Digital down converter with baseband equalization

Abstract

A digital down-converter with baseband equalization comprises a composite analog-to-digital converter (ADC) adapted to convert an applied RF analog signal to be processed to a digital signal, and then down-convert the digital signal to a baseband frequency region, and then perform equalization on the down-converted digital signal, thereby reducing distortions caused by introduction of spurious signals by the ADC.

Claims

1. A digital down-converter system, comprising: A. a composite analog-to-digital converter (ADC), including M sub-ADCs characterized by time-interleaved sampling signals at frequency f.sub.s, where M is an integer, and wherein the M sub-ADCs have a common input for receiving an applied RF analog signal, and have M outputs, wherein each output is associated with a different one of the M sub-ADCs, wherein the composite ADC is adapted to convert the applied analog signal into a set of M partial digital signals, each partial digital signal having a frequency f.sub.s/M, and being synchronous with an associated one of the time-interleaved sampling signals and wherein the M partial digital signals are each applied to an associated one of the M sub-ADC outputs; and whereby the composite ADC is adapted to convert the applied RF analog signals to the M partial digital signals, and B. a down-converter adapted to separately down-convert the M partial digital signals to a baseband frequency region, and then perform equalization separately on each of M1 of the down-converted partial digital signals, thereby reducing distortions caused by introduction of spurious signals by the respective sub-ADCs of the composite ADC.

2. The digital down converter system according to claim 1, wherein the down-converter comprises: a. a set of M IQ demodulators, each IQ demodulator having an input connected to an associated output of the composite ADC, an in-phase output and a quadrature output, wherein each of the IQ demodulator transforms its applied one of the M partial digital signals into an output complex baseband signal, wherein the output complex baseband signal has a real part and an imaginary part, wherein, for each of the IQ demodulators, the real part of the complex baseband signal is applied to the In-phase output I of the IQ demodulator and the imaginary part of the complex baseband signal is applied to the quadrature output Q of the IQ demodulator; b. a set of N misalignment equalizers, where NM1, wherein each misalignment equalizer has an in-phase input connected to the in-phase output of an associated one of the M IQ demodulators, a quadrature input connected to the quadrature output an associated one of the M IQ demodulators, an in-phase output and a quadrature output, wherein each misalignment equalizer is adapted to produce a complex signal, wherein a real part of the produced complex signal is applied to the in-phase output I.sub.R of the misalignment equalizer and the imaginary part of the produced complex signal is applied to the quadrature output Q.sub.R of the misalignment equalizer; c. a first adder having M first adder inputs and a first adder output, wherein the in-phase output I of a first of the IQ demodulators is connected to a first of the M first adder inputs, and the in-phase outputs I.sub.R of the N misalignment equalizers are connected to an associated one of the remaining first adder inputs, wherein the first adder is operative to provide a sum of the first adder inputs PreI at the first adder output; d. a second adder having M second adder inputs and a second adder output, wherein the quadrature output Q of the first of the IQ demodulators is connected to a first of the M second adder inputs, and the quadrature outputs Q.sub.R of the N misalignment equalizers are connected to an associated one of the remaining second adder inputs, wherein the second adder is operative to provide a sum of the second adder inputs PreQ at the second adder output; and e. a common equalizer with a frequency response K.sub.CE(f), having a common in-phase input connected to the first adder output, a common quadrature input connected to the second adder output, a common in-phase output and a common quadrature output, wherein the common equalizer is operative to multiply a spectrum of a complex input signal comprising real part PreI and imaginary part PreQ by the common equalizer frequency response K.sub.CE(f) to obtain a complex product, and apply a real part of the product Out-I to the common in-phase output and apply an imaginary part of the product Out-Q to the common quadrature output.

