Patent Grant
9509331
The present invention pertains to systems, methods and techniques for converting a sampled, quantized (discrete-time) signal into a continuous-time continuously variable (linear) signal. It is particularly applicable to converters that operate at very high sampling rates and with high instantaneous bandwidth.
Many applications in modern electronics require that discrete-time signals, generated using computers and digital signal processors, be converted to linear (analog) signals, e.g., for transmission as electromagnetic signals. Typically, this transformation is made using a conventional digital-to-analog converter (DAC). However, the present inventor has discovered that each of the presently existing converters exhibits shortcomings that limit overall performance at very high sampling rates.
Due to parallel processing and other innovations, the digital information processing bandwidth of computers and signal processors has advanced beyond the capabilities of state-of-the art DACs. Therefore, converters with higher instantaneous bandwidth are desired. Existing solutions are limited by instantaneous bandwidth (sampling rate), effective conversion resolution (accuracy), or both.
The resolution of a DAC is a measure of the precision with which a quantized signal can be transformed into a continuous-time continuously variable signal, and typically is specified as a ratio of the total signal power to the total noise plus distortion power at the DAC output. This signal-to-noise-and-distortion ratio (SNDR) of a DAC is commonly expressed on a logarithmic scale in units of decibels (dB). When a discrete-time discretely variable (digital) signal is converted into a continuous-time continuously variable (analog) signal, the quality of the analog signal is corrupted by various limitations and errors introduced during the conversion process. Examples include: 1) the finite granularity of the DAC output levels, which produces quantization noise; 2) the imprecise (e.g., nonlinear) mapping of digital inputs to corresponding discrete output voltage or current levels, which introduces distortion in the form of rounding inaccuracies (rounding errors); 3) the imperfect timing between transitions in output voltages or currents relative to transitions in digital inputs, which causes noise in the form of sampling jitter; and 4) the thermal noise associated with active devices (e.g., switches and amplifiers) which couples onto the DAC output. High-resolution converters transform discrete signals into continuously variable signals using a rounding operation with fine granularity, and a more linear mapping of digital inputs to output voltage and/or current. Instantaneous conversion bandwidth is limited by the Nyquist criterion to a theoretical maximum of one-half the converter sampling rate (the Nyquist limit), such that aliasing occurs when the converted signal contains frequency components which exceed the Nyquist limit. High-resolution conversion (of ≧10 bits) conventionally has been limited to instantaneous bandwidths of about a few gigahertz (GHz) or less.
Converters that transform digital signals into analog signals with fine granularity (i.e., transform a digital signal using many discrete output levels) and a sampling rate fS that is equal to, or just greater than, twice the maximum frequency fMAX spanned by the digital signal, are conventionally known as Nyquist-rate converters, or Nyquist converters. Conventional Nyquist-rate converters include those implemented using resistor ladder networks (e.g., R-2R ladders), or those employing switched current/voltage sources with unitary (i.e., equal) weighting or binary weighting. A conventional R-2R ladder DAC, such as that shown in
Conventional Nyquist-rate converters potentially can achieve very high instantaneous bandwidths, but as discussed in greater detail below, the present inventor has discovered that component mismatches in the resistor ladder network, or in the switched current sources, can introduce rounding errors that significantly limit attainable resolution. In addition, the resolution of conventional Nyquist-rate converters is limited by other practical implementation impairments such as sampling jitter and thermal noise. Although in theory, Nyquist-rate converters potentially could realize high resolution at instantaneous bandwidths greater than 10 GHz, this potential has been unrealized in conventional Nyquist-rate converters due to the foregoing problems.
A conventional approach that attempts to reduce quantization noise and errors uses an oversampling technique. Conventional Nyquist-rate converters transform digital input samples into variable-level output samples (i.e., as a voltage or a current), such that a single input sample is represented by a single output sample, and the value of each output sample is proportional to the digital input. In contrast, conventional oversampling converters transform digital input samples into outputs which are pseudorandom sequences of two-level samples (i.e., output samples having a single positive level or a single negative level), such that: 1) a single input sample is represented by multiple output samples; and 2) the average of these multiple output samples is proportional to the digital input. Therefore, oversampling converters generate coarse (e.g., two-level) analog voltages or currents at a sampling rate (i.e., fS) that is much higher than twice the occupied bandwidth fB of the input signal (i.e., fS>>fB), where: 1) fB is equal to the Nyquist frequency fMAX for lowpass (baseband) input signals; and 2)
is conventionally referred to as the oversampling ratio of the converter. A continuously variable output that is proportional to the digital inputs is produced from the pseudorandom sequences of two-level outputs, using a filtering operation that effectively averages the output samples. Although this averaging process reduces the instantaneous bandwidth of the oversampling converter (i.e., the maximum frequency that can be converted without exceeding the Nyquist limit), it has the benefit of improving the converter resolution by attenuating quantization noise (i.e., the noise introduced by using only two levels to represent a continuously variable signal) and errors resulting from component mismatches, sampling jitter, and thermal noise. The extent of this benefit is directly related to the output sampling rate fS (i.e., the benefit increases as the sampling rate increases), and is conventionally enhanced using oversampling in conjunction with an operation referred to as noise-shaped quantization, that ideally attenuates conversion noise and errors in the signal bandwidth without also attenuating the signal itself. Through this noise-shaped quantization operation and subsequent filtering (i.e., output averaging), oversampling converters transform a high-rate intermediate signal having low resolution, into a relatively low bandwidth output signal having improved resolution.
is the response of integrator operation 13 and z−1 represents a unit delay equal to 1/fCLK (i.e., z−1 represents a delay corresponding to one cycle of the modulator clocking rate fCLK). Converter 5A, shown in
Generally speaking, the delta-sigma modulator processes the signal with one transfer function (i.e., the signal transfer function or STF) and the quantization noise with a different transfer function (i.e., the noise transfer function or NTF). Conventional transfer functions (i.e., after accounting for the implicit delay of the clocking operation on two-level quantizer 10) are of the form STF(z)=z−k and NTF(z)=(1−z−1)P, where k is an integer and P is called the order of the modulator (or order of the noise-shaped response). Converter circuits 5A&B employ first-order ΔΣ modulation (i.e., P=1) that produces STF frequency response 30 and NTF frequency response 32 that are shown in
For a given converter resolution, the bandwidth of a conventional oversampling converter typically is increased by increasing the clocking frequency fCLK of the ΔΣ modulator (i.e., increasing the sampling rate fS), thereby making the oversampling ratio N higher. Similarly, for a given bandwidth, a higher oversampling ratio N results in improved converter resolution. Generally speaking, the present inventor has determined that the resolution B of a conventional oversampling converter is given by
where ΔQ is the number of bits at the output of quantization circuit 10 (i.e., level of coarse quantization which typically is equal to one bit) and F(e2πjfT) is the frequency response of output filter 12. Increasing the clock frequency fCLK of the ΔΣ modulator requires circuitry with higher speed capability, and generally, higher power dissipation. Alternatively, higher bandwidth and/or improved resolution are realized by increasing the order P of the ΔΣ modulator. Compared to converter circuits 5A&B, lowpass oversampling converter 5C, illustrated in
The delta-sigma converters 5A-C illustrated in
is the response of integration operation 14 and z−1represents a unit delay equal to 1/fCLK. After accounting for the implicit delay of the clocking operation on two-level quantizer 10, conventional bandpass ΔΣ modulator 42 has a STF(z)=z−1 and a second-order NTF(z)=1−z−2. Like converter circuits 5A&C, bandpass oversampling converter circuit 40A is an interpolative structure that produces signal response 70, shown in
which is at the center of the converter Nyquist bandwidth. Producing a NTF with a spectral null at a frequency other than zero hertz requires a real ΔΣ modulator with, at minimum, a second-order response (i.e., the delay operator z is raised to a power of −2), and in general, the NTF of a bandpass ΔΣ modulator is of the form (1+ρ·z−1+z−2)P, where −2≦ρ≦+2. Although signal response 70 of circuit 40A is all-pass, the present inventor has discovered that, in general, the STF of bandpass oversampling converters is not all-pass when interpolative modulator structures are employed. Conversely, the present inventor has discovered that bandpass oversampling converters which utilize the alternative error-feedback structure of
Although oversampling with noise-shaped quantization can reduce quantization noise and other conversion errors, the output filtering (i.e., smoothing) operations generally limit the utility of oversampling converters to applications requiring only low instantaneous bandwidth (e.g., input signals with low frequency content). Conventional schemes for overcoming the bandwidth and resolution limitations of data converters generally have been devised with a focus on the conversion of analog (linear) signals to digital (discrete) signals (i.e., analog-to-digital conversion), rather than on the conversion of digital signals to analog signals (i.e., digital-to-analog conversion), which is the subject of the present invention. The present inventor has discovered that these conventional schemes for improving bandwidth and/or resolution in analog-to-digital conversion suffer from significant disadvantages, particularly in attempts to directly adapt these schemes for use in digital-to-analog conversion applications.
For example, one attempt to overcome the instantaneous bandwidth limitations of high-resolution, analog-to-digital (A/D) converters is conventional hybrid filter bank (HFB) converter 50, illustrated in
A second attempt to overcome the instantaneous bandwidth limitations of high-resolution, analog-to-digital (A/D) converters is conventional Multi-Band Delta-Sigma (MBΔΣ) converter 70, shown in
In addition to the conventional frequency-interleaved schemes employed by converters 50 and 70 (i.e., schemes involving spectral decomposition of the converter input signal), another attempt at overcoming the instantaneous bandwidth limitations of high-resolution, analog-to-digital converters involves the use of conventional time-interleaving to increase the bandwidth, or equivalently, the sampling rate of a ΔΣ modulator. Circuits 80A&B, which are illustrated in
In circuits 80A&B of
delay equal to 2/fS, where fS is the sampling rate of the complete converter (i.e., the full-rate before polyphase decomposition). Circuit 80A is a lowpass modulator with a NTF response that is first-order (i.e. , P=1), and circuit 80B is a lowpass modulator with a NTF response that is approximately second-order (i.e., P=2). But rather than decomposing the entire modulator into parallel (polyphase) circuits, in such conventional converters the difference function of the modulator (i.e., subtractors 8A&B of circuit 80A&B) and the quantization function of the modulator (i.e., quantizers 10A&B in circuits 80A&B) are simply replicated m times and distributed across the m parallel processing paths. See R. Khoini-Poorfard, L. B. Lim, and D. A. Johns, “Time-Interleaved Oversampling A/D Converters: Theory and Practice,” IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, 1997. As discussed in greater detail below, the present inventor has discovered that simple replication and distribution of the difference and quantization functions (i.e., using time-interleaving through simple replication rather than through polyphase decomposition) causes conventional time-interleaved ΔΣ modulators to exhibit undesirable properties that prevent their use in very high-rate converter applications.
