implemented time delays are performed on either of said amplitude-modulated and phase-modulated signals.

22. The circuit of Claim 17 wherein said digital delay circuit comprises a digital filter.

23. The circuit of Claim 22 wherein said digital filter comprises a linear filter, an IIR filter, an FIR filter, or an all-pass filter.

24. The circuit of Claim 22 wherein said digital filter comprises an ILR filter having two or more coefficients.

25. The circuit of Claim 22 wherein said digital filter comprises an FLR filter having two or more coefficients.

26. An electrical circuit, comprising: an amplitude path having circuitry for amplitude modulating an input signal to produce an amplitude-modulated signal; a phase path having circuitry for phase modulating the input signal to produce a phase- modulated signal; and means for digitally correcting for timing misalignments between said amplitude- modulated and phase-modulated signals.