3. The digital down converter system according to claim 2, wherein the sub-ADCs each include a front-end circuit, wherein the front-end circuit of the m.sup.th sub-ADC, where 0mM1, are characterized by a frequency response H.sub.m(f) and: A. the number N of misalignment equalizers equals M1; B. the frequency responses K.sub.ME,m(f) of the misalignment equalizers are determined by a solution of a system of M1 linear equations, whereas the equation with a number k, where 1kN, has a form $H_0\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right)+\underset{m=1}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right).Math.K_{\mathrm{ME},m}\left(f\right)=0;\mathrm{and}$ C. the common equalizer has a frequency response K.sub.CE(f) $K_{\mathrm{common}}\left(f\right)=T\left(f\right)/\left(H_0\left(f+f_{\mathrm{lo}}\right)+\underset{m=1}{\overset{M-1}{.Math.}}{H}_m\left(f+f_{\mathrm{lo}}\right).Math.K_{\mathrm{ME},m}\left(f\right)\right),$ where T(f) is a target frequency response of the down-converter.

4. The digital down converter system according to claim 2, wherein A. the number N of misalignment equalizers equals M; B. the set of frequency responses K.sub.ME,m(f) of the misalignment equalizers is determined by a solution of a system of M linear equations, wherein the equations have a form $\begin{array}{cc}\underset{m=0}{\overset{M-1}{.Math.}}{K}_{\mathrm{ME},m}\left(f\right).Math.H_m\left(f+f_{\mathrm{LO}}-k.Math.f_s/M\right)=0,& \mathrm{if}k>0,\mathrm{and}\\ \underset{m=0}{\overset{M-1}{.Math.}}{K}_{\mathrm{ME},m}\left(f\right).Math.H_m\left(f+f_{\mathrm{LO}}\right)=D\left(f+f_{\mathrm{LO}}\right),& \mathrm{if}k=0,\end{array}$ where D(f) is a target frequency response at the output of the misalignment equalizers; and C. the common equalizer has a frequency response K.sub.CE(f) equal to
K.sub.common(f)=T(f)/D(f+f.sub.LO), where T(f) is a target frequency response of the down converter.

5. The digital down converter system according to claim 2, wherein the common equalizer has a frequency response K.sub.CE(f)=1 while a set of frequency responses K.sub.ME,m(f) of the misalignment equalizers is found as a solution of a system of M linear equations, wherein the equations have a form $\begin{array}{cc}\underset{m=0}{\overset{M-1}{.Math.}}{K}_{\mathrm{ME},m}\left(f\right).Math.H_m\left(f+f_{\mathrm{LO}}-k.Math.f_s/M\right)=0,& \mathrm{if}k>0,\mathrm{and}\\ \underset{m=0.}{\overset{M-1}{.Math.}}{K}_{\mathrm{ME},m}\left(f\right).Math.H_m\left(f+f_{\mathrm{LO}}\right)=T\left(f+f_{\mathrm{LO}}\right),& \mathrm{if}k=0.\end{array}$

6. The digital down converter system according to claim 5, wherein the number N of the misalignment equalizers equals the number M of the sub-ADCs in the composite ADC.

7. The digital down converter system according to claim 5, wherein the number N of the misalignment equalizers equals M1, and the set of frequency responses K.sub.ME,m(f) of the misalignment equalizers is found as an approximate solution of the system of M linear equations.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 depicts, in block diagram form, an exemplary prior art down-converter with equalization;

(2) FIG. 2 depicts, in block diagram form, an exemplary digital down-converter with equalization according to the present technology;

(3) FIG. 3 depicts, in block diagram form, an exemplary IQ demodulator included in the down-convertor of FIG. 2;

(4) FIG. 4 depicts typical (a) inter-slice amplitude- and phase-frequency responses, and (b) impulse response, of a 32 GS/s composite ADC;

(5) FIG. 5 depicts a QAM64 EVM simulation for a 32 GS/s composite ADC with 2 sub-ADCs (a) without misalignment correction, and (b) with 7-taps misalignment correction;

(6) FIG. 6 depicts a QAM64 EVM simulation for a 32 GS/s composite ADC with 4 sub-ADCs (a) without misalignment equalization, and (b) with three misalignment equalizers using 7 taps; and

(7) FIG. 7 depicts a comparison of EVM level using a conventional 159 taps full-bandwidth equalizer (curve a), a conventional 1150 taps full-bandwidth equalizer (curve b) and a down-converter with baseband equalization of the present technology comprising a 115 taps common equalizer following 7 taps misalignment equalizers (curve c).