Referring to conventional circuit 80A of
where the Q(•) operator represents quantization (rounding). The above difference equation results in a STF(z)=z−1 (i.e., an all-pass response) and a NTF(z)=1−z−1, where z−1 represents a full-rate delay, equal to one period of the effective sampling rate fS. Although the STF and NTF of the time-interleaved ΔΣ modulator are equal to those of a full-rate ΔΣ modulator with a first-order shaping (i.e., a ΔΣ modulator which is not time-interleaved), the output of the circuit is a function of a previous output Q(yn−1) which has been delayed by only one full-rate delay (i.e., one period of the effective sampling rate fS) relative to the output Q[yn]. The present inventor has observed that the need to feed back and process outputs that are delayed by only one full-rate period of 1/fS (i.e., the need for processing computations to take place within one full-rate cycle), rather than one sub-rate period of m/fS=2/fS, creates a race condition that forces circuit 80A to operate at speeds equivalent to the full-rate sampling frequency of fS, rather than the intended sub-rate speed of
This race condition occurs because the difference and quantization functions of the time-interleaved modulator are not implemented as true polyphase (multirate) operations. A similar race condition occurs in the implementation of circuit 80B, which has an output Q(yn) that is described by the difference equation
(i.e., assuming no implicit quantizer delay). Since the output of circuit 80B also depends on an output Q(yn−1) that has been delayed by only one full-rate period equal to 1/fS, the circuit also must operate at a full-rate of fS, rather than the intended sub-rate of
In addition to the computational race condition, the present inventor has discovered that circuit 80B exhibits three other undesirable properties: 1) the time-interleaved modulator has a signal transfer function STF(z)=z−2·(1+z−1)2/(1+z−1+z−3) which deviates from a true all-pass response of STF(z)=z−k; 2) the time-interleaved modulator has a noise transfer function NTF(z)=(1−z−1)2/(1+z−1+z−3) which deviates from the desired form of (1−z−1)P for a full-rate lowpass modulator; and 3) the cascaded integrator structure of the second-order modulator is impractical for use in bandpass converter applications because second-order (bandpass) NTFs of the form (1+ρ·z−1+z−2)P, where −2≦ρ≦+2, cannot be factored into cascaded first-order functions of the form (1+α·z−1)·(1+β·z−1).
Besides utilization in analog-to-digital (A/D) converter applications, time-interleaved ΔΣ modulators are employed in conventional circuits which attempt to improve the effective sampling rates and/or instantaneous bandwidths of digital-to-analog (D/A) converters. An example is conventional D/A converter circuit 80C, illustrated in
The present inventor has discovered that like conventional circuits which utilize time-interleaved ΔΣ modulators for A/D conversion, race conditions (i.e., conditions where outputs must be fed back as inputs for processing within one full-rate cycle) limit the effective sampling rate and instantaneous bandwidth of conventional circuits which utilize time-interleaved ΔΣ modulators for D/A conversion. More specifically, race conditions occur because in these conventional D/A converters (e.g., circuit 80C), polyphase decomposition is applied to only a portion of the time-interleaved ΔΣ modulator. As illustrated by conventional converter 80C, shown in
computational results must propagate from the first modulator output (i.e., the output of quantizer 10B) to the input of the mth truncation element (i.e., the input of quantizer 10A) in a time of less than m full-rate cycles (i.e., a total time of less than m/fS). Consequently, the modulator computations occur at a speed equivalent to the full-rate sampling frequency of fS, or with a computational latency of 1/fS, to ensure that m computations (i.e., one computation per parallel path) can traverse through the parallel processing blocks during the allotted time of m/fS. The maximum effective sampling rate of conventional converter 80C is further constrained because the conventional m-to-1 multiplexer (e.g., multiplexer 16B), which combines modulator outputs, must run at the full-rate of fS, requiring the multiplexer to be implemented using high-speed circuitry with correspondingly high power consumption.
The present inventor has discovered that conventional lowpass ΔΣ converters, as illustrated in
The present invention provides an improved DAC, particularly for use at very high sampling rates and instantaneous bandwidths (e.g., maximum input frequencies) approaching the Nyquist limit. The improved DAC overcomes the resolution limitations of conventional Nyquist-rate converters and the bandwidth limitations of conventional oversampling converters.
Thus, one specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal and includes: 1) an input line for accepting a high-rate input signal that is discrete in time and in value; 2) a demultiplexer coupled to the input line that distributes the samples of the high-rate input signal to multiple, parallel outputs having a lower sampling rate; 3) a discrete-time noise-shaping/quantization circuit having a plurality of parallel processing paths, such that the inputs of the parallel paths are either coupled to the multiple outputs of the demultiplexer or are coupled to the outputs of other parallel paths, with each such parallel path generating a different subsampling phase of a complete signal that is output by the discrete-time noise-shaping/quantization circuit; 4) a multiplexer coupled to the parallel outputs of the discrete-time noise-shaping/quantization circuit that converts low-rate, parallel inputs into a serial (multi-level) output signal which reflects high-rate sampling; 5) a multi-bit-to-variable-level signal converter coupled to an output of the discrete-time noise-shaping/quantization circuit; and 6) an analog bandpass filter coupled to an output of the discrete-time noise-shaping/quantization circuit. Together, the parallel paths of the discrete-time noise-shaping/quantization circuit produce a conversion-noise frequency response with a minimum magnitude at a frequency which corresponds to a band selected by the analog bandpass filter. The parallel outputs of the discrete-time noise-shaping/quantization circuit operate at a subsampling rate (fCLK) which is m times less than the sampling rate (fS) of the high-rate input signal, such that
where m is the number of parallel paths, and the outputs of at least some of the parallel paths are a function only of: 1) samples of the input signal (i.e., inputs to the parallel paths); and 2) previous outputs from the parallel paths that have been delayed by at least one period of the subsampling rate at which a parallel path operates (i.e., a minimum delay of 1/fCLK=m/fS).
An alternate specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal and includes: 1) an input line for accepting an input signal that is discrete in time and in value; 2) an adder having a first input coupled to the input line, a second input and an output; 3) a multi-bit, discrete-time noise-shaping/quantization circuit coupled to the output of the adder; 4) a multi-bit-to-variable-level signal converter coupled to the output of the discrete-time noise-shaping/quantization circuit; 5) an analog bandpass filter coupled to the output of the discrete-time noise-shaping/quantization circuit; 4) a nonlinear bit-mapping component, having an input coupled to the output of the discrete-time noise-shaping/quantization circuit and also having an output; and 5) a feedback-loop filter having a first input coupled to the output of the nonlinear bit-mapping component, a second input coupled to the output of the adder, and an output coupled to the second input of the adder. The discrete-time noise-shaping/quantization circuit has a conversion-noise frequency response with a minimum magnitude at a frequency which corresponds to a band selected by the analog bandpass filter. The nonlinear bit-mapping component scales different bits of a multi-bit signal at its input, using different multi-bit factors to produce an output with an intentionally imperfect binary weighting, such as an imperfect binary weighting that is matched to the unintentionally imperfect binary weighting of a conventional R-2R ladder network.
A third specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal and includes: 1) an input line for accepting an input signal that is discrete in time and in value; 2) multiple processing branches coupled to the input line; and 3) an adder. Each of the processing branches includes: (a) a discrete-time noise-shaping/quantization circuit; (b) a multi-bit-to-variable-level signal converter coupled to an output of the discrete-time noise-shaping/quantization circuit; and (c) an analog bandpass filter coupled to the output of the discrete-time noise-shaping/quantization circuit. The adder is coupled to an output of the analog bandpass filter in each of the processing branches. The discrete-time noise-shaping/quantization circuits in different ones of the processing branches have conversion-noise frequency responses with minima at different frequencies, and each of the discrete-time noise-shaping/quantization circuits produce a conversion-noise frequency response with a minimum magnitude at a frequency which corresponds to a band selected by the analog bandpass filter in the same processing branch. The analog filters in the various processing branches have frequency response orders that are not greater than 10 and preferably have standard analog filter responses (e.g., Butterworth, Chebychev, coupled-resonator), where the center frequencies, bandwidths, and/or orders of one or more filters has been made intentionally unequal to the others to minimize the amplitude and group delay distortion introduced by the composite filter bank response (i.e., the summed frequency responses of the filters in the various processing branches). It is noted that in applications where conversion at zero frequency (i.e., DC) is desired, one of the processing branches preferably includes an analog bandpass filter that is centered at zero frequency to produce a lowpass response.
A fourth specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal that includes: 1) an input line for accepting an input signal that is discrete in time and in value; 2) a digital pre-distortion linearizer (DPL) coupled to the input line; 3) multiple processing branches coupled to the DPL; and 4) an adder. Each of the processing branches includes: (a) a discrete-time noise-shaping/quantization circuit; (b) a multi-bit-to-variable-level signal converter coupled to an output of the discrete-time noise-shaping/quantization circuit; and (c) an analog bandpass filter coupled to an output of the discrete-time noise-shaping/quantization circuit. The adder is coupled to an output of the analog bandpass filter in each of the processing branches. The discrete-time noise-shaping/quantization circuits in different ones of the processing branches have conversion-noise frequency responses with minima at different frequencies, and each of the discrete-time noise-shaping/quantization circuits produce a conversion-noise frequency response with a minimum magnitude at a frequency which corresponds to a band selected by the analog bandpass filter in the same processing branch. Unlike conventional HFB schemes that use analog filters to divide an input signal into narrowband segments, the DPL is a digital filter that does not perform such a frequency decomposition function in the representative embodiment of the invention. Instead, the frequency response of the DPL preferably has intentional amplitude and group delay (i.e., phase) variation, such as intentional amplitude and group delay variation that is equal and opposite to the unintentional amplitude and group delay variation that occurs within an analog filter bank which is constructed from analog filters with standard frequency responses (e.g., Butterworth, Chebychev, coupled-resonator, etc.). It is noted that the cascaded response of the DPL and the imperfect analog filter bank preferably is approximately all-pass, and therefore, forms a filter bank with near-perfect signal reconstruction properties.
A fifth specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal and includes: 1) an input line for accepting an input signal that is discrete in time and in value, and is sampled at Nyquist-rate; 2) a demultiplexer (deserializer) coupled to the input line that converts an input signal into a high-rate (oversampled) output signal, the samples of which are distributed to multiple, parallel outputs having a lower sampling rate, such that in combination, the low-rate parallel outputs represent a high-rate signal; 3) a discrete-time noise-shaping/quantization circuit having a plurality of parallel processing paths, such that the inputs of parallel paths are either coupled to the multiple outputs of the demultiplexer or are coupled to the outputs of other parallel paths, with each such parallel path generating a different subsampling phase of a complete signal that is output by the discrete-time noise-shaping/quantization circuit; 4) a multiplexer coupled to the parallel outputs of the discrete-time noise-shaping/quantization circuit that converts low-rate, parallel inputs into a serial (multi-level) output signal which reflects high-rate sampling; 5) a multi-bit-to-variable-level signal converter coupled an output of the discrete-time noise-shaping/quantization circuit; and 6) an analog bandpass filter coupled to an output of the discrete-time noise-shaping/quantization circuit. Through parallel processing (i.e., within the plurality of parallel paths), the discrete-time noise-shaping/quantization circuit has a conversion-noise frequency response with a minimum magnitude at a frequency which corresponds to a band selected by the analog bandpass filter. The parallel outputs of the discrete-time noise-shaping/quantization circuit operate at a sampling rate (fCLK) which is no more than twice the maximum frequency spanned by the input signal, such that fCLK≦2·fMAX, and the outputs of at least some of the parallel paths are a function only of: 1) samples of the input signal (i.e., inputs to the parallel paths); and 2) previous outputs from the parallel paths that have been delayed by more than a sample period of the input signal times a total number of the parallel paths (i.e., a minimum delay of m/(2·fMAX) for a number of parallel paths equal to m).
A sixth specific embodiment of the invention is directed to an apparatus for converting a discrete-time quantized signal into a continuous-time, continuously variable signal and includes: 1) an input line for accepting the full-rate samples of an (underlying) input signal that is discrete in time and in value; 2) a parallel signal processor having an input coupled to said input line and also having a plurality of sub-rate outputs; 3) a multi-bit-to-variable-level signal converter which is coupled to a sub-rate output of the parallel signal processor, and which operates at a sampling rate that is less than or equal to the full-rate sampling frequency of the input signal; and 5) a signal combiner coupled to an output of the multi-bit-to-variable-level signal converter. The parallel signal processor includes a serial-to-parallel demultiplexing operation that sequentially couples input samples to multiple sub-rate outputs, such that the sequence of samples provided at each sub-rate output represents a subsampled version of the underlying (complete) input signal at a particular subsampling phase, with the different sub-rate outputs representing different subsampling phases. The samples on each sub-rate output are provided at a sampling rate (i.e., a sub-rate) which is less than or equal to the full-rate sampling frequency of the input signal. Additionally, the relationship between the sampling rate (i.e., sub-rate) on each sub-rate output and the sampling rate associated with the underlying input signal depends on the processing which occurs within the parallel signal processor: 1) the samples on each sub-rate output are provided at a lower sampling rate than the sampling rate of the underlying input signal when the parallel signal processor includes no upsampling (e.g., the full-rate sampling frequency of the input signal divided by the number of processing branches); and 2) the samples on each sub-rate output are provided at a sampling rate which can equal (or even exceed) the sampling rate of the underlying input signal when the sub-rate signal processor includes some form of upsampling, such as zero insertion, sample repetition, and/or interpolation. Using delay and summation operations, the sub-rate outputs of multi-bit-to-variable level signal converters are combined as continuous-time signals to produce an output sequence of values which represents a filtered version of the full-rate, underlying input signal. Typically, the filter response is a lowpass (e.g., a sin (x)/x) function with a bandwidth smaller than the full-rate sampling frequency and larger than a maximum frequency component of the underlying input signal. The delay operations introduce different time-offsets in increments equaling integer multiples of a corresponding full-rate period (i.e., the period associated with full-rate sampling). In one variation of the present embodiment, the delay operations introduce time-offsets via digital resampling on different phases of a sub-rate clock, and outputs of the delay operations are coupled to inputs of the multi-bit-to-variable-level signal converters. In an alternate variation of the present embodiment, the delay operations introduce time-offsets via signal propagation through continuous-time circuitry (e.g., passive delay lines, ladder networks, etc.), and inputs of the delay operations are coupled to outputs of the multi-bit-to-variable-level signal converters.
According to another aspect of any of the foregoing embodiments, the invention also encompasses an apparatus for combining the multiple sub-rate outputs of a polyphase (parallel) processor into a single full-rate output, and includes: 1) a plurality of input lines for accepting the sub-rate samples of an input signal; 2) a plurality of delay elements which are coupled to the input lines and which introduce different time-offsets in increments equaling integer multiples of a corresponding full-rate period; and 3) a signal combiner with inputs that are coupled to the outputs of the delay elements. The sub-rate samples on each of the plurality of input lines preferably are a sequence of values representing different subsampling phases of a full-rate and complete underlying signal. The plurality of output signals provided by the delay elements are summed within the combiner to produce an output sequence of values which represents a full-rate and filtered version of the complete signal. Typically, the filter response is a lowpass (e.g., a sin (x)/x) function with a bandwidth smaller than the full-rate sampling frequency.
A digital-to-analog (D/A) converter apparatus created by incorporating one or more of the specific embodiments of the invention described above, typically can provide a better combination of high resolution and wide bandwidth than is possible with conventional D/A converters and can be used for various commercial, industrial and military applications, e.g., in various direct conversion transmitters, software-defined or cognitive radios, multi-channel communication transmitters, all-digital RADAR systems, and high-speed arbitrary waveform generators.
The foregoing summary is intended merely to provide a brief description of certain aspects of the invention. A more complete understanding of the invention can be obtained by referring to the claims and the following detailed description of the preferred embodiments in connection with the accompanying figures.
In the following disclosure, the invention is described with reference to the attached drawings. However, it should be understood that the drawings merely depict certain representative and/or exemplary embodiments and features of the present invention and are not intended to limit the scope of the invention in any manner. The following is a brief description of each of the attached drawings.
In a manner somewhat comparable to conventional, oversampling digital-to-analog (D/A) converters, a preferred discrete-to-linear converter according to the present invention employs a form of “oversampling” (as that term is broadly used herein) in conjunction with noise-shaped quantization to mitigate the resolution-degrading effects of coarse quantization, rounding errors (i.e., distortion), and thermal noise. However, a converter according to the preferred embodiments of the present invention incorporates one or more of the following technological innovations to improve instantaneous bandwidth and resolution: 1) multiple oversampling converters (e.g., each processing a different frequency band) are operated in parallel to overcome the bandwidth limitations of conventional oversampling converters; 2) multirate (i.e., polyphase) delta-sigma modulators (preferably second-order or higher) are used in place of conventional delta-sigma (ΔΣ) modulators, or conventional time-interleaved ΔΣ modulators, so that the effective oversampling ratio of the modulator is not strictly dependent on the modulator clocking frequency fCLK (or the switching/sampling speed of digital modulator circuits); 3) multi-bit quantizers are used in conjunction with multi-bit-to-variable-level signal converters, such as resistor ladder networks or current source networks, to allow stable operation with noise-shaped responses that are higher than second-order; 4) nonlinear bit-mapping is used to compensate for mismatches (rounding errors) in the multi-bit-to-variable-level signal converters (e.g., by replicating such mismatches so that the resulting distortion is shaped into a frequently range where it will be attenuated by a corresponding bandpass filter); 5) multi-band (e.g., programmable NTF response) delta-sigma modulators are used in place of single-band (i.e., fixed NTF response) delta-sigma modulators to enable a single modulator circuit to be configured for operation on arbitrary frequency bands; and 6) a digital pre-distortion linearizer (DPL) is used so that an analog signal reconstruction filter bank, based on standard analog filter structures of low order, can effectively attenuate conversion noise and errors without introducing appreciable amplitude and phase distortion. Certain combinations of such techniques are sometimes is referred to herein as Multi-Channel Bandpass Oversampling (MBO). An MBO converter can in some respects be thought of as comprising unique and novel methods of combining two distinct conventional techniques: 1) continuous-time, bandpass oversampling; and 2) multi-channel, frequency-decomposition. As discussed in more detail below, the use of such techniques often can overcome the problems of limited conversion resolution and precision at very high instantaneous bandwidths.
Simplified block diagrams of converters 110A&B and 200A-C according to certain preferred embodiments of the present invention are illustrated in
In any event, parallel outputs 108A of demultiplexer 107A (i.e., signals xm−1 . . . x0) are coupled to the parallel inputs of discrete-time noise-shaping/quantization circuit 112A, which processes demultiplexer outputs 108A using parallel paths (e.g., ΔΣ processing paths 105A-C) to produce low-resolution, noise-shaped outputs 108B (i.e., signals ym−1 . . . y0). Outputs 108B of parallel processing paths 105A-C are fed back as inputs to the parallel paths, and are also coupled to the parallel inputs of multiplexer 107B. Multiplexer 107B, in conjunction with multi-bit-to-variable-level converter 113A (e.g., a resistor ladder network or current source network), combines parallel outputs 108B of discrete-time noise-shaping/quantization circuit 112A, to produce a serial output signal (i.e., analog output signal 109B) which reflects coarse quantization and a high effective sampling rate. High-rate, analog output 109B is then coupled to bandpass filter 115, which in addition to smoothing the output of multi-bit-to-variable-level converter 113A, attenuates the shaped quantization noise at the outputs of ΔΣ processing paths 105A-C. The parallel operation of discrete-time noise-shaping/quantization circuit 112A is based on polyphase decomposition, except that unlike conventional approaches where only a portion (i.e., only the integrator or loop filter) of a ΔΣ modulator is decomposed, race conditions are eliminated by preferably decomposing the entire noise-shaping/quantization circuit into a polyphase structure, using the means described in greater detail in the Noise-Shaping and Quantizing Considerations section below. Generally speaking, the outputs of parallel paths are fed back into the inputs of parallel paths, with subsequent preprocessing ensuring that the mean level of high-rate, coarsely-quantized output 109B, is proportional to the value of digital signal 102 (i.e., the signal input on line 103). Through such preprocessing, the residual quantization noise at the output of the noise-shaping/quantization circuit is shifted away (i.e., noise-shaped) from the frequency band occupied by digital input signal. As used herein, the term “coupled”, or any other form of the word, is intended to mean either directly connected or connected through one or more other processing blocks, e.g., for the purpose of preprocessing.