DETAILED DESCRIPTION

(8) A block diagram of a first embodiment of a digital down converter with equalization according to the present technology, is shown in FIG. 2. In this diagram, a composite ADC with a sampling frequency f.sub.s, consisting of M time interleaved sub-ADCs, transforms the input analog signal into M partial digital signals, each partial digital signal having a sampling frequency f.sub.s/M.

(9) The front-end circuits of the sub-ADC with the number m, where 0mM1, have a frequency response H.sub.m(f), alternatively referred to as H.sub.m() in FIG. 2. Due to the imperfections of the front-end circuits, the frequency responses H.sub.m(f) are distorted, and are different for different m. If the input analog signal has a spectrum F[f], then the signal, that comes to the sample-and-hold unit in a sub-ADC, has a spectrum F[f].Math.H.sub.m(f).

(10) Since a sub-ADC operates with under-sampling, the signal at its output contains frequency components reflected from the frequencies k.Math.f.sub.s/M, 1kM1. As a result, the partial digital signal, or output signal, produced by the m.sup.th sub-ADC, has a spectrum:

(11) $S_{\mathrm{pds},m}\left(f\right)=1/M\underset{k=0}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f-k.Math.f_s/M\right).Math.F\left(f-k.Math.f_s/M\right)$
(see, for example, J. G. Proakis and D. G. Manolakis, Digital signal processing, 1996, pp 785-787).

(12) One of basic operations performed in a down converter, is a frequency transformation, which transfers a processed signal into a baseband frequency region. In the digital down converter with equalization according to the present technology, the frequency transformation is carried out in parallel, for each partial digital signal separately. That schematic solution makes it possible to correct the distortions caused by the differences between frequency responses of the different sub-ADCs in the baseband frequency region after the frequency transformation.

(13) In the configuration of FIG. 2, the frequency transformation for the output signals of the respective sub-ADCs, is performed by a set of M IQ demodulators. The IQ demodulators are well-known devices, which transform input RF signals into output complex baseband signals, using a local oscillator with the frequency f.sub.lo. The real parts of the output complex baseband signals are placed at the In-phase outputs of the IQ demodulators (outputs I.sub.m in FIG. 2). The imaginary parts of the output complex baseband signals are placed at the quadrature outputs of the IQ demodulators (outputs Q.sub.m in FIG. 2). A detailed block diagram of an exemplary IQ demodulator is shown in FIG. 3. Each IQ demodulator includes two mixers 31 and 33, each having a first input and a second input, and an output coupled to an associated input of one of two combined low pass filters/decimators (LPF-DECs) 35 and 37. In each IQ demodulator, an input signal (from an associated one of the sub-ADCs) is applied to the first input of each of the two mixers 31 and 33. A common local oscillator at frequency .sub.LO provides two sinusoid signals, having a phase difference of 90 degrees; a first sinusoid signal cos (.sub.LOt) is applied to the second input of mixer 31, and a second sinusoid signal sin (.sub.LOt) is applied to the second input of mixer 33. The output of LPF-DEC 35 is an In-Phase baseband partial digital signal, Output I, and the output of LPF-DEC 37 is a Quadrature baseband partial digital signal, Output Q.

(14) The spectrum S.sub.BB,m(f) of the baseband partial digital signal produced by an IQ demodulator with the number m, equals:

(15) $S_{\mathrm{BB},m}\left(f\right)=S_{\mathrm{pds},m}\left(f+f_{\mathrm{lo}}\right)=1\underset{k=0}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right).Math.F\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right),$
where f.sub.lo is the frequency of the local oscillator used in the frequency transformation.

(16) The spectrum S.sub.BB,m(f) consists of the principal part H.sub.m(f+f.sub.lo).Math.F(f+f.sub.lo), corresponding to k=0, and a set of reflections H.sub.m(f+f.sub.lok.Math.f.sub.s/M).Math.F(f+f.sub.lok.Math.f.sub.s/M) from the frequencies k.Math.f.sub.s/M, 1kM1. The total reflection from the frequency k.Math.f.sub.s/M is obtained by summing up over all baseband partial digital signals. The spectrum S.sub.refl,k(f) of the total reflection with the number k equals

(17) $S_{\mathrm{refl},k}\left(f\right)=F\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right)/\underset{k=0}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right).$

(18) In an ideal composite ADC, the sub-ADCs would be identical, so that all frequency responses H.sub.m(f+f.sub.lok.Math.f.sub.s/M) would be the same. In such a situation, all addends in the sum of the last equation cancel each other, so that this sum equals zero. This means that the total reflection from the frequency k.Math.f.sub.s/M equals zero as well. However, in real composite ADCs, the frequency responses H.sub.m(f+f.sub.lok.Math.f.sub.s/M) are different, and the combined reflections are distinct from zero, resulting in the appearance of spurious frequency components in the output signal of a digital down converter.