It should be noted that converter 110A (shown in
In the preferred embodiments of the invention, the noise-shaping/quantization circuit (e.g., circuit 112A) utilizes ΔΣ modulation (or other noise-shaped quantization methods) to produce NTFs with noise-shaped responses that are second-order or greater. And when the order of the noise-shaped response is greater than two, multi-bit quantizers (e.g., quantizer 114 shown in
To maximize discrete-to-linear (i.e., digital-to-analog) conversion bandwidth and resolution, multiple converters can be operated in parallel using a structure that is somewhat similar to conventional MBΔΣ approaches for analog-to-digital conversion, but with key differences that will become clear below. Such a technique of operating multiple converters in parallel, with each converter processing a different portion of the frequency band occupied by the input signal, sometimes is referred to herein as Multi-Channel Bandpass Oversampling (MBO). Simplified block diagrams of MBO converters 200A-C according to the preferred embodiments of the present invention are illustrated in
In certain conventional frequency-interleaved converters, such as an HFB analog-to-digital converter, each sub-converter in the interleaved array operates at a submultiple of the effective sampling rate
due to the reduced signal bandwidth in each of the subdivided bands. In contrast, converters 200A-C according to the present invention separately processes M different frequency bands, with each band preferably operating at the effective sampling rate of fS, rather than at a submultiple of the effective sampling rate. This approach results in an oversampling ratio of M, sometimes referred to herein as an “interleaved oversampling ratio” or “interleave factor”. It should be noted that the interleave factor M is different from the excess-rate oversampling ratio N of a conventional oversampling converter, but generally has the same or similar effect on conversion noise and errors. It is noted that, except to the extent specifically indicated to the contrary, the term “oversampling” is used herein in a broad sense, referring to processing techniques in which a signal, or some portion of the signal, is digitally represented during some intermediate stage at a higher sampling rate (but typically at a lower resolution) than the signal, or portion thereof, that ultimately is output. In the preferred embodiments of the present invention, input digital signal 102 is processed in different channels or branches (e.g., branches 110 and 120), the purpose of each being to convert a different frequency band. It is noted that each such processing branch could be implemented, e.g., using either of the structures shown in
Referring to
In the present embodiment of converter 200A, the samples of input digital signal 102 are first coupled, or distributed, to M different branches (e.g., branches 110 and 120), each processing a different frequency band and each preferably including: 1) a discrete-time noise-shaping/quantization circuit (e.g., noise-shaped quantizer 112 or 122); 2) a multi-bit-to-variable-level signal converter, such as resistor ladder network 113B; and 3) a bandpass (signal reconstruction) filter (e.g., filter 115 or 125). Lastly, adder 131 sums the outputs of these M branches (more specifically, the outputs of the signal reconstruction filters) to produce final output signal 135. As used herein, the term “distributes”, or any other form of the word, is intended to mean provides, either through direct connection or through one or more other processing blocks, e.g., for the purpose of preprocessing. Rather than replicating the finite impulse response (FIR) of the relatively high-order, transversal window filters (e.g., Hann, Hamming, etc.) used in conventional MBΔΣ schemes, each of the bandpass filters (e.g., filter 115 and 125) at the output of a processing branch preferably is a relatively low-order filter (i.e., order of 7-10 or less) with a standard analog filter response, such as a Butterworth, Chebychev, Bessel or coupled-resonator response. Particularly at high frequency (e.g., gigahertz frequencies), these standard analog filter responses can be realized as passive structures without excessive circuit complexity. The center frequency, bandwidth, and/or order of the filters in each of the multiple processing branches preferably are independently adjusted to minimize the amplitude and group delay distortion introduced by all the filter responses in combination (i.e., the amplitude and group delay distortion introduced by imperfect signal reconstruction). Preferably, the filter responses are adjusted to produce amplitude variation of less than ±1.5 dB and group delay variation of less than ±12.5 periods of the effective sampling rate fS. Often, for ease of reference, the following discussion refers only to the components of branch 110, it being understood that similar processing preferably occurs in each of the other branches (e.g., branch 120).
Similar processing to that described above occurs within converters 200B&C of
Although the representative embodiments described above and illustrated in
The term “adder”, as used herein, is intended to refer to one or more circuits for combining two or more signals together, e.g., through arithmetic addition and/or (by simply including an inverter) through subtraction. The term “additively combine” or any variation thereof, as used herein, is intended to mean arithmetic addition or subtraction, it being understood that addition and subtraction generally are interchangeable through the use of signal inversion. The term “bandpass”, as used herein, refers to a filter or other circuit that provides higher gain for a desired band of frequencies as compared to other frequencies within the input signal, where the desired band could be centered at zero (in which case it could be referred to as a lowpass filter) or at any other frequency.
Furthermore, in the present embodiments, the typically multi-bit output of each noise-shaping/quantization circuit 112 is converted into a single variable-level signal, which, via a resistor ladder network (e.g., R-2R network 113B), switches among a fixed number of discrete levels when the output of the corresponding noise-shaping/quantization circuit 112 changes. However, other multi-bit-to-variable-level signal converters known in the art, such as binary-weighted or unitary-weighted current sources, instead may be used. Also, as in converter 200C shown
In accordance with one aspect of certain preferred embodiments, the present invention overcomes the problems of limited conversion resolution and precision at high instantaneous bandwidths via a novel method of combining two established techniques—bandpass oversampling and a variant of frequency interleaving. By combining multiple bandpass noise-shaped channels in parallel, such that each noise-shaping/quantization circuit minimizes conversion noise in a particular region of the converter's Nyquist bandwidth, the present invention can provide a frequency interleaved or time-interleaved converter simultaneously having high resolution and high instantaneous bandwidth.
Noise-Shaping and Quantizing Considerations
In the embodiments described above, each of the noise-shaping/quantization circuits (e.g., 112 and 122) preferably is constructed differently from those shown in
A simplified block diagram of an exemplary noise-shaping/quantization circuit 112C, employing a programmable feedback-loop filter (e.g., loop filter 150) in combination with a multi-bit quantization circuit (e.g., quantizer 114), is shown in
Whereas a conventional delta-sigma (ΔΣ ) modulator with a clocking rate of fCLK, is limited by race conditions (i.e., processing of outputs within one full-rate sampling cycle) and/or circuit construction to an oversampling ratio
(i.e., where fB equals fMAX for lowpass operation), the multirate delta-sigma (μΔΣ) modulators illustrated in
where m is the polyphase decomposition factor of the μΔΣ modulator. The excess-rate oversampling ratio of μΔΣ modulator is m times greater than that of conventional time-interleaved modulators because the entire μΔΣ modulator, including its arithmetic (i.e., summation and difference) functions and quantization operations, is distributed among parallel processing paths using polyphase decomposition. In general, the circuit complexity of the μΔΣ modulator (e.g., the number of quantizers 114) increases as m2. It should be noted that although the μΔΣ modulator is a parallel processing structure, a μΔΣ modulator is different from a conventional MASH (i.e., Multi-stAge SHaping) modulator, which conventionally is sometimes referred to as a “parallel” modulator. In a conventional MASH structure, full-rate ΔΣ modulators are grouped in a parallel arrangement to increase the order P of the noise-shaped response. In contrast, the parallel μΔΣ modulator architecture increases the effective oversampling ratio N′, regardless of the order P of the noise-shaped response. It should further be noted that the μΔΣ modulator is different from a conventional, time-interleaved ΔΣ modulator in that polyphase decomposition is applied to the entire modulator circuit, rather than simply to the function of the loop filter. Therefore, the parallel, sub-rate (i.e., subsampling) outputs of the μΔΣ modulator are a function only of: 1) delayed samples of the input signal; and 2) previous output samples that have been delayed by a least one sub-rate delay of m/fS, where m is the number of parallel μΔΣ modulator outputs (i.e., the polyphase decomposition factor is m).
In an exemplary μΔΣ modulator, as most clearly illustrated in
for the same oversampling ratio N; or 2) can achieve twice the oversampling ratio (i.e., 2·N) for the same clocking rate of fCLK.
Generally speaking, in reference to converter 112C of
Feedback-loop filter 150 of μΔΣ modulator 112C, introduces frequency-dependent delaying and frequency-dependent amplitude variation to feedback signal 145, such that the noise transfer function (NTF) of the μΔΣ modulator has a bandstop response with a null at a predetermined frequency (i.e., a frequency determined by feedback-loop filter parameter ρ). In the present embodiment, feedback-loop filter 150 uses multiplier 118, adder 119 and delay register 111A to produce a frequency response with the correct amount of frequency-dependent delaying and frequency-dependent amplitude variation. As will be readily appreciated, multiplier 118 can be replaced by a combination of shift and add components to potentially reduce feedback-loop filter complexity, especially for the case where the feedback-loop filter parameter ρ can be represented by a small number of digital bits (i.e., ρ's binary representation contains few terms). The term “adder”, as used herein, is intended to refer to one or more circuits for combining two or more signals together, e.g., through arithmetic addition and/or (by simply including an inverter) through subtraction. The term “additively combine” or any variation thereof, as used herein, is intended to mean arithmetic addition or subtraction, it being understood that addition and subtraction generally are interchangeable through the use of signal inversion.
As illustrated in
Like conventional ΔΣ modulators, the μΔΣ modulator processes input signal 102 with one transfer function (STF) and the conversion noise (e.g., from quantizer 114 in reference to
NTF(z)=1+H(z).
Therefore, the signal response is all-pass and the noise response depends on the loop filter transfer function, H(z), of the μΔΣ modulator. To produce quantization noise nulls at predetermined frequencies across the Nyquist bandwidth of the converter, the feedback-loop filter 150 preferably has a second-order transfer function of the form
H(z)=ρ·z−1+z−2,
where ρ is a programmable value. Then, the noise transfer function is given by
NTF(z)=1+H(z)=1+ρ·z−1+z−2
and the location of the noise minimum is determined by the coefficient ρ. To produce a noise minimum at an arbitrary frequency within the operating bandwidth of the overall converter, it is preferable for ρ to be capable of varying over a range of −2 to +2. Specifically, a value of
ρ=−2·cos(2·π·f/(m·fCLK)),
produces a noise minimum, or null, at a frequency equal to f (i.e., the center frequency of a given processing branch), where fCLK is the quantizer/modulator clocking rate. In the absence of quantization noise (i.e., εQ=0) and input signal (i.e., x=0), the output 142A (y1) of the sampling/quantization circuit is
y1=εM·(ρ·z−1+z−2),
and the output 142B (y2) of the nonlinear bit-mapping component is
y2=y1+εD=εM·(ρ·z−1+z−2)+εD,
where: 1) εM is the intentional nonlinear distortion introduced by nonlinear bit-mapping component 161; and 2) εD is the unintentional nonlinear distortion introduced by multi-bit-to-variable-level converter 113B. When the nonlinear distortion introduced by nonlinear bit-mapping component 161 is equal to the nonlinear distortion introduced by multi-bit-to-variable-level converter 113B, such that εM=εD, then the overall distortion transfer (DTF=y2/ε) is
DTF(z)=1+ρ·z−1+z−2=NTF(z),
and therefore, distortion (εD) is subjected to the same noise-shaped response as quantization noise (εQ).
The effective oversampling ratio of an MBO converter, according to the preferred embodiments of the invention, is given by the product of the interleaved oversampling ratio M equal to the number of parallel processing branches, and the excess-rate oversampling ratio N′, equal to
Therefore, the resolution performance of an MBO converter can be increased independently of N′ by increasing the number M of parallel processing branches (e.g., branch 110 or 120). Furthermore, the excess-rate oversampling ratio N′ can be increased independently of the clocking rate fCLK, by increasing the order m of the polyphase decomposition (i.e., the number of parallel outputs of the μΔΣ modulator). However, processing branches are added at the expense of increasing the number of analog bandpass filters (e.g., filters 115 and 125) in the filter bank that performs output signal reconstruction, while simultaneously increasing the minimum quality factor (Q=fC/BW3 dB) of each such filter. Problems with controlling the amplitude and phase distortion of the filter-bank, coupled with the design complexities associated with building multiple high-Q analog filters, generally makes increasing the interleave factor, M, a less desirable alternative for increasing the effective oversampling ratio of the converter, than increasing the excess-rate oversampling ratio, N′. Therefore, the MBO converter preferably has an excess-rate oversampling ratio N′>1.
Conventionally, increasing the excess-rate oversampling ratio N is realized by increasing the clocking rate (fCLK) of the noise-shaping modulator. As mentioned previously, however, the effective excess-rate oversampling ratio N′ of a μΔΣ modulator is not limited by fCLK due to the multirate (i.e., polyphase) decomposition of the entire μΔΣ modulator circuit. Polyphase decomposition of the entire μΔΣ modulator into parallel paths eliminates race conditions and allows the effective sampling rate (fS) of the converter to increase without increasing the clocking rate (fCLK) of the modulator, at the expense of additional circuitry (i.e., at the expense of addition arithmetic and quantization/rounding operations). For illustrative purposes, consider a noise-shaping/quantization circuit 112C as illustrated in
H(z)=ρ+z−1
and
NTF(z)=1+ρ·z−1+z−2.
The quantized output 142 of noise-shaping/quantization circuit 112C, Q(yn), can be represented by the difference equation
and therefore, the difference equations for the first two output samples (i.e., n=0, 1) are
and
Substitution of y0 into y1 results in
which can be generalized to
The above equation differs from the equation in the '079 application in that the last two terms (i.e., Q(yn−2) and yn−2), which appear in the preceding equations for y0 into y1, were inadvertently excluded from the final result in the '079 application (i.e., an error was made substituting y0 into y1). Also, the block diagram of
for the same oversampling ratio N, or at twice the oversampling ratio for the same clocking rate. This novel polyphase decomposition approach, described above for a polyphase decomposition factor of m=2, can be extended to higher polyphase decomposition factors and to arbitrary feedback-loop filter functions (H(z)). This is an important consideration, particularly for converters that operate at a high sampling rate.
Each of the μΔΣ circuits shown in
Typically, the coefficients (or parameters) ρ0, ρ1, and ρ2 of the noise-shaping circuit are equal, such that the zeros of the noise transfer function occur at a common frequency. In the case of roots having equal magnitudes, the resulting noise transfer function simplifies to
H1(z)=HNOISE(z)−1=(1−ρ·z−1+z−2)3−1=−3·ρ·z−1+6·ρ2·z−2−7·ρ3·z−3+6·ρ2·z−4−3·ρ·z−5+z−6.
However, this simplified condition is not necessarily optimal with respect to minimizing output noise, particularly for small interleave factors (M) where there is a correspondingly small number of analog output filters. Conventionally, a high-order modulator is said to be “zero-optimized” when output noise is minimized by employing a NTF with unequal zeros. See K. Chao, S. Nadeem, W. Lee, and C. Sodini, “A Higher Order Topology for Interpolative Modulators for Oversampling A/D Converters,” IEEE Transactions on Circuits and Systems, 1990. A zero-optimized NTF enables the bandwidth of the NTF bandstop response to be increased at the expense of reducing the depth of the noise null. For small interleave factors M, this difference in noise response can result in improved converter resolution.
For the μΔΣ modulator, however, a NTF with unequal zeros also can reduce the circuit complexity associated with the multirate architecture. When feedback structures, such as μΔΣ modulators, are implemented using parallel-processing methods, such as polyphase decomposition, coefficient dynamic range expansion can reduce digital precision and cause the NTF response to deviate from the preferred NTF response. This occurs because in polyphase feedback structures, input and output values are multiplied by the same coefficient (i.e., ρ) multiple times, causing needed arithmetic precision to grow geometrically. A large number of binary terms (i.e., large bit-widths) is needed to represent values with high precision. This resulting increase in complexity can be offset by approximating an optimal NTF with an NTF that has unequal zeros, and has rational coefficients which can be represented as simple binary fractions (i.e., fractions with denominators that are powers of two). In signal processing applications, the technique of approximating high-precision values with the sum of binary fractions is conventionally referred to as canonic-signed-digit (CSD) representation (see for example Pham 2008). Use of coefficients that can be represented by simple binary fractions (e.g., values represented by no more than 3-8 bits), allows the multipliers comprising the loop filter of the μΔΣ modulator to be replaced by less complex circuits consisting of adders and/or bit-shifting operations. This complexity-reduction technique is sometimes referred to herein as “bit-optimization”. Therefore, in the preferred embodiments of the invention, μΔΣ modulators with a bit-optimized NTF are employed. It should be noted that zero-optimization for the purpose of reducing complexity (i.e., bit-optimization) is different from zero-optimization for the purpose of noise reduction. However, sometimes bit-optimization can result in NTFs having beneficial responses compared to NTFs with equal zeros.
Due to faster accumulation of quantization errors caused by greater amplification of quantization noise in out-of-band regions, when using noise-shaped quantization circuits of high-order, it is preferable to use greater than single-bit quantization to ensure that the noise-shaped output remains bounded. As a result, quantizer 114 shown in
Bandpass (Signal Reconstruction) Filter Considerations
The primary considerations for the bandpass filters (e.g., filters 115 and 125) used in MBO signal reconstruction according to the preferred embodiments of the present invention are: 1) design complexity (preferably expressed in terms of filter quality factor and order); 2) frequency response (particularly stopband attenuation); and 3) amplitude and phase distortion. With regard to quantization noise attenuation and conversion resolution, the best performance is obtained for output filters (i.e., bandpass or signal reconstruction filters 115) having frequency responses that exhibit high stopband attenuation, which generally increases with increasing filter order. To minimize complexity, however, the implementation of the analog filters preferably is based on relatively low-order (i.e., 5th to 7th order) standard analog filter responses (e.g., Butterworth, Chebychev, and coupled-resonator), rather than on direct transformation (e.g., impulse invariance and bilinear transformations) of the FIR window filters used in MBΔΣ analog-to-digital converters. In addition, it is preferable that the filter responses introduce as little amplitude and group delay (phase) distortion as possible to minimize the complexity of circuits that can equalize the distortion, such as digital pre-distortion linearizer (e.g., DPL 104A&B). The performance improvement realized by increasing the converter interleave factor (M) is contingent on a proportionate increase in the quality factor of the reconstruction filters, defined as the ratio of the filter center frequency to the filter 3 dB bandwidth (i.e. fC/f3 dB). For an MBO converter, according to the preferred embodiments of the invention, the limiting quality factor is the one calculated for the highest-frequency filter in the reconstruction filter bank
Therefore, the preferred quality factor for the analog filters (e.g., filters 115 and 125) is directly related to the interleave factor of the converter and, more preferably, is equal to M
Conventionally, the quality factor for standard lumped-element or distributed-element analog filters is limited to about 30. As a result, a practical limitation on the interleave factor for the MBO converter is a typical value of M≈32. However, because of the complexity associated with an analog reconstruction filter bank comprised of 32 filters, the preferred embodiments of the invention limit the interleave factor to M=16 or less (i.e., a bank of 16 or fewer analog filters 115).
For an interleave factor of M=16, 5th- to 7th-order Butterworth filter responses (i.e., with a response given by Fk(jω)) provide sufficient stopband attenuation of conversion noise. However, the overall response, F(jω)=ΣFk(jω), of a bank of these analog filters does not exhibit the properties necessary for perfect signal reconstruction in frequency-interleaved applications, namely low amplitude and group delay (phase) distortion. For example, curve 90 labeled “No Predistortion Response” in
where L(z) is a physically realizable transfer function (e.g., stable and causal). This second filter with transfer function L(z) intentionally predistorts input signal 102 with added phase and/or amplitude distortion, such that the added intentional distortion cancels the unintentional distortion of the analog reconstruction filter bank (i.e., the aggregate distortion across all of bandpass filters 115, 125, etc.). As represented in the equation above, the transfer function L(z) of DPL 104A preferably employs both feed-forward and feedback components (preferably simple weighted delay components), represented by coefficients βi and αi, respectively.
The coefficients, βi and αi, for a fixed pre-distortion linearizer (e.g., DPL 104A) that maximally equalizes the impulse response of a particular analog filter bank, can be determined using conventional methods for solving simultaneous linear equations (e.g., zero-forcing or minimum mean square error solutions), or can be determined using conventional adaptive techniques, such as those based on a least mean squares (LMS) principle. Under conditions where the overall response of the analog filter bank (i.e., the filter bank comprised of analog bandpass filters 115, 125, and the filters in the remainder of the processing branches) varies, for example due to temperature or other environmental conditions, the coefficients of DPL 104B are variable and preferably continuously adapt based on the measured amplitude and phase characteristics of the data converter output. Converter 140 of
where the * superscript represents complex conjugate and j is equal to √{square root over (−1)}. Input spectrum analyzer 141A computes the 2·K-point, discrete Fourier transform (DFT) of real input signal 102, at frequency points k=0, . . . , K−1, using: 1) multipliers 153A; 2) cosine sequence 152A and sine sequence 152B, both having an angular frequency of ωk; 3) moving-average filters 148; and 4) downsample-by-K functions 143. Output spectrum analyzer 141B performs similar processing on output signal 135. In the preferred embodiments, moving average-filter 148 is single-stage, K-point rectangular window filter, but more preferably, the magnitude of the DFT side lobes is reduced using cascaded moving-average filters of the type described in U.S. Pat. No. 8,089,382, titled “Sampling/Quantization Converters”, which is incorporated by reference herein as though set forth herein in full. Furthermore, to minimize residual output amplitude and phase distortion at the output of converter 140, DPL 104B has an impulse response of length K≧2·M (i.e., K coefficients) in the preferred embodiments, where M is the number of MBO processing branches. In applications where higher power dissipation and circuit complexity are tolerable, DPL 104B preferably has an impulse response of length K≧4·M.
As illustrated in
Reduced analog filter bank complexity is one reason why the preferred embodiments of the invention employ one or more pre-distortion linearizing filters (e.g., DPL 104A&B). A second reason is that linearizers of this type can be employed to correct signal skew caused by propagation delay differences between converter branches or channels (e.g., branches 110 and 120), and between parallel paths in configurations employing polyphase noise-shaping.
To reduce the complexity of the digital pre-distortion linearizer (e.g., DPL 104A&B), or to allow the DPL to be eliminated in certain applications which are less sensitive to amplitude and phase distortion, the responses for the bandpass filters (e.g., filters 115 and 125) that make up the analog filter bank preferably are selected to minimize the amplitude and phase distortion which produce passband variation and group delay variation (phase dispersion), respectively. To minimize amplitude and phase distortion in the preferred embodiments, individual analog filter bank responses preferably are optimized with respect to: 1) frequency response, 2) filter order, 3) center frequency, and/or 4) bandwidth. For example, a conventional analog filter bank comprised of 5th-order Butterworth filters having uniformly distributed center frequencies (i.e., center frequencies distributed evenly across the converter Nyquist bandwidth) and equal bandwidths, has a frequency response with magnitude 92, shown in
Polyphase decomposition techniques can be applied to the digital pre-distortion linearizer (DPL) to form a parallel processing structure and reduce the clocking rates of the digital multipliers and adders that are used to implement the DPL. For example, fixed DPL 104A preferably is a recursive (i.e., infinite-impulse response or IIR) structure with transfer function L(z), which performs the discrete-time convolution of the data converter input sequence x(n) and the filter coefficients l(n) according to
y(n)=x(n)*l(n)Y(z)=X(z)·L(z)=X·L.
Assuming, without loss of generality, a pre-distortion linearizer with three coefficients (i.e., β0, β1, and α1) and transfer function
the operation of the pre-distortion linearizer can be represented by the difference equation
yn=β0xn+β1xn−1−α1yn−1.
Therefore, the difference equations for the first two output samples (i.e., n=1, 2) are
y2=β0x2+β1x1−α1y1
and
y1=β0x1+β1x0−α1y0,
and substitution of y1 into y2 results in
The above equation can be generalized to
yn=β0xn+(β1−α1β0)xn−1−α1β1xn−2+α12yn−2.