(19) According to the present technology, the spurious components are eliminated by passing each baseband partial digital signal (except the baseband partial digital signal, produced by IQ demodulator with the number m=0) through a misalignment equalizer with a frequency response K.sub.ME,m(f), 1mM1. The expression for the spectrum S.sub.eq,k(f) of the total reflection with the number k after the equalization becomes:

(20) $S_{\mathrm{eq},k}f)=F\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right)/M.Math.\left(H_0\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right)++\underset{m=1}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right).Math.K_{\mathrm{ME},m}\left(f\right)\right).$

(21) The frequency responses K.sub.ME,m((f), 1mM1, are calculated in such a way as to make the spectrums S.sub.eq,kf) equal to zero for each k, 1kM1. This requirement is equivalent to a set of M1 equalities:

(22) $H_0\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right)+\underset{m=1}{\overset{M-1}{.Math.}}\exp \left(-j2.Math.m.Math.k/M\right).Math.H_m\left(f+f_{\mathrm{lo}}-k.Math.f_s/M\right).Math.K_{\mathrm{ME},m}\left(f\right)=0.$

(23) All together these equalities form a system of M1 linear equations with the unknowns K.sub.ME,m((f). Solution of this system for each frequency fin the down conversion band of interest allows a determination of the desired frequency responses K.sub.ME,m((f) of the M1 misalignment equalizers.

(24) There are two adders in the block diagram of the FIG. 2: the adder #0 and the adder #1. The inputs of the adder #0 are connected to the in-phase output of the IQ demodulator #0 and to the in-phase outputs of the misalignment equalizers. The inputs of the adder #1 are connected to the quadrature output of the IQ demodulator #0 and to the quadrature outputs of the misalignment equalizers.

(25) The outputs of the two adders form a combined complex signal PreI+j.Math.PreQ with a spectrum S.sub.ccs(f). Since the use of the misalignment equalizers suppresses all the reflections, the spectrum S.sub.cs(f) equals:

(26) $S_{\mathrm{ccs}}\left(f\right)=F\left(f+f_{\mathrm{lo}}\right).Math.\left(H_0\left(f+f_{\mathrm{lo}}\right)+\underset{m=1}{\overset{M-1}{.Math.}}{H}_m\left(f+f_{\mathrm{lo}}\right).Math.K_{\mathrm{ME},m}\left(f\right)\right).$

(27) The combined complex signal is processed by a common equalizer. The purpose of this equalizer is to correct the frequency distortions common to all sub-ADCs and to compensate changes of the frequency responses, introduced by the misalignment equalizers. The frequency response of the common equalizer K.sub.common(f) is made equal to:

(28) $K_{\mathrm{common}}\left(f\right)=l/\left(H_0\left(f+f_{\mathrm{lo}}\right)+\underset{m=1}{\overset{M-1}{.Math.}}{H}_m\left(f+f_{\mathrm{lo}}\right).Math.K_{\mathrm{ME},m}\left(f\right)\right).$

(29) The outputs of the common equalizer are used as the outputs of the digital down converter with the equalization. The spectrum S.sub.out(f) of the output complex signal OutI+j.Math.OutQ equals
S.sub.out(f)=S.sub.ccs(f).Math.K.sub.common(f)=F(f+f.sub.lo),
as it should be in a down converter, where all the distortions have been corrected.

(30) A modification of the embodiment of the present technology described above, is possible. The system of M1 linear equations, which is used to find the frequency responses K.sub.ME,m(f) of the misalignment equalizers, may be extended with an additional equation:

(31) $H_0\left(f+f_{\mathrm{lo}}\right)+\underset{m=1}{\overset{M-1}{.Math.}}{H}_m\left(f+f_{\mathrm{lo}}\right).Math.K_{\mathrm{ME},m}(\left(f\right)=1.$

(32) When such extended system of M linear equations is satisfied, the spectrum of the combined complex signal at the adders outputs equals S.sub.ccs(f)=F(f+f.sub.lo), and the common equalizer becomes unnecessary. The extended system of M linear equations contains as before M1 unknowns. Such a system may be solved only approximately, for example by a method of the least squares.