The above equation differs from the equation in the '079 application, in that the coefficient of the last term is now raised to a power of 2 (i.e., α12yn−2), correcting for an error that was made substituting y1 into y2 in the '079 application. As before, however, it can be noted that yn only depends on inputs and every other output for the above example, demonstrating that, like the μΔΣ modulator, the digital pre-distortion linearizer function can be implemented as a parallel processing structure with two parallel paths (i.e., with a polyphase decomposition factor of m=2). In the above example, parallel processing enables the DPL clocking rate fCLK to be one-half and effective sampling rate fS, such that
Through conventional methods for factoring the denominator of the linearizer transfer function, this polyphase decomposition approach can be extended to higher polyphase decomposition factors (i.e., m>2) and arbitrary DPL transfer functions (L(z)), including transfer functions with only numerator terms (i.e., finite impulse response), to allow the DPL to run at a sub-multiple of the effective sampling rate of the converter. Polyphase decomposition into parallel paths results in an m-times reduction in clocking rate at the expense of no greater than an m-times increase in circuit complexity. This penalty in circuit complexity is generally a favorable alternative in the case of converters that operate at very high sampling rates.
Multi-Bit-To-Variable-Level Signal Converter Considerations
In the preferred embodiments of the invention, the binary weighted outputs of the noise-shaping/quantization circuits (e.g., circuits 112A&112B which are collectively referred to as circuit 112 herein), shown in
More specifically, the preferred embodiments of the invention use an R-2R ladder network that has been modified for bipolar operation, where R is matched to the characteristic impedance of analog filter 115. This impedance is generally between 50 ohms and 100 ohms.
An important consideration for the resistor ladder network is the relative matching of the constituent resistive elements. It is conventionally understood that a perfect resistor ladder creates an analog output by weighting each digital input according to a binary scaling factor. Mismatches in the resistive elements of the ladder distort this binary scaling, producing a nonlinear response. This nonlinear response distorts the output waveform and, therefore, degrades the quality of the converted analog signal. In conventional converters that employ resistive ladder networks, the matching requirement (εD) for the resistive elements is determined by the converter precision according to
where B in the above equation is the effective resolution of the converter in bits. Therefore, the resistor ladder matching is ˜0.2% for 8-bit effective resolution.
The oversampled operation of an MBO converter according to the preferred embodiments of the invention, affords two advantages over conventional converters that are based on resistor ladder networks. One advantage is that because of noise-shaped quantization and filtering, oversampled converters require resistor ladders with fewer inputs to achieve the same effective resolution as Nyquist-rate converters. Thus, oversampling reduces the overall complexity of the resistor ladder network. The reduction in the number of resistor ladder inputs is a function of: 1) the converter effective oversampling ratio (N″·M); 2) the noise-shaping order (P) of the μΔΣ modulators within the noise-shaping/quantization circuits (e.g., circuit 112); and 3) the stopband attenuation of the signal reconstruction filters (e.g., bandpass filter 115). To reduce resistor network complexity (i.e., the number of discrete resistor elements and the number of input lines), the preferred embodiment of the invention uses resistor ladder networks with eight or fewer inputs (i.e., eight or fewer digital inputs to the resistor ladder network in each processing branch).
A second and more significant advantage is that oversampling enables the distortion introduced by mismatches, and other imperfections such as the signal amplitude-dependent gain (i.e., buffer amplifier compression in resistor ladder 113C of converter 200C), to be shaped by noise-shaping/quantization circuit 112 and then largely removed by bandpass filter 115. Such distortion shaping and removal is preferably realized through the inclusion of nonlinear bit-mapping, e.g., as illustrated in the representative embodiment of converter 110B in
(e.g., a Taylor's series). In
Applying relatively high-resolution weighting factors to each such bit output from quantizer 114, prior to feeding signal 146B back to adder 116 through feedback-loop filter 150, makes it possible to more accurately match the binary scaling imperfections of the resistor ladder network (or other multi-bit-to-variable-level signal converter). More precisely, the nonlinear bit-mapping coefficients, C0 . . . Cn−1, shown in
In practice, the nonlinear bit-mapping coefficients C0 . . . Cn−1 preferably are calibrated once upon startup (e.g., using a known signal) and then are dynamically adjusted in real time in order to account for drift in resistance values (e.g., due to thermal changes). In the preferred embodiments, such dynamic adjustments are made on the order of once per second so as to allow for a sufficient amount of time to evaluate the effect of any changes.
Although not shown in
For a conventional ladder-based converter, the matching accuracy of the resistors in the ladder network determines the precision of the converter. In contrast, the precision of the preferred MBO converter is a function of the converter oversampling ratio (N′·M), the order (P) of the noise-shaped response, and the stopband attenuation of the reconstruction filters. Therefore, oversampling enables high-accuracy converters to be implemented using low-accuracy resistor ladder networks. The preferred embodiments of the invention use resistor ladder networks with accuracies of just 1%, or better, to reduce the tuning range of the nonlinear bit-mapping components.
Overall Converter Considerations
The noise-shaping/quantization operation of the MBO converter is most effective when the spectral null in the noise transfer function (NTF) is precisely aligned with the center frequency of the bandpass filter in a corresponding processing branch. When the NTF spectral null and bandpass filter center frequency are precisely aligned, the noise level, and therefore the signal-plus-noise level, at the bandpass filter output is a minimum. Because the spectral null in the NTF response is determined by parameters ρi of feedback-loop filter 150, the configuration illustrated in
Because the digital pre-distortion linearizer (e.g., DPL 104A&B) and the μΔΣ modulators within the noise-shaping/quantization circuits (e.g., circuit 112) can be implemented as multirate (polyphase) structures, the instantaneous bandwidth of the converter technology illustrated in
Although the foregoing MBO converter has up to 10 GHz of instantaneous bandwidth at effective sampling rates fS of 20 GHz (i.e., a frequency range of 0 Hz to 10 GHz in the preferred embodiments), inclusion of conventional upconversion techniques should be considered within the scope of the invention as a means of shifting the converter output to frequency bands that exceed the Nyquist limit of
For example, an output signal can be shifted from a band centered at 5 GHz to a band centered at 15 GHz, using a conventional upconverter with a 10 GHz local oscillator (LO), such that the resulting 15 GHz output signal can be converted with an MBO processing branch configured for 5 GHz operation (i.e., the quantization noise response is configured for a spectral null at 5 GHz). An exemplary converter 100A shown in
z=y′inphase·cos(ωkt31y′quadrature·sin(ωkt),
where y′inphase and y′quadrature are formed within quadrature combiner 309, and represent phase-shifted versions of the quantized output of the noise-shaping/quantization circuit. In addition to quadrature combiner 309, each quadrature upconverter consists of: 1) a local oscillator source with frequencies ω0 and ωm (e.g., frequencies which generate each of signals 306A&B, respectively); 2) a quadrature hybrid (e.g., each of hybrid splitters 307) that divides the local oscillator signal into quadrature (i.e., sine) and in-phase (i.e., cosine) components; and 3) dual mixers (e.g., mixers 308A&B) that produce frequency-shifted images of the quantized output from the noise-shaping/quantization circuit. In the preferred embodiments, a quadrature upconverter (i.e., image reject mixer) is used instead of a simple upconverter (i.e., single mixer), because a simple upconverter produces unwanted lower images of the quantized signal (i.e., ω−ω0 and ω−ωm), in addition to the desired upper images of the quantized signal (i.e., ω+ω0 and ω+ωm)
The present inventor has discovered that in addition to extending a usable frequency range, output quadrature upconverters can be combined with input quadrature downconverters, as illustrated in
An exemplary MBO converter 100B, shown in
The quadrature downconverter produces an in-phase output (yinphase) and a quadrature output (yquadrature) by processing input signal 106 (x) according to:
yinphase=x·cos(ωt)
yquadrature=x·A·sin(ωt+θ),
where parameters A and θ preferably are set (e.g., pursuant to a manufacturing trim operation), or dynamically adjusted, to compensate for amplitude and phase imbalances, respectively, in the quadrature upconverter (e.g., circuits 305A&B). Upconverter amplitude and phase imbalances produce unwanted spurious responses at the output of the reconstruction filter (e.g., each of filters 115 and 125), that get smaller when parameters A and θ are matched (i.e., equal and opposite) to the inherent imbalances of quadrature upconverter 305. Preferably, the parameter A is approximately equal, or more preferably exactly equal, to the multiplicative inverse of the amplitude imbalance of the quadrature upconverter. Similarly, the parameter θ preferably is approximately equal, or more preferably exactly equal, to the additive inverse of the phase imbalance of the quadrature upconverter. Similarly to the digital pre-distortion linearizer (DPL), the quadrature downconverter can be implemented using polyphase decomposition techniques to reduce the clocking/processing rates of digital multipliers and sine/cosine sequence generators.
Exemplary block diagrams of MBO converters according to the preferred embodiments of the invention are illustrated in
and are combined into a single output with an effective sampling rate of fS, using a novel moving-average summation operation (e.g., circuit element 179) which requires no upsampling (i.e., the moving-average summation process requires no upsampling from a sub-rate of
to a full-rate of fS). In contrast, a conventional converter implementation, such as circuit 80C shown in
to a full-rate of fS, and then concatenated to form a single digital output (i.e., an operation which can be functionally represented as upsampling, delaying, and summing). Although the conventional multiplexer has an all-pass response, upsampling requires the multiplexing circuitry to switch at the full sampling rate (i.e., instead of the modulator's subsampling rate), and consequently, limits the effective excess-rate oversampling ratio N of the overall converter. The moving-average summation (i.e., parallel-to-serial reformatting) operation of the preferred embodiments, however, combines the multiple sub-rate outputs
of a parallel ΔΣ modulator (i.e., a μΔΣ modulator) into a single full-rate output (e.g., a single output at a rate of fS), and, e.g., can be limited to including: 1) a plurality of delay elements coupled to the sub-rate outputs, each of which introduces a different time-offset in increments of ΔΦ=1/fS (e.g., within circuit elements 178A-C), using for example, phase offset resampling at a sub-rate of
(i.e., latches or flip-flops that are registered on m different phases of the sub-rate clock) and/or conventional passive or active delay lines; and 2) a signal combiner that sums (e.g., within analog adders 177A&B) the time-offset signals which are provided by the delay elements, and which reflect sub-rate sampling of
Therefore, the circuitry comprising the preferred combining operation switches at a subsampled rate
and for a constant switching speed, the excess-rate oversampling ratio of the preferred converter is m times higher than that of a conventional oversampling converter.
A more generalized depiction of a converter, which utilizes moving-average summation according to the preferred embodiments of the preset invention, is converter 95B illustrated in
relative to effective full-rate (fS) sampling at the output of the converter (e.g., analog output 135). The sub-rate samples on each of the m parallel outputs (e.g., outputs 108C-E) preferably are a sequence of values representing different subsampling phases of the underlying (complete) input signal (e.g., input signal 103). According to different preferred embodiments of the present invention, parallel signal processor 107B performs different functions. For example, in the embodiment of converter 95A in
where
and fS is the effective sample rate of the overall converter). In an alternate embodiment, parallel signal processor 107B performs only serial-to-parallel demultiplexing to transform an input sequence comprising relatively high-resolution, low-rate samples in serial format (e.g., multi-bit samples x on line 103 with a rate of fCLK), to an output sequence comprising relatively high-resolution, low-rate samples in parallel format (e.g., each output yi on one of the lines 108C-E provides multi-bit samples at a rate of
where
and fS=fCLK is the effective sample rate of the overall apparatus). Unlike the multi-bit-to-variable-level signal converters included within converter 95A (e.g., converters 113A) which operate at a sampling rate of
the multibit-to-variable-level signal converters of converter 95B (e.g., converters 113C) operate at a potentially lower rate of
In still other embodiments, parallel signal processor 107B combines serial-to-parallel demultiplexing with other signal processing operations, such as: 1) pre-emphasis filtering for equalization and/or sin(x)/x or other (e.g., similar) correction; 2) signal companding for dynamic range reduction; and/or 3) estimation and mitigation of sampling clock imperfections, such as jitter and skew. For embodiments employing estimation and mitigation of sampling clock imperfections, the methods described in the '284 Application are preferred. The outputs of the parallel signal processor (e.g., output 108C-E) are combined using a moving-average summation operation comprising: 1) delay with multi-bit-to-variable-level conversion (e.g., within delay paths 188A-C); and 2) continuous-time summation (e.g., within adders 177A&B).
In embodiments where parallel signal processor 107B estimates and mitigates sampling skew (i.e., a condition where the sub-rate clocks are offset in time by increments which do not equal exact multiples of the full-rate clock period ΔΦ=1/fS), a circuit similar to that illustrated in
and m is the number of parallel outputs generated by parallel signal processor 107B. Referring to
The process of moving-average summation is depicted in the timing diagram given in
are combined to produce a resultant signal with transitions that reflect full-rate sampling (i.e., switching at an effective rate of fS). It can be shown that phase-offset resampling and summing (i.e., moving-average summation), according to the preferred embodiments, introduces what is conventionally referred to as a moving-average filter response, which has a continuous-time transfer function given by
where: 1) m is the polyphase decomposition factor equal to the number of multirate outputs from the parallel processor (e.g., processor 107B); and 2) ΔΦ=1/fS is the incremental time (i.e., clock phase) offset associated with the resampling clocks. The above transfer function produces a lowpass response with a sin(x)/x or sinc(x) shape and a 3 dB cutoff frequency of approximately 1/(2·m·ΔΦ), which without compensation, limits the instantaneous bandwidth of the overall converter to fS/(2·m). The magnitude versus frequency response of the moving-average summation operation is given in
In each of
and the outputs of the latches are respectively offset in by time increments of ΔΦ=1/fS, where m is the polyphase decomposition factor of the μΔΣ modulator (i.e., m=2 for exemplary circuit 170B of
Accordingly, the output of the adder in each summation circuit 176A-C represents a full-rate signal with a sampling rate of fS. Therefore, to reduce the switching speed of the digital output logic, the adder preferably is implemented as an analog (i.e., continuous-time) adder, using for example, resistive or reactive combiner networks (e.g., Wye splitters, Wilkinson combiners).
A structure that is similar to that of exemplary converter 170B (i.e., shown in
The purpose of IMA filters 174A-C is to compensate for the sin(x)/x response introduced by each moving-average summation circuit 176A-C (or the corresponding summation structure shown in
As discussed above, using moving-average summation to combine the multirate outputs of noise-shaping/quantization circuits 112A-C, as illustrated in
where m is the polyphase decomposition factor, equal to the number of multirate outputs from each noise-shaping/quantization circuit 112A-C (i.e., m=2 in
For a polyphase decomposition factor of m=4, an IMA filter has the frequency response illustrated in
where fS is the effective sampling rate of the overall converter. As a result, an IMA filter with finite gain (i.e., finite word lengths for a digital filter), cannot perfectly compensate for the nulls produced at input frequencies near
by the moving-average summation operation at the output of each noise-shaping/quantization circuit. In general, the moving-average response produces such spectral nulls at frequencies equal to
This means that for a polyphase decomposition factor of m=2, the spectral null in the moving-average response occurs at the Nyquist frequency, which can be eliminated from the MBO converter output with little or no consequence in terms of overall converter bandwidth. In applications where the maximum frequency at the input of the converter exceeds fS/(2·m), therefore, combining the multirate outputs of each noise-shaping/quantization circuit 112A-C using moving-average summation, is preferable only for combining up to two multirate outputs (i.e., m=2).
As illustrated in
Several of the embodiments described above incorporate both IMA filters (e.g., filters 174A-C in
The instantaneous bandwidth of the MBO converter technology (e.g., as shown in
As noted previously, however, the resolution performance of MBO converters 200A-C (collectively referred to as converter 200 herein) is not limited by the effective sampling rate fS, because the resolution is also a function of the interleave factor (i.e., the number of parallel processing branches M), the order P of the noise-shaped quantization, and the properties of the bandpass (reconstruction) filter. In addition, like conventional oversampling converters, the MBO converter technology can be implemented so as to be relatively insensitive to impairments such as sampling jitter and thermal noise that degrade the performance of other high-speed converter architectures. Specifically, impairments such as quantizer thermal noise can be made subject to a noise-shaped response in a similar manner to quantization noise, exhibiting a frequency response that enables significant attenuation by the analog bandpass (reconstruction) filters (e.g., filters 115 and 125).
Simulated resolution performance results for the MBO converter 200 are given in Table 1 for a noise-shaped response of 6th-order, for various interleave factors M, and for analog reconstruction filters of various order.
System Environment
Generally speaking, except where clearly indicated otherwise, all of the systems, methods, functionality and techniques described herein can be practiced with the use of one or more programmable general-purpose computing devices. Such devices (e.g., including any of the electronic devices mentioned herein) typically will include, for example, at least some of the following components coupled to each other, e.g., via a common bus: (1) one or more central processing units (CPUs); (2) read-only memory (ROM); (3) random access memory (RAM); (4) other integrated or attached storage devices; (5) input/output software and circuitry for interfacing with other devices (e.g., using a hardwired connection, such as a serial port, a parallel port, a USB connection or a FireWire connection, or using a wireless protocol, such as radio-frequency identification (RFID), any other near-field communication (NFC) protocol, Bluetooth or a 802.11 protocol); (6) software and circuitry for connecting to one or more networks, e.g., using a hardwired connection such as an Ethernet card or a wireless protocol, such as code division multiple access (CDMA), global system for mobile communications (GSM), Bluetooth, a 802.11 protocol, or any other cellular-based or non-cellular-based system, which networks, in turn, in many embodiments of the invention, connect to the Internet or to any other networks; (7) a display (such as a cathode ray tube display, a liquid crystal display, an organic light-emitting display, a polymeric light-emitting display or any other thin-film display); (8) other output devices (such as one or more speakers, a headphone set, a laser or other light projector and/or a printer); (9) one or more input devices (such as a mouse, one or more physical switches or variable controls, a touchpad, tablet, touch-sensitive display or other pointing device, a keyboard, a keypad, a microphone and/or a camera or scanner); (10) a mass storage unit (such as a hard disk drive or a solid-state drive); (11) a real-time clock; (12) a removable storage read/write device (such as a flash drive, any other portable drive that utilizes semiconductor memory, a magnetic disk, a magnetic tape, an opto-magnetic disk, an optical disk, or the like); and/or (13) a modem (e.g., for sending faxes or for connecting to the Internet or to any other computer network). In operation, the process steps to implement the above methods and functionality, to the extent performed by such a general-purpose computer, typically initially are stored in mass storage (e.g., a hard disk or solid-state drive), are downloaded into RAM, and then are executed by the CPU out of RAM. However, in some cases the process steps initially are stored in RAM or ROM and/or are directly executed out of mass storage.
Suitable general-purpose programmable devices for use in implementing the present invention may be obtained from various vendors. In the various embodiments, different types of devices are used depending upon the size and complexity of the tasks. Such devices can include, e.g., mainframe computers, multiprocessor computers, one or more server boxes, workstations, personal (e.g., desktop, laptop, tablet or slate) computers and/or even smaller computers, such as personal digital assistants (PDAs), wireless telephones (e.g., smartphones) or any other programmable appliance or device, whether stand-alone, hard-wired into a network or wirelessly connected to a network.
In addition, although general-purpose programmable devices have been described above, in alternate embodiments one or more special-purpose processors or computers instead (or in addition) are used. In general, it should be noted that, except as expressly noted otherwise, any of the functionality described above can be implemented by a general-purpose processor executing software and/or firmware, by dedicated (e.g., logic-based) hardware, or any combination of these approaches, with the particular implementation being selected based on known engineering tradeoffs. More specifically, where any process and/or functionality described above is implemented in a fixed, predetermined and/or logical manner, it can be accomplished by a processor executing programming (e.g., software or firmware), an appropriate arrangement of logic components (hardware), or any combination of the two, as will be readily appreciated by those skilled in the art. In other words, it is well-understood how to convert logical and/or arithmetic operations into instructions for performing such operations within a processor and/or into logic gate configurations for performing such operations; in fact, compilers typically are available for both kinds of conversions.
It should be understood that the present invention also relates to machine-readable tangible (or non-transitory) media on which are stored software or firmware program instructions (i.e., computer-executable process instructions) for performing the methods and functionality of this invention. Such media include, by way of example, magnetic disks, magnetic tape, optically readable media such as CDs and DVDs, or semiconductor memory such as various types of memory cards, USB flash memory devices, solid-state drives, etc. In each case, the medium may take the form of a portable item such as a miniature disk drive or a small disk, diskette, cassette, cartridge, card, stick etc., or it may take the form of a relatively larger or less-mobile item such as a hard disk drive, ROM or RAM provided in a computer or other device. As used herein, unless clearly noted otherwise, references to computer-executable process steps stored on a computer-readable or machine-readable medium are intended to encompass situations in which such process steps are stored on a single medium, as well as situations in which such process steps are stored across multiple media.
The foregoing description primarily emphasizes electronic computers and devices. However, it should be understood that any other computing or other type of device instead may be used, such as a device utilizing any combination of electronic, optical, biological and chemical processing that is capable of performing basic logical and/or arithmetic operations.
In addition, where the present disclosure refers to a processor, computer, server, server device, computer-readable medium or other storage device, client device, or any other kind of apparatus or device, such references should be understood as encompassing the use of plural such processors, computers, servers, server devices, computer-readable media or other storage devices, client devices, or any other such apparatuses or devices, except to the extent clearly indicated otherwise. For instance, a server generally can (and often will) be implemented using a single device or a cluster of server devices (either local or geographically dispersed), e.g., with appropriate load balancing. Similarly, a server device and a client device often will cooperate in executing the process steps of a complete method, e.g., with each such device having its own storage device(s) storing a portion of such process steps and its own processor(s) executing those process steps.
As used herein, the term “coupled”, or any other form of the word, is intended to mean either directly connected or connected through one or more other elements or processing blocks.
Additional Considerations
In the preceding discussion, the terms “operators”, “operations”, “functions” and similar terms can refer to method steps or hardware components, depending upon the particular implementation/embodiment.