(33) The digital down converter may also be built in such a way that each the baseband partial digital signal produced by an associated IQ demodulator, is processed by a misalignment equalizer. In such a situation the block diagram of the FIG. 2 contains M misalignment equalizers and the extended system of M linear equations may be solved exactly. In the general case, the complex signal spectrum at the output of M misalignment equalizers, may be written as:

(34) $\underset{m=0}{\overset{M-1}{.Math.}}{H}_m\left(f+f_{\mathrm{lo}}\right)-K_{\mathrm{ME},m}(\left(f\right)=D\left(f\right)$
where D(f) is a desirable frequency response at the input of common equalizer. For example, D(f) may be chosen to be equal to an average frequency response of all sub-ADCs. If D(f)=1, then common equalizer is not required.

(35) When the functions of the common equalizer are transferred to the misalignment equalizers, the number of multipliers in the misalignment equalizers increases. The final choice of the structure of the digital down-converter in each specific case, cis best performed after a comparison of required computing recourses.

(36) Usually frequency responses of individual sub-ADCs have moderate deviations without abrupt frequency variations. For example, in FIG. 4, part (a) shows amplitude and phase deviations between two 16 GS/s ADC sub-ADCs. As seen, the amplitude and phase variations are small and exhibit wavy low-frequency dependence. In FIG. 4, part (b) shows reconstructed impulse response (magnified) corresponding to slice deviation. This response is concentrated within 2 ns from the main peak. A composite ADC with a 32 GS/s sampling rate, requires a conventional equalizer (FIR) with 64 taps to correct features of the responses of the sub-ADCs. In a digital down-converter according the present technology, since a misalignment equalizer is applied after decimation, the required FIR length is reduced to 64/M taps, where M is the decimation factor. In this example, only a 7 taps misalignment equalizer is sufficient for M=10.

(37) FIG. 5 shows results of an EVM calculation for a QAM64 signal with a 32 GS/s composite ADC having 2 sub-ADCs, one 7 taps misalignment equalizer and a 115 taps common equalizer. The QAM64 signal with a 2 GHz bandwidth was down-converted with a decimation factor DF=10. A main spur region can be seen around 8 GHz (curve (a) in FIG. 5, calculated without slice correction). However, that spur signal is fully eliminated using the 7 taps misalignment equalizer and the 115 taps common equalizer, with the resulting EVM close to 0.12% (curve (b) in FIG. 5).

(38) A composite 32 GS/s ADC having 4 sub-ADCs, generates spurs at 4 GHz and 8 GHz, corresponding to reflections from 16 and 8 GHz (curve (a) in FIG. 6). An EVM calculation for a down-converter with equalization, which uses 3 misalignment equalizers with 7 taps and a 115 taps common equalizer, shows suppression of both spurs with low a residual EVM level (curve (b) in FIG. 6).

(39) In FIG. 7, curve (a) illustrates achievable EVM levels in a conventional down-converter that uses a 159 taps full-bandwidth equalizer, resulting in 1 to 2% of EVM. A conventional full bandwidth digital down-converter equalizer, correcting all frequency response distortions to a 0.1% EVM level, requires a 1150 taps (EVM levels for that conventional configuration are shown in FIG. 7, curve (b)). Such a very long FIR filter in that configuration cannot be implemented in real time due to the limited number of multipliers that is practical. However, with a digital down-converter implementing the technology, a same residual EVM level is achieved using 7 taps misalignment equalizers followed by a 115 taps common equalizer (the EVM levels for the digital down-converter implementing the technology are shown in FIG. 7, curve (c), where curve (c) is essentially the same as curve (b)).

(40) Although the foregoing description of the embodiment of the present technology contains some details for purposes of clarity of understanding, the technology is not limited to the detail provided. There are many alternative ways of implementing the technology. The disclosed embodiment is illustrative and not restrictive.