Unless clearly indicated to the contrary, words such as “optimal”, “optimize”, “minimize”, “best”, as well as similar words and other words and suffixes denoting comparison, in the above discussion are not used in their absolute sense. Instead, such terms ordinarily are intended to be understood in light of any other potential constraints, such as user-specified constraints and objectives, as well as cost and processing constraints.
In the event of any conflict or inconsistency between the disclosure explicitly set forth herein or in the attached drawings, on the one hand, and any materials incorporated by reference herein, on the other, the present disclosure shall take precedence. In the event of any conflict or inconsistency between the disclosures of any applications or patents incorporated by reference herein, the disclosure having the most recent priority date shall take precedence.
Several different embodiments of the present invention are described above and in the documents incorporated by reference herein, with each such embodiment described as including certain features. However, it is intended that the features described in connection with the discussion of any single embodiment are not limited to that embodiment but may be included and/or arranged in various combinations in any of the other embodiments as well, as will be understood by those skilled in the art.
In the above discussion, certain methods are explained by breaking them down into steps listed in a particular order. However, it should be noted that in each such case, except to the extent clearly indicated to the contrary or mandated by practical considerations (such as where the results from one step are necessary to perform another), the indicated order is not critical but, instead, that the described steps can be reordered and/or two or more of such steps can be performed concurrently.
References herein to a “criterion”, “multiple criteria”, “condition”, “conditions” or similar words which are intended to trigger, limit, filter or otherwise affect processing steps, other actions, the subjects of processing steps or actions, or any other activity or data, are intended to mean “one or more”, irrespective of whether the singular or the plural form has been used. For instance, any criterion or condition can include any combination (e.g., Boolean combination) of actions, events and/or occurrences (i.e., a multi-part criterion or condition).
Similarly, in the discussion above, functionality sometimes is ascribed to a particular module or component. However, functionality generally may be redistributed as desired among any different modules or components, in some cases completely obviating the need for a particular component or module and/or requiring the addition of new components or modules. The precise distribution of functionality preferably is made according to known engineering tradeoffs, with reference to the specific embodiment of the invention, as will be understood by those skilled in the art.
In the discussions above, the words “include”, “includes”, “including”, and all other forms of the word should not be understood as limiting, but rather any specific items following such words should be understood as being merely exemplary.
Thus, although the present invention has been described in detail with regard to the exemplary embodiments thereof and accompanying drawings, it should be apparent to those skilled in the art that various adaptations and modifications of the present invention may be accomplished without departing from the spirit and the scope of the invention. Accordingly, the invention is not limited to the precise embodiments shown in the drawings and described above. Rather, it is intended that all such variations not departing from the spirit of the invention are to be considered as within the scope thereof as limited solely by the claims appended hereto.
This application claims the benefit of U.S. Provisional Patent Application Ser. No. 62/103,160, filed Jan. 14, 2015, and is a continuation in part of U.S. patent application Ser. No. 14/537,122, filed on Nov. 10, 2014, which claims the benefit of U.S. Provisional Patent Application Ser. No. 61/970,846, filed Mar. 26, 2014, and is a continuation in part of U.S. patent application Ser. No. 13/647,301, filed Oct. 8, 2012 (now U.S. Pat. No. 8,896,471), which is a continuation of U.S. patent application Ser. No. 13/400,019, filed Feb. 17, 2012 (now U.S. Pat. No. 8,294,605), which: (1) claims the benefit of U.S. Provisional Patent Application Ser. Nos. 61/444,643, 61/450,617 and 61/507,568, filed on Feb. 18, 2011, Mar. 8, 2011 and Jul. 13, 2011, respectively; and (2) is a continuation in part of U.S. patent application Ser. No. 12/970,379, filed on Dec. 16, 2010 (now U.S. Pat. No. 8,264,390), which claimed the benefit of U.S. Provisional Patent Application Ser. No. 61/287,079, filed on Dec. 16, 2009. The present application also is a continuation in part of U.S. patent application Ser. No. 14/629,442, filed Feb. 23, 2015, which is a continuation in part of U.S. patent application Ser. No. 14/056,917, filed on Oct. 17, 2013 (now U.S. Pat. No. 9,000,967), which is a continuation in part of U.S. patent application Ser. No. 13/535,037, filed on Jun. 27, 2012 (now U.S. Pat. No. 8,581,768) which claimed the benefit of: U.S. Provisional Patent Application Ser. No. 61/549,739, filed on Oct. 20, 2011, and titled “Linear to Discrete Quantization Conversion with Reduced Sampling Variation Errors”; U.S. Provisional Patent Application Ser. No. 61/501,284 (the '284 Application), filed on Jun. 27, 2011; U.S. Provisional Patent Application Ser. No. 61/536,003, filed on Sep. 18, 2011; and U.S. Provisional Patent Application Ser. No. 61/554,918, filed on Nov. 2, 2011. Each of the foregoing applications is incorporated by reference herein as though set forth herein in full.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5341135 | Pearce | Aug 1994 | A |
| 5907242 | Gard | May 1999 | A |
| 6888484 | Kiss et al. | May 2005 | B2 |
| 7633417 | Yeh | Dec 2009 | B1 |
| 8155238 | Muller et al. | Apr 2012 | B2 |
| 8264390 | Pagnanelli | Sep 2012 | B2 |
| 8294605 | Pagnanelli | Oct 2012 | B1 |
| 8581768 | Pagnanelli | Nov 2013 | B1 |
| 8896471 | Pagnanelli | Nov 2014 | B1 |
| 9000967 | Pagnanelli | Apr 2015 | B2 |
| 9048865 | Pagnanelli | Jun 2015 | B2 |
| 9225353 | Pagnanelli | Dec 2015 | B2 |
| 9276602 | Pagnanelli | Mar 2016 | B1 |
| 20040208249 | Risbo et al. | Oct 2004 | A1 |
| 20040213356 | Burke | Oct 2004 | A1 |
| 20050099327 | Robinson et al. | May 2005 | A1 |
| Number | Date | Country |
|---|---|---|
| WO9215153 | Sep 1992 | WO |
| Entry |
|---|
| International search report and written opinion from corresponding PCT application Serial No. PCT/US2010/060789. |
| Eshraghi A et al: “A Comparative Analysis of Parallel Delta—Sigma ADC Architectures”, IEEE Transactions on Circuits and Systems Part I: Regular Papers, IEEE Service Center, New York, NY, US, vol. 51, No. 3, Mar. 1, 2004, pp. 450-458, XP011109280, ISSN: 1057-7122, DOI: D01:10.1109/TCSI.2004.823663. |
| Harris F J et al: “Implementation considerations and limitations for dynamic range enhanced analog to digital converters”, IEEE International Conference on Acoustics, Speech, and Signal Processing, 1989. ICASSP-89., May 3, 1989, May 23, 1989, pp. 1286-1289, XP010083173. |
| D. Anastassiou “Error Diffusion Coding for A/D Conversion,” IEEE Transactions on Circuits and Systems, vol. 36, 1989. |
| A. Petraglia and S. K. Mitra, “High Speed A/D Conversion Using QMF Banks,” Proceedings: IEEE Symposium on circuits and Systems, 1990. |
| Aziz, P., “Multi Band Sigma Delta Analog to Digital Conversion”, IEEE International Conference on Acoustics, Speech, and Signal Processing, 1994. |
| S. R. Velazquez, T. Q. Nguyen, and S. R. Broadstone, “Design of Hybrid Filter Banks for Analog/Digital Conversion,” IEEE Transactions on Signal Processing, 1998. |
| A. Fernandez-Vazquez and G. Jovanovic-Dolecek, “Design of Real and Complex Linear-Phase IIR Modified QMF Banks,” IEEE Asia Pacific Conference on Circuits and Systems, 2006. |
| P. C. R. Brandao and A. Petraglia, “A Switched-Capacitor Hadamard Filter Bank in 0.35 m CMOS,” Proceedings: 48th Midwest Symposium on Circuits and Systems, 2005. |
| P: M. M Furth and A. G. Andreou, “A Design Framework for Low Power Analog Filter Banks”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, 1995. |
| R. Khoini-Poorfard, L. B. Lim, and D. A. Johns, “Time-Interleaved Oversampling A/D Converters: Theory and Practice,” IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, 1997. |
| K. Chao, S. S. Nadeem, W. Lee, and C. Sodini, “A Higher Order Topology for Interpolative Modulators for Oversampling A/D Converters,” IEEE Transactions on Circuits and Systems, 1990. |
| Prosecution history of parent U.S. Appl. No. 13/400,019 (now U.S. Pat. No. 8,294,605). |
| J. Pham and A.C. Carusone, “A Time-Interleaved [Delta-Sigma]-DAC Architecture Clocked at the Nyquist Rate,” IEEE Transactions on Circuits and Systems II: Express Briefs, 2008. |
| Prosecution history of parent U.S. Appl. No. 13/647,301 (now U.S. Pat. No. 8,896,471). |
| Report/Translation of Office Action in corresponding Japanese Patent Application No. 2012-544832. |
| Krall, C. , et al., “Time-Interleaved Digital-to-Analog Converters for UWB Signal Generation”, IEEE International Conference on Ultra-Wideband, 2007. ICUWB 2007, pp. 366-371, ISBN: 978-1-4244-0521-3. |
| Prosecution history of parent U.S. Appl. No. 14/537,122 (now U.S. Pat. No. 9,048,865). |
| Prosecution history of parent U.S. Appl. No. 14/056,917 (now U.S. Pat. No. 9,000,967). |
| Prosecution history of parent U.S. Appl. No. 12/970,379 (now U.S. Pat. No. 8,264,390). |
| Prosecution history of parent U.S. Appl. No. 13/535,037 (now U.S. Pat. No. 8,581,768). |
| Prosecution history of parent U.S. Appl. No. 114/629,442 (now U.S. Pat. No. 9,225,353). |
| Prosecution history of parent U.S. Appl. No. 14/697,574. |
| Office Action dated Aug. 17, 2015, from European Patent Office in related European application EP 10798430.4. |
| Number | Date | Country | |
|---|---|---|---|
| 61970846 | Mar 2014 | US | |
| 61444643 | Feb 2011 | US | |
| 61450617 | Mar 2011 | US | |
| 61507568 | Jul 2011 | US | |
| 61287079 | Dec 2009 | US | |
| 61554918 | Nov 2011 | US | |
| 61501284 | Jun 2011 | US | |
| 61549739 | Oct 2011 | US | |
| 61536003 | Sep 2011 | US | |
| 62103160 | Jan 2015 | US | |
| 62266479 | Dec 2015 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14697574 | Apr 2015 | US |
| Child | 14997504 | US | |
| Parent | 13400019 | Feb 2012 | US |
| Child | 13647301 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14537122 | Nov 2014 | US |
| Child | 14697574 | US | |
| Parent | 13647301 | Oct 2012 | US |
| Child | 14537122 | US | |
| Parent | 12970379 | Dec 2010 | US |
| Child | 13400019 | US | |
| Parent | 14629442 | Feb 2015 | US |
| Child | 14697574 | US | |
| Parent | 14056917 | Oct 2013 | US |
| Child | 14629442 | US | |
| Parent | 13535037 | Jun 2012 | US |
| Child | 14056917 | US